Literature review: white blood cell classification and segmentation — research directions, comparison of representative work, research gaps and innovations

White blood cell (WBC) classification is a crucial step in assessing a patient’s health and validating medical treatment in the medical domain. Hence, efficient computer vision solutions to the classification of WBC will be an effective aid to medical practitioners. Computer-aided diagnosis (CAD) reduces manual intervention, avoids errors, speeds up medical analysis, and provides accurate medical reports. Though a lot of research has been taken up to develop deep learning models for efficient classification of WBCs, there is still scope for improvement to support the data insufficiency issue in medical data sets. Data augmentation and normalization techniques increase the quantity of data but don’t enhance the quality of the data. Hence, deep learning models though performing well can still be made efficient and effective when quality data is fused along with the available image dataset. This paper aims to utilize domain knowledge and image data to improve the classification performance of pre-trained models namely Inception V3, DenseNet 121, ResNet 50, MobileNet V2, and VGG 16. The models performance, with and without domain knowledge infused, is analyzed on the BCCD and LISC datasets. On the BCCD dataset, the average accuracies increased from 82.7%, 98.8%, 98.38%, 98.56%, and 98.5%–99.38%, 99.05%, 99.05%, 98.67%, and 98.75% for Inception V3, DenseNet 121, ResNet 50, MobileNet V2, and VGG 16, respectively. Similarly, on the LISC dataset, the accuracies improved from 86.76%, 92.2%, 91.76%, 92.8%, and 94.4%–92.05%, 95.88%, 95.58%, 95.2%, and 95.2%, respectively.

No abstract available

No abstract available

No abstract available

No abstract available

Accurate diagnosis of white blood cells from cytopathological images is a crucial step in evaluating leukaemia. In recent years, image classification methods based on fully convolutional networks have drawn extensive attention and achieved competitive performance in medical image classification. In this paper, we propose a white blood cell classification network called ResNeXt-CC for cytopathological images. First, we transform cytopathological images from the RGB color space to the HSV color space so as to precisely extract the texture features, color changes and other details of white blood cells. Second, since cell classification primarily relies on distinguishing local characteristics, we design a cross-layer deep-feature fusion module to enhance our ability to extract discriminative information. Third, the efficient attention mechanism based on the ECANet module is used to promote the feature extraction capability of cell details. Finally, we combine the modified softmax loss function and the central loss function to train the network, thereby effectively addressing the problem of class imbalance and improving the network performance. The experimental results on the C-NMC 2019 dataset show that our proposed method manifests obvious advantages over the existing classification methods, including ResNet-50, Inception-V3, Densenet121, VGG16, Cross ViT, Token-to-Token ViT, Deep ViT, and simple ViT about 5.5–20.43% accuracy, 3.6–23.56% F1-score, 3.5–25.71% AUROC and 8.1–36.98% specificity, respectively.

No abstract available

White blood cells (WBCs) are useful for diagnosing infectious diseases and infections. Machine learning and deep learning have been used to classify WBCs from blood smear images. Despite advances in machine learning, there has been little research on applying medical domain knowledge to convolutional neural networks (CNNs) to improve WBC classification. The existing models are often inaccurate, rely on manual input, and fail to incorporate external medical knowledge into decision-making. This study used the blood cell count and detection dataset which contains images of monocytes, lymphocytes, neutrophils, and eosinophils for WBC classification. In this paper, we propose a CNN model for WBC classification called WBC-KICNet (knowledge-infused convolutional neural network). The present work uses two CNN models: the first model generates the knowledge vector from input images and the domain expert (hematologist); the second model extracts deep features from the input image. A feature fusion mechanism is then used to combine these two features to classify the WBCs. Several metrics have been used to evaluate the performance of the WBC-KICNet model. These measures yielded impressive results. Accuracy, precision, recall, specificity, and F1-score were rated 99.22%, 99.25%, 99%, 99.77%, and 99.25%, respectively. In each of the WBC classes, accuracy rates are: 98.7% for eosinophils, 99.83% for lymphocytes, 100% for monocytes, and 98.32% for neutrophils. As a result, the proposed WBC-KICNet classifies WBCs accurately and without much misclassification, and the results have been confirmed by a statistical hypothesis test (t-test).

The White Blood Cell (WBC) differential test ranks as the second most frequently performed diagnostic assay. It requires manual confirmation of the peripheral blood smear by experts to identify signs of abnormalities. Automated digital microscopy has emerged as a solution to reduce this labor-intensive process and improve efficiency. Several publicly available datasets provide various WBC subtypes of differing quality and resolution. These datasets have contributed to advancing WBC classification using machine learning techniques. However, digital microscopy of blood cells with high magnification often requires a wider depth of field, posing challenges for automatic digital microscopy that necessitates capturing multiple stacks of focal planes to obtain complete images of specific blood cells. Our dataset provides 25,773 image stacks from 72 patients. The image labels consist of 18 classes encompassing normal and abnormal cells, with two experts reviewing each label. Each image includes 10 z-stacks of cropped 200 by 200 pixel images, captured using a 50X microscope with 400 nm intervals. This study presents a comprehensive multi-focus dataset for WBC classification.

The morphological classification of nucleated blood cells is fundamental for the diagnosis of hematological diseases. Many Deep Learning algorithms have been implemented to automatize this classification task, but most of the time they fail to classify images coming from different sources. This is known as “domain shift”. Whereas some research has been conducted in this area, domain adaptation techniques are often computationally expensive and can introduce significant modifications to initial cell images. In this article, we propose an easy-to-implement workflow where we trained a model to classify images from two datasets, and tested it on images coming from eight other datasets. An EfficientNet model was trained on a source dataset comprising images from two different datasets. It was afterwards fine-tuned on each of the eight target datasets by using 100 or less-annotated images from these datasets. Images from both the source and the target dataset underwent a color transform to put them into a standardized color style. The importance of color transform and fine-tuning was evaluated through an ablation study and visually assessed with scatter plots, and an extensive error analysis was carried out. The model achieved an accuracy higher than 80% for every dataset and exceeded 90% for more than half of the datasets. The presented workflow yielded promising results in terms of generalizability, significantly improving performance on target datasets, whereas keeping low computational cost and maintaining consistent color transformations. Source code is available at: https://github.com/mc2295/WBC_Generalization

Categorizing white blood cells (WBCs) in blood specimens is crucial for early disease detection in hematology and pathology. Numerous research endeavors have been undertaken to investigate the precise classification of WBCs in microscopic blood smear images, utilizing machine learning and deep learning methodologies. However, the notable accuracy achieved by most of these techniques relies on the assumption that the conditions under which the test microscope images are acquired closely resemble those of the training images, including the use of identical microscopes and consistent lighting conditions. The primary aim of this research is to design a novel network layer that can be used as a component in combination with other backone network, and improve the generalization ability of this backbone for classification tasks on WBC data. In this investigation, we predominantly employ the color invariance technique to fashion a trainable convolutional layer, which serves as the tool for extracting edge features from the original image. This layer is amenable to integration with various backbone architectures. Ultimately, the model was trained utilizing the Raabin-WBC dataset, and its efficacy was evaluated by appraising its performance on WBC images acquired through a variety of microscope instruments.

The assessment of a patient’s blood sample is a critical obligation in the healthcare industry. Different health problems are caused by aberrant blood cell growth. One of the vital elements of blood are white blood cells (WBC). Manually identifying WBC under a microscope requires a significant amount of time and effort, and this procedure is susceptible to human error, which might result in wrong interpretation. The automated systems for recognizing and categorizing the various types of WBC have major medicinal applications in this field. Despite the fact that the deep convolution neural network method gave decent results for categorizing WBC images. In the proposed study, a CNN-based deep learning algorithm is used to identify WBC. The dataset contains a variety of white blood cells, including neutrophils, lymphocytes, eosinophils, basophils, and monocytes. The CNN model was tested using multiple optimizers with batch sizes of 32 and 10 epochs. For the CNN model study, performance metrics such as Precision, Recall, F1-Score, and Accuracy were employed. The suggested model’s greatest accuracy is 93%. The proposed methodology provides exact and speedy results, which may assist save patients’ lives.

No abstract available

White blood cells protects human body from the various diseases. This study aimed to address the critical issue of distinguishing between different subtypes of Leukocytes (WBS) using deep learning techniques. Recent progress in biomedical diagnostics has been made possible by convolutional neural network (CNN) models, which have showed promise in automating the classification of white blood cell images. This proposed work introduces a unique CNN model architecture that can categorize white blood cell pictures into Polynuclear and Mononuclear classification. 12,500 images of white blood cells made up a large dataset that was used. With a batch size of 128 and a default learning rate, the suggested CNN model gone through intense training that lasted 30 iterations. The outcomes of this experiment show how successful the proposed technique is. When correctly categorizing the various white blood cell classification, the CNN model performed exceptionally well, obtaining a fantastic accuracy rate of 95%. For the discipline of medical image analysis, these discoveries have significant implications. This research adds to the creation of automated and precise systems for classifying white blood cell subtypes by utilizing the power of deep learning, particularly the CNN architecture. This CNN model's successful deployment sets the way for more complex AI-driven solutions in medical imaging, transforming healthcare procedures and enhancing overall medical decision-making. This proposed strategy can help with better categorization planning by allowing for the early identification of white blood cell subgroups.

In the realm of computer vision, the application of deep learning techniques has brought about significant transformations, particularly in the domain of image classification. This study delves into an in-depth exploration of the Raabin White Blood Cell (WBC) dataset, leveraging the power of the Deep Vision Transformer (ViT), an advanced deep learning model, to advance the classification of white blood cells. Although the Raabin WBC dataset may be relatively lesser-known, its inherent value lies in its specialized focus on white blood cell images. To harness the potential of this dataset, we meticulously curated it for compatibility with DeepViT, followed by rigorous training and fine-tuning. Our efforts yielded exceptional results, with an impressive accuracy rate of 97%, thus surpassing previous benchmarks. This remarkable performance underscores the dataset’s significance in the context of medical image analysis, highlighting its potential for various applications, particularly in the field of medical research and diagnosis. Furthermore, this study serves as a compelling testament to the adaptability and efficacy of DeepViT when applied to specialized datasets. These groundbreaking results accentuate the importance of exploring lesser-known datasets in conjunction with advanced deep learning models, as they hold the promise of delivering exceptional performance in specific domains. This work paves the way for further research and applications in medical image analysis, setting a benchmark for future white blood cell classification studies.

Deep learning (DL) has made significant advances in computer vision with the advent of vision transformers (ViTs). Unlike convolutional neural networks (CNNs), ViTs use self-attention to extract both local and global features from image data, and then apply residual connections to feed these features directly into a fully networked multilayer perceptron head. In hospitals, hematologists prepare peripheral blood smears (PBSs) and read them under a medical microscope to detect abnormalities in blood counts such as leukemia. However, this task is time-consuming and prone to human error. This study investigated the transfer learning process of the Google ViT and ImageNet CNNs to automate the reading of PBSs. The study used two online PBS datasets, PBC and BCCD, and transferred them into balanced datasets to investigate the influence of data amount and noise immunity on both neural networks. The PBC results showed that the Google ViT is an excellent DL neural solution for data scarcity. The BCCD results showed that the Google ViT is superior to ImageNet CNNs in dealing with unclean, noisy image data because it is able to extract both global and local features and use residual connections, despite the additional time and computational overhead.

White blood cells, or leukocytes, are indispensable for the optimal functioning of the immune system. They play a critical role in protecting the body against infections, diseases, and other foreign invaders by identifying and fighting harmful bacteria and pathogens that can cause illness. Additionally, they contribute to the elimination of dead and damaged cells from the body and facilitate tissue healing and repair processes. The absence of white blood cells would render the body defenceless against infections and diseases, exposing it to a variety of harmful pathogens. This could result in significant health issues and potentially even lead to death in severe instances. White blood cell classification is an important task in medical diagnosis and treatment because healthcare professionals diagnose and treat a variety of immune system-related diseases and conditions, including autoimmune disorders, infections, and cancers by identifying the structure, characteristics and functions of white blood cells. In this work, a convolutional neural network (CNN) model has been trained to classify white blood cells. The proposed model has achieved an accuracy of 88.78%, which has been identified as the highest among all the models implemented by various authors in the literature review. This implies that the proposed model has correctly classified white blood cells in almost 9 out of 10 cases. Moreover, the error rate of the model is only 0.108967 which indicates that the model is very reliable and consistent in its predictions. Additionally, this work shows the promising result for white blood cell classification using deep learning techniques. Furthermore, with improvements and refinements in the future, it can be possible to achieve higher levels of accuracy and precision, which could have a significant impact on medical diagnosis and treatment.

No abstract available

White blood cell (WBC) classification plays an important role in human pathological diagnosis since WBCs will show different appearance when they fight with various disease pathogens. Although many previous white blood cell classification have been proposed and earned great success, their classification accuracy is still significantly affected by some practical issues such as uneven staining, boundary blur and nuclear intra-class variability. In this paper, we propose a deep neural network for WBC classification via discriminative region detection assisted feature aggregation (DRFA-Net), which can accurately locate the WBC area to boost final classification performance. Specifically, DRFA-Net uses an adaptive feature enhancement module to refine multi-level deep features in a bilateral manner for efficiently capturing both high-level semantic information and low-level details of WBC images. Considering the fact that background areas could inevitably produce interference, we design a network branch to detect the WBC area with the supervision of segmented ground truth. The bilaterally refined features obtained from two directions are finally aggregated for final classification, and the detected WBC area is utilized to highlight the features of discriminative regions by an attention mechanism. Extensive experiments on several public datasets are conducted to validate that our proposed DRFA-Net can obtain higher accuracies when compared with other state-of-the-art WBC classification methods.

The accurate classification of white blood cells is crucial for diagnosing and monitoring various medical conditions. Traditional methods of diagnosis are expensive, time-consuming, and depend on specialists' expertise. Image processing-based techniques provide a faster and simpler approach to detecting abnormalities in white blood cells. In this study, a computer-aided diagnosis system was proposed that utilizes pre-trained networks for accurate classification of white blood cells. The dataset used in this study is sourced from Kaggle, and the classification process is performed without image segregation or feature extraction techniques. Pretrained series networks are employed for classification, and a classification accuracy of 99% is achieved using the Inception-ResNet-v2 network with the Adam optimizer. The comparative analysis aids in understanding the performance and suitability of different network architectures for white blood cell classification tasks. The results showcase the potential of the proposed computer-aided diagnosis system for accurate and efficient white blood cell classification, contributing to advancing the automation and reliability of white blood cell analysis and supporting improved disease diagnosis and monitoring.

The research focuses on the segmentation and classification of leukocytes, a crucial task in medical image analysis for diagnosing various diseases. The leukocyte dataset comprises four classes of images such as monocytes, lymphocytes, eosinophils, and neutrophils. Leukocyte segmentation is achieved through image processing techniques, including background subtraction, noise removal, and contouring. To get isolated leukocytes, background mask creation, Erythrocytes mask creation, and Leukocytes mask creation are performed on the blood cell images. Isolated leukocytes are then subjected to data augmentation including brightness and contrast adjustment, flipping, and random shearing, to improve the generalizability of the CNN model. A deep Convolutional Neural Network (CNN) model is employed on augmented dataset for effective feature extraction and classification. The deep CNN model consists of four convolutional blocks having eleven convolutional layers, eight batch normalization layers, eight Rectified Linear Unit (ReLU) layers, and four dropout layers to capture increasingly complex patterns. For this research, a publicly available dataset from Kaggle consisting of a total of 12,444 images of four types of leukocytes was used to conduct the experiments. Results showcase the robustness of the proposed framework, achieving impressive performance metrics with an accuracy of 97.98% and precision of 97.97%. These outcomes affirm the efficacy of the devised segmentation and classification approach in accurately identifying and categorizing leukocytes. The combination of advanced CNN architecture and meticulous pre-processing steps establishes a foundation for future developments in the field of medical image analysis.

No abstract available

Abstract. Significance Leukocytes are mainly composed of neutrophils, basophils, eosinophils, monocytes, and lymphocytes. The number and proportion of different types of leukocytes correspond to different diseases, so an accurate segmentation of each type of leukocyte is important for the diagnosis of disease. However, the acquisition of blood cell images can be affected by external environmental factors, which can lead to variable light and darkness, complex backgrounds, and poorly characterized leukocytes. Aim To address the problem of complex blood cell images collected under different environments and the lack of obvious leukocyte features, a leukocyte segmentation method based on improved U-net is proposed. Approach First, adaptive histogram equalization-retinex correction was introduced for data enhancement to make the leukocyte features in the blood cell images clearer. Then, to address the problem of similarity between different types of leukocytes, convolutional block attention module is added to the four skip connections of U-net to focus the features from spatial and channel aspects, so that the network can quickly locate the high-value information of features in different channels and spaces. It avoids the problem of large amount of repeated computation of low-value information, prevents overfitting, and improves the training efficiency and generalization ability of the network. Finally, to solve the problem of class imbalance in blood cell images and to better segment the cytoplasm of leukocytes, a loss function combining focal loss and Dice loss is proposed. Results We use the BCISC public dataset to verify the effectiveness of the proposed method. The segmentation of multiple leukocytes using the method of this paper can achieve 99.53% accuracy and 91.89% mIoU. Conclusions The experimental results show that the method achieves good segmentation results for lymphocytes, basophils, neutrophils, eosinophils, and monocytes.

The number, relative ratio, and appearance of peripheral blood leukocytes can assist doctors to diagnose diseases such as lymphoma and leukemia. Therefore, segmentation of peripheral blood leukocytes in blood smear images plays a crucial role in the diagnosis of certain diseases. We propose a semisupervised leukocyte segmentation method under the framework of adversarial learning, in which a discriminator is trained to differentiate the segmentation maps coming either from the ground truth or from the segmentation network. Specifically, we first construct a lightweight leukocyte segmentation network by modifying DeepLabV3+ with simplified MobileNetV2. In addition, a reconstruction decoder is developed as a constraint on feature extraction to recover the original leukocyte image from its segmentation result. Then, we design the discriminator in a fully convolutional manner, which as well enables semi-supervised learning by inferring reliable supervisory information for unlabeled images. Quantitative and qualitative experimental results on four image datasets demonstrate the superiority of the proposed method over several state-of-the-art methods, achieving the highest average $F1$ score of 0.9717 and mIoU of 0.9493 with only 20% labeled training samples.

To address the issues of uneven illumination and inconspicuous leukocyte properties in the gathered cell pictures, a leukocyte segmentation method based on adaptive retinex correction and U-net was proposed. The procedure begins by processing a peripheral blood image to clearly distinguish leukocytes from other components in the image. The adaptive retinex correction, which is based on multiscale retinex with colour replication (MSRCR), redefines the colour recovery function by introducing Michelson contrast. Then, the image is trained with the U-net convolutional neural network, and the leukocyte segmentation is completed. The innovation is in the manner of processing peripheral blood images, which improves the accuracy of leukocyte segmentation. This study conducts experimental evaluations on the Cellavision, BCCD, and LISC datasets. The experimental results show that the method in this study is better than the current best method, and the segmentation accuracy rate reaches 98.87%.

Accurate leukocyte segmentation remains challenging in automated hematological analysis due to staining variability, heterogeneous imaging conditions, and morphological diversity across cytological datasets, severely limiting deep learning model generalization. This work proposes a dual-module framework designed to achieve stain-invariant and robust leukocyte segmentation. The first module performs explicit stain standardization by combining a VGG-based encoder, a transformer bottleneck, and a convolutional decoder to harmonize diverse inputs toward a Wright–Giemsa reference appearance. The second module introduces a multi-encoder segmentation architecture integrating complementary spatial, leukocyte-specific, and nucleus-focused representations extracted from multiple color spaces. The framework is evaluated on six public and clinical datasets covering multiple staining protocols, magnifications, and imaging scenarios. Experimental results demonstrate consistent high performance, with Dice coefficients exceeding 96% on most datasets and systematic improvements over state-of-the-art methods. Extensive ablation studies confirm the synergistic contributions of stain-standardization and multi-encoder fusion to model robustness and cross-dataset generalization. This framework overcomes stain variability and domain shift, offering a practical tool for automated leukocyte analysis in clinical settings.

To address the urgent need for precise malaria diagnosis and white blood cell analysis in resource-limited regions, this paper proposes an intelligent blood image analysis system that integrates a portable microscope with deep learning algorithms. The system employs a self-developed high-magnification mobile microscopy platform and adopts an end-cloud-end collaborative architecture to achieve efficient acquisition and remote processing of blood smear images. To cope with the complex distribution of white blood cells and Plasmodium par-asites in the images, we innovatively introduce a multimodal fusion segmentation and feature validation pipeline, efficiently screening candidate Plasmodium regions via a modified AlexNet model, while achieving fine-grained classification of five white blood cell types based on a medically optimized GoogLeNet architecture. Experiments conducted on a real clinical dataset spanning multiple regions and qualities validate the system's superior performance in Plasmodium detection, white blood cell segmentation, and classification tasks, with an overall malaria parasite detection accuracy of 97.74 %, and white blood cell segmentation and classification accuracies of 93.78 % and 94.70 %, respectively, significantly enhancing processing speed and model generalization capability. The results indicate that the system provides efficient and reliable technical support for on-site rapid malaria screening and auxiliary diagnosis of blood diseases, and is expected to be promoted in primary healthcare and mobile diagnostic scenarios.

Advances in image acquisition and computational processing technologies have significantly transformed histological and hematological analysis. However, to maximize the clinical impact of these technologies, it is essential to develop automatic or semi-automatic segmentation techniques. The objective of this work is to develop an automatic multi-class segmentation technique that achieves competitive performance while minimizing computational costs. The proposed technique consists of an unsupervised algorithm that models a neutrosophic image by combining only two features: the luminance channel L* from the CIELUV color space and the HH sub-band of the discrete wavelet transform obtained from the grayscale image. The neutrosophic components are enhanced to promote robust clustering and segmentation, especially in regions of uncertainty. Finally, morphological operations are applied as a post-processing stage. Experimental results demonstrate strong performance in multiclass segmentation, with average standard index values above 90% and an overall correct classification rate of 97.07%. The developed algorithm has publicly available open-source code, operates directly on color images without requiring any training and achieves competitive results with a low-cost implementation.

The quantity of leukocytes in papillary thyroid carcinoma (PTC) potentially have prognostic and treatment predictive value. Here, we propose a novel method for training a convolutional neural network (CNN) algorithm for segmenting leukocytes in PTCs. Tissue samples from two retrospective PTC cohort were obtained and representative tissue slides from twelve patients were stained with hematoxylin and eosin (HE) and digitized. Then, the HE slides were destained and restained immunohistochemically (IHC) with antibodies to the pan-leukocyte anti CD45 antigen and scanned again. The two stain-pairs of all representative tissue slides were registered, and image tiles of regions of interests were exported. The image tiles were processed and the 3,3′-diaminobenzidine (DAB) stained areas representing anti CD45 expression were turned into binary masks. These binary masks were applied as annotations on the HE image tiles and used in the training of a CNN algorithm. Ten whole slide images (WSIs) were used for training using a five-fold cross-validation and the remaining two slides were used as an independent test set for the trained model. For visual evaluation, the algorithm was run on all twelve WSIs, and in total 238,144 tiles sized 500 × 500 pixels were analyzed. The trained CNN algorithm had an intersection over union of 0.82 for detection of leukocytes in the HE image tiles when comparing the prediction masks to the ground truth anti CD45 mask. We conclude that this method for generating antibody supervised annotations using the destain-restain IHC guided annotations resulted in high accuracy segmentations of leukocytes in HE tissue images.

No abstract available

Segmentation of blood cells is a prerequisite step in automated morphological analysis of blood smear images, cell count determination and diagnosis of various diseases such as leukemia. It is extremely challenging due to different sizes, shapes, morphological characteristics and overlapping of blood cells. Due to its complicated nature, it is generally performed as a sequence of steps. However, sequential segmentation results in restricted accuracy due to cascading of errors that creep during each stage. On the contrary, pixel-wise segmentation of blood cells is a single step task and gives promising results. In this paper, we propose LeukoSegmenter, a double encoder-decoder for precise pixel-wise segmentation of leukocytes from blood smear images. It uses pre-trained ResNet18 based encoders and U-Net based decoders. Feature maps obtained from the first network are utilised as attention maps. These are used as input in conjunction with the original 3-channel image to obtain final mask from the second network. This mechanism allows the latter encoder-decoder pair to focus explicitly on leukocytes and ignore other blood cells and debris, thus improving the segmentation accuracy. Experiments on ALL-IDB1 dataset show that the proposed LeukoSegmenter achieves intersection-over-union score of 94.6827% and Dice score of 97.1987% which is superior than that of state-of-the-art methods.

No abstract available

Leukocytes are an essential component of the human defense system, accurate segmentation of leukocyte images is a crucial step towards automating detection. Most existing methods for leukocyte images segmentation relied on fully supervised semantic segmentation (FSSS) with extensive pixel-level annotations, which are time-consuming and labor-intensive. To address this issue, this paper proposes a weakly supervised semantic segmentation (WSSS) approach for leukocyte images utilizing improved class activation maps (CAMs). Firstly, to alleviate ambiguous boundary problem between leukocytes and background, preprocessing technique is employed to enhance the image quality. Secondly, attention mechanism is added to refine the CAMs generated by improving the matching of local and global features. Random walks, dense conditional random fields and hole filling were leveraged to obtain final pseudo-segmentation labels. Finally, a fully supervised segmentation network is trained with pseudo-segmentation labels. The method is evaluated on BCCD and TMAMD datasets. Experimental results demonstrate that by employing the pseudo segmentation annotations generated through this method can be utilized to train UNet as close as possible to FSSS. This method effectively reduces manual annotation cost while achieving WSSS of leukocyte images.

No abstract available

BACKGROUND: Blood cell count, also known as differential count of various types of blood cells, provides valuable information in order to assess variety of diseases like AIDS, leukemia and blood cancer. Manual techniques are still used in diseases diagnosis that is very lingering and tedious process. However, machine based automatic analysis of leukocyte is a powerful tool that could reduce the human errors, improve the accuracy, and minimize the required time for blood cell analysis. However, leukocyte segmentation is a challenging process due to the complexity of the blood cell image; therefore, this task remains unresolved issue in the blood cell segmentation. OBJECTIVE: The aim of this work is to develop an efficient leukocyte cell segmentation and classification system. METHODS: This paper presents an efficient strategy to segment cell images. This has been achieved by using Wiener filter along with Curvelet transform for image enhancement and noise elimination in order to elude false edges. We have also used combination of entropy filter, thresholding and mathematical morphology for obtaining image segmentation and boundary detection, whereas we have used back-propagation neural network for leukocyte classification into its sub classes. RESULTS: As a result, the generated segmentation results are fruitful in a sense that we have overcome the problem of overlapping cells. We have obtained 100%, 96.15%, 92.30%, 92.30% and 96.15% accuracy for basophil, eosinophil, monocyte, lymphocyte and neutrophil respectively.

Medical images play a fundamental role in disease screening, and automated evaluation of these images is widely preferred in hospitals. Recently, Convolutional Neural Network (CNN) supported medical data assessment is widely adopted to inspect a set of medical imaging modalities. Extraction of the leukocyte section from a thin blood smear image is one of the essential procedures during the preliminary disease screening process. The conventional segmentation needs complex/hybrid procedures to extract the necessary section and the results achieved with conventional methods sometime tender poor results. Hence, this research aims to implement the CNN-assisted image segmentation scheme to extract the leukocyte section from the RGB scaled hematological images. The proposed work employs various CNN-based segmentation schemes, such as SegNet, U-Net, and VGG-UNet. We used the images from the Leukocyte Images for Segmentation and Classification (LISC) database. In this work, five classes of the leukocytes are considered, and each CNN segmentation scheme is separately implemented and evaluated with the ground-truth image. The experimental outcome of the proposed work confirms that the overall results accomplished with the VGG-UNet are better (Jaccard-Index = 91.5124%, Dice-Coefficient = 94.4080%, and Accuracy = 97.7316%) than those of the SegNet and U-Net schemes Finally, the merit of the proposed scheme is also confirmed using other similar image datasets, such as Blood Cell Count and Detection (BCCD) database and ALL-IDB2. The attained result confirms that the proposed scheme works well on hematological images and offers better performance measure values.

No abstract available

No abstract available

A leukocyte segmentation method based on S component and B component images is proposed. Threshold segmentation operation is applied to get two binary images in S component and B component images. The samples used in this study are peripheral blood smears. It is easy to find from the two binary images that gray values are the same at every corresponding pixels in the leukocyte cytoplasm region, but opposite in the other regions. The feature shows that "IMAGE AND" operation can be employed on the two binary images to segment the cytoplasm region of leukocyte. By doing "IMAGE XOR" operation between cytoplasm region and nucleus region, the leukocyte segmentation can be retrieved effectively. The segmentation accuracy is evaluated by comparing the segmentation result of the proposed method with the manual segmentation by a hematologist. Experiment results show that the proposed method is of a higher segmentation accuracy and it also performs well when leukocytes overlap with erythrocytes. The average segmenta...

No abstract available

No abstract available

No abstract available

Medical image segmentation is very important for detecting illness and treatment in a precise manner. The diseases like leukemia, malaria, and other blood disorders identification mainly depends on the anomalies segmentation in microscopic blood smear images. Traditional methods for medical image segmentation are laborious to unpredictability because they need a frequent domain-specific expertise and manual intervention. The study provides a novel method to segment microscopic blood smears using Trans-SegNet, a deep transfer learning algorithm. By using pre-trained Convolutional Neural Networks (CNNs), the proposed method transfer learned features efficiently to the blood smear analysis domain, so that segmentation accuracy can be improved. A pre-trained CNN is fine-tuned using blood smear images properties and for accurate segmentation, the proposed method uses a Fully Convolutional Network (FCN). Experimentation uses publicly accessible blood smear image dataset and results show that the proposed method outperforms traditional segmentation techniques. The Trans-SegNet obtained an Intersection over Union (IoU) of 89.7% and Dice Similarity Coefficient (DSC) of 92.5%. Furthermore, precision with 91.3% and recall with 90.8% were recorded.

Leukemia diagnosis from peripheral blood smear images is difficult because of the variability in staining, overlapping cells, intensity inhomogeneity, and similarities in the morphological patterns of different types of leukemia. To overcome these problems, this paper presents Stain-Adaptive Lukeman (SAL-Net), a deep learning framework optimized for preprocessing, which is guided by the SAPHIRE (Stain-normalized Adaptive Preprocessing with Hybrid Intelligent Recognition and Enhancement) Algorithm for efficient segmentation and multiclass classification of white blood cells and leukemia. The proposed work combines a stain-adaptive preprocessing module, which incorporates contrast normalization, stain normalization, and edge-preserving filtering to reduce inter-slide variability and improve the definition of cell boundaries before feature extraction. Unlike existing approaches, where the preprocessing step is fixed, SAL-Net incorporates adaptive enhancement as part of the learning process to enhance feature robustness and generalization. The pre-processed images are then processed using a modified U-Net with hybrid loss optimization for accurate segmentation, followed by a hybrid deep classification module with multi-scale feature learning for accurate classification of different types of leukemia and healthy cells. The proposed framework is unique in its stain-adaptive learning paradigm, two-stage segmentation-classification architecture, and optimization approach designed for robustness against staining variability. The proposed framework has been shown to provide 3-5%. In addition, the proposed SAL-Net framework has a strong ability to generalize over different datasets and acquisition conditions, ensuring diagnostic performance. The SAPHIRE algorithm further improves the discriminability of features by adaptive optimization, which is less sensitive to noise and illumination changes. The ablation study validates the important role of the stain-adaptive preprocessing module in the overall segmentation and classification accuracy. The computational complexity analysis indicates that the framework has efficient inference time for real-time clinical screening. In summary, the suggested system is an intelligent and scalable one for next-generation computer-aided leukemia diagnosis systems.

The focus of this study is on biomedical image data processing and analysis. The primary objective is to identify leukocytes in blood smear samples and determine their quantities. Since the manual laboratory processing of this data is a laborious task, it is susceptible to subjectivity and time-consuming. Hence, the objective of this study is to examine contemporary processing methodologies and existing technologies that have the potential to speed up and automate this procedure. This study presents a concept that utilizes existing methods to address a problem, including a unique sequence of operations. Finally, it evaluates the accuracy of this approach for future research. This approach has the potential to be used in the diagnosis of a range of disorders, including blood malignancies or infections.

White blood cell (WBC) segmentation of microscopic blood smear images (MBSI) is a significant process in the diagnosis of blood-related diseases like anemia, leukemia, and parasites. Traditional manual segmentation is tiresome and sensitive to inter-observer variation, which limits its application to a big-scale clinical process. To address these issues, this paper presents a hybrid DINO-Net approach, which works on WBC segmentation. The suggested solution uses a DINO-Net encoder with self-supervised training to obtain discriminative features and context-sensitive features, which are then decoded by a U-Net-based network to obtain the correct segmentation masks. The effectiveness of the offered framework is contrasted with popular segmentation models, i.e., U-Net, DeepLabv3+, and a DINOv2-based one. The experimental outcomes demonstrate that the suggested strategy is always more successful in comparison to the comparison models during the training, validation, and testing stages. Precisely, Dice coefficients of $93.00 \%, 89.25 \%$, and 90.35 %, as well as IoU scores of $87.10 \%, 81.90 \%$, and 85.23 %, are obtained, respectively. Also, the recall values of $92.24 \%, 80.35 \%$, and 81.64 % show that WBC regions are reliably detected. These results prove that self-supervised transformer-Convolutional Neural Network (CNN) addition makes it a possibility to segment WBC.

No abstract available

Leukemia is a blood cancer that attacks human white blood cells. This disease is divided into four types, including Acute Myeloid Leukemia (AML). AML is the most common type of acute leukemia, and it has eight types of subtypes distinguished by the level of cell maturation. Medical personnel determines the type of AML based on microscopic images of blood cell smears that contain white blood cells, red blood cells, and pieces of blood. This research builds a segmentation system that can determine the boundary of an object with the surrounding area, where the object sought is white blood cells contained in microscopic images of blood cell smears. White blood cells are sought based on ROI using the Haar Cascade Classifier, and then segmentation is carried out on the nucleus and cytoplasm. AML sub-types used as objects in this study are M4, M5, and M7. Based on the results of experimental data on the segmentation system, the nucleus segmentation in each cell of M4, M5, and M7 with an accuracy of 87.5%, 90.4%, 84.6% in sequence, and the results of cytoplasm segmentation are 75%, 71.4%, and 80.76%, respectively.

Blood tests play an important role in the detection of various types of diseases. The problem arises when large quantities of blood samples need to be processed. As the number of samples increases, so does the risk of error. It limits the speed and precision of blood sampling. Blood cells can be divided into three types: Red blood cells (RBC), White blood cells (leukocytes) and platelets. Of these globules, white blood cells play an important role in disease detection because their conventional method of detecting and classifying white blood cells under the microscope is inaccurate and leads to the development of new methods. This paper presents an effective computer method for WBC classification. This method consists of three steps. They are preprocessing, segmentation and classification using an ANN classifier. The experiments were performed on 65 digital blood smear images of 5 different types of WBC cells.

No abstract available

Leukemia is a life-threatening blood disorder marked by an abnormal increase in white blood cells, known as blast cells, in the bone marrow. Early detection of these blast cells is critical for effective treatment. Traditionally, diagnosis relies on manual examination of blood smear images by pathologists, a process that is not only time-consuming but also prone to human error. With the rise of ML along with the integration of deep learning , there is a growing opportunity to automate and enhance the diagnostic process. This study explores the application of the U-Net architecture, a neural network model engineered for image segmentation, to automatically detect and segment blast cells in leukemia diagnosis. By automating this process, the aim is to reduce diagnostic time, minimize errors, and improve the overall accuracy of leukemia detection. Index Terms—Leukemia, image segmentation, U-Net, ma- chine learning, deep learning.

Acute Lymphoblastic Leukemia (ALL) is a dangerous form of leukemia, which disturbs the bone marrow and irregular White Blood Cells (WBC) for the persons of all age groups. Early diagnosis of leukemia is necessary to provide appropriate care and cure the patients. Still, the detection of ALL from Peripheral Blood Smear (PBS) images is complicated due to the size and shape of the cells. This variability creates more difficulty for the segmentation and classification of Leukemia. To solve these issues, the Deep Kronecker Wide Residual Network (DKWRN) is developed for classifying Leukemia. The blood smear images are preprocessed using adaptive Gaussian filtering; hence, the noise from the image is diminished. After that, RefineNet effectively segments the Blast cell from the image. Then, augmentation such as flipping, resizing, and rotation enhances the dimension of the image. Following augmentation, essential features like Binary Pattern of Phase Congruency (BPPC) with Discrete Cosine Transform (DCT), and statistical features are extracted. Finally, the DKWRN classifies the Leukemia into early Pre-B, Pre-B, Pro-B, and Hematogones. Here, the proposed DKWRN is the integration of WideResNet (WRN) and Deep Kronecker Network (DKN). Furthermore, the proposed model offered optimal accuracy, True Negative Rate (TNR), recall, precision, and F1-score of 92.12 %, 91.40 %, 90.36 %, 91.56 %, and 90.96 %, respectively. Furthermore, the proposed DKWRN model demonstrated notable enhancements in classification accuracy when compared to existing techniques. Specifically, its performance exceeded that of Bayesian-based optimized convolutional neural network (CNN) for ALL detection (BO-ALLCNN) by 5.91 %, Support vector machine (SVM) by 4.63 %, ResNet-18 combined with SVM by 4.04 %, Residual Convolutional Neural Network (ResNet-152) by 2.27 %, DKN by 1.51 %, and WRN by 0.55 %, respectively.

The classification and recognition of leukocytes or WBCs in blood smear images presents a key role in the corresponding diagnosis of specific diseases, such as leukemia, tumor, hematological disorders, etc. The computerized framework for automated segmentation & classification of WBCs nucleus contributes an important role for the recognition of WBCs related disorders. Therefore, this work emphasizes WBCs nucleus segmentation using modified U-Net architecture and the segmented WBCs nucleus are further classified into their subcategory i.e., basophil, eosinophil, neutrophil, monocyte and lymphocyte. The classification and nucleus characterization task has been performed using VGGNet and MobileNet V2 architecture. Initially, collected instances are passed to the preprocessing phase for image rescaling and normalization. The rescaled and normalized instances are passed to the U-Net model for nucleus segmentation. Extracted nucleus are forwarded to the classification phase for their class identifications. Furthermore, the functioning of the intended design will be compared with other modern methods. By the end of this study a successful model classifying various nucleus morphologies such as Basophil, Eosinophil, Lymphocyte, Monocyte and Neutrophil was obtained where overall test accuracy achieved was 97.0% for VGGNet classifier and 94.0% for MobileNet V2 classifier.

Blood cell classification and counting are vital for the diagnosis of various blood-related diseases, such as anemia, leukemia, lymphoma, and thrombocytopenia. The manual process of blood cell classification and counting is time-consuming, prone to errors, and labor-intensive. Therefore, we have proposed a deep learning (DL)-based automated system for blood cell classification and counting from microscopic blood smear images. We classify a total of nine types of blood cells, including Erythrocyte, Erythroblast, Neutrophil, Basophil, Eosinophil, Lymphocyte, Monocyte, Immature Granulocytes, and Platelet. Several preprocessing steps like image resizing, rescaling, contrast enhancement and augmentation are utilized. To segment the blood cells from the entire microscopic images, we employed the U-Net model. This segmentation technique aids in extracting the region of interest (ROI) by removing complex and noisy background elements. Both pixel-level metrics such as accuracy, precision, and sensitivity, and object-level evaluation metrics like Intersection over Union (IOU) and Dice coefficient are considered to comprehensively evaluate the performance of the U-Net model. The segmentation model achieved impressive performance metrics, including 98.23% accuracy, 98.40% precision, 98.26% sensitivity, 95.97% Intersection over Union (IOU), and 97.92% Dice coefficient. Subsequently, a watershed algorithm is applied to the segmented images to separate overlapped blood cells and extract individual cells. We have proposed a BloodCell-Net approach incorporated with custom light weight convolutional neural network (LWCNN) for classifying individual blood cells into nine types. Comprehensive evaluation of the classifier’s performance is conducted using metrics including accuracy, precision, recall, and F1 score. The classifier achieved an average accuracy of 97.10%, precision of 97.19%, recall of 97.01%, and F1 score of 97.10%. A 5-fold cross-validation technique is applied to split the data, which not only aids in reducing overfitting but also helps in generalizing the model.

No abstract available

No abstract available

Abstract Leukemia is an abnormal proliferation of leukocytes in the bone marrow and blood and it is usually diagnosed by the pathologists by observing the blood smear under a microscope. The count of various cells and their morphological features are used by the pathologists to identify and classify leukemia. An abnormal increase in the count of immature leukocytes along with a reduced count of other blood cells may be an indication of leukemia. The Pathologist may then recommend for bone marrow examination to confirm and identify the specific type of leukemia. These conventional methods are time consuming and may be affected by the skill and expertise of the medical professionals involved in the diagnostic procedures. Image processing based methods can be used to analyze the microscopic smear images to detect the incidence of leukemia automatically and quickly. Image segmentation is one of the very important tasks in processing and analyzing medical images. In the proposed paper an attempt has been made to review the available works in the area of medical image processing of blood smear images, highlighting automated detection of leukemia. The available works in the related area are reviewed based on the segmentation method used. It is learnt that even though there are many studies for detection of acute leukemia only a very few studies are there for the detection of chronic leukemia. There are a few related review studies available in the literature but, none of the works classify the previous studies based on the segmentation method used.

A peripheral blood smear is a basic test for hematological disease diagnosis. This test is performed manually in many places worldwide, which requires both time and qualified staff. Large laboratories are equipped with digital morphology analyzers, some of which are based on deep learning methods. However, it is difficult to explain to scientists how they work. In this paper, we proposed to add an explanatory factor to enhance the interpretability of deep learning models in leukocyte classification.

Both in clinical practice and translational research, cell differentiation of leukocytes provides important information used for diagnostics or insights into pathophysiological mechanisms. The current gold-standard method for bronchoalveolar lavage fluid (BALF) analysis involves histochemical staining of cytospins, followed by manual morphological quantification. However, this approach is labor-intensive, time-consuming, and highly operator-dependent, limiting its efficiency and throughput. This study proposes a deep learning framework for rapid, automated 3D leukocyte differentiation using label-free higher harmonic generation microscopy (HHGM). 3D leukocyte characterization was performed with label-free HHGM. Two deep learning models, ResNet 3D-50 and Vision Transformer (ViT) 3D, were trained, validated and tested for leucocyte differentiation on both BALF and blood fraction samples from 14 interstitial lung disease (ILDs) and 12 acute respiratory distress syndrome (ARDS) patients. Deep-learning model-prediction and cytospin analysis were performed by separate investigators. Results were compared using Bland-Altman analysis. The deep learning algorithm achieved >96% accuracy in quantifying neutrophils, eosinophils, lymphocytes, and macrophages/monocytes. Bland-Altman analysis showed mean differences of less than 3% between cytospin analysis and the deep learning based approach across all cell types. By integrating the label-free imaging capabilities of HHGM with deep learning, this study established a fast, accurate and high-throughput leukocyte differentiation in fresh BALF and blood samples. By significantly improving efficiency and reproducibility, this technology has the potential to transform clinical workflows and advance precision medicine.

Digital hematology is an emerging domain within the broader transformation of healthcare through artificial intelligence (AI). Manual examination of peripheral blood smears remains a time-consuming and expertise-dependent task, essential for diagnosing hematological disorders. This paper introduces a deep learning framework based on the YOLOv8 architecture for automated leukocyte detection and classification. Trained on a large multi-center dataset of high-resolution microscopy images spanning 23 clinically relevant leukocyte classes, the system integrates standardized preprocessing, class imbalance mitigation, and anchor-free object detection to directly process full-slide images. Experimental results demonstrate strong performance, achieving a mean Average Precision at IoU 0.5 (mAP@0.5) of 0.935 and an F1 score of 0.86, highlighting the potential of AI-driven approaches to enhance accuracy, scalability, and consistency in digital hematology workflows.

Human blood leucocyte formula analysis is a common laboratory test performed among the first to assess human health. Concurrently, blood cell recognition is a timeconsuming and error-prone procedure Existing systems that automate the process of leukocyte counting are related to the operations of localisation of cells in the blood sample image, and further, to the operation of classification of the selected cells. The existing methods and algorithms for recognising cells in a blood sample are prone to a large number of errors and are computationally expensive. In this paper it is proposed to divide the task into two stages - the actual classification of blood cells into three classes of erythrocytes, leukocytes and platelets, with further specification of the type of leukocyte. The developed software system is based on microservice architecture and includes separate neural network models for each of the steps. The proposed method allowed the use of more compact neural network models, which accelerated the recognition process without loss of accuracy.

White blood cells (WBCs) are crucial components of the immune system, responsible for detecting and eliminating pathogens. Accurate detection and classification of WBCs are essential for various clinical diagnostics. This study aims to develop an AI framework for detecting and classifying WBCs from microscopic images using a customized YOLOv5 model with three key modifications. Firstly, the C3 module in YOLOv5's backbone is replaced with the innovative C3TR structure to enhance feature extraction and reduce background noise. Secondly, the BiFPN is integrated into the neck to improve feature localization and discrimination. Thirdly, an additional layer in the head enhances detection of small WBCs. Experiments on the BCCD dataset, comprising 352 microscopic blood smear images with leukocytes, demonstrated the framework's superiority over state‐of‐the‐art methods, achieving 99.4% accuracy. Furthermore, the model exhibits computational efficiency, operating over five times faster than existing YOLO models. These findings underscore the framework's promise in medical diagnostics, showcasing deep learning's supremacy in automated cell classification.

Background In diseases such as interstitial lung diseases (ILDs), patient diagnosis relies on diagnostic analysis of bronchoalveolar lavage fluid (BALF) and biopsies. Immunological BALF analysis includes differentiation of leukocytes by standard cytological techniques that are labor-intensive and time-consuming. Studies have shown promising leukocyte identification performance on blood fractions, using third harmonic generation (THG) and multiphoton excited autofluorescence (MPEF) microscopy. Objective To extend leukocyte differentiation to BALF samples using THG/MPEF microscopy, and to show the potential of a trained deep learning algorithm for automated leukocyte identification and quantification. Methods Leukocytes from blood obtained from three healthy individuals and one asthma patient, and BALF samples from six ILD patients were isolated and imaged using label-free microscopy. The cytological characteristics of leukocytes, including neutrophils, eosinophils, lymphocytes, and macrophages, in terms of cellular and nuclear morphology, and THG and MPEF signal intensity, were determined. A deep learning model was trained on 2D images and used to estimate the leukocyte ratios at the image-level using the differential cell counts obtained using standard cytological techniques as reference. Results Different leukocyte populations were identified in BALF samples using label-free microscopy, showing distinctive cytological characteristics. Based on the THG/MPEF images, the deep learning network has learned to identify individual cells and was able to provide a reasonable estimate of the leukocyte percentage, reaching >90% accuracy on BALF samples in the hold-out testing set. Conclusions Label-free THG/MPEF microscopy in combination with deep learning is a promising technique for instant differentiation and quantification of leukocytes. Immediate feedback on leukocyte ratios has potential to speed-up the diagnostic process and to reduce costs, workload and inter-observer variations.

Leukocytes, sometimes referred to as white blood cells (WBCs), are crucial to the healthy operation of the human body. WBC distribution in human body are biological markers that determine the immunity of human body to fight against infectious diseases. WBC detection and classification plays an important role in medical application. However, using manual microscopic evaluation is complicated and time consuming. To tackle the limitations associated with traditional methods, recently deep learning (D.L) based methods are much experimented and explored. In this paper, we implemented various D.L models for automatic classification of WBCs. A comparative study among pretrained networks namely Inceptionv3, MobileNetV3 and VGG-19 was performed using transfer learning on publicly available WBC images from Kaggle. Classification accuracy of WBC images using Inceptionv3, MobileNetV3 and VGG-19 is 99.76%, 99.25% and 86.50% respectively. Inceptionv3 was further compared with the existing works in the literature and is found to be superior.

Acute Myeloid Leukemia (AML) is a rare and rapidly progressing disease characterized by genetic mutations that disrupt the normal development of blood cells. Accurate diagnosis and classification of AML sub-categories are crucial for effective treatment strategies. In this paper, we propose an approach using deep learning classification techniques to automate the differentiation of leukocyte morphology and identify malignant cells. Our study utilizes a large dataset of individual cell images from peripheral blood smears and evaluates various state-of-the-art deep learning architectures through transfer learning and data augmentation techniques. Additionally, we employ a hybrid approach that combines deep learning-based featurization technique with machine learning models applied to the extracted features. Data augmentation techniques are employed to enhance dataset richness, and the featurized dataset is trained using three distinct algorithms: K Nearest Neighbors (KNN), Random Forest, and Multilayer Perceptron (MLP). Using the ConvNextLarge model for featurization with an MLP classifier achieved a leading 95.98% accuracy across 15 categories, surpassing prior algorithms. Our findings contribute to the field of AML diagnosis by providing a computationally efficient and accurate automated system for cytomorphological examination of leukocytes, paving the way for improved personalized treatment strategies,

Data augmentation is reported as a useful technique to generate a large amount of image datasets from a small image dataset. The aim of this study is to clarify the effect of data augmentation for leukocyte recognition with deep learning. We performed three different data augmentation methods (rotation, scaling, and distortion) as pretreatment on the original images. The subjects of clinical assessment were 51 healthy persons. The thin-layer blood smears were prepared from peripheral blood and stained with MG. The effect of data augmentation with rotation was the only significant effective technique in AI model generation for leukocyte recognition. On contrast, the effect of data augmentation with image distortion or image scaling was poor, and accuracy improvement was limited to specific leukocyte categories. Although data augmentation is one effective method for high accuracy in AI training, we consider that a highly effective method should be selected.

The conventional technique of leukocyte cell classification involves segmenting the required portion of cells from input image, extracting features of the segmented nuclei, reducing and optimizing these features and then implements the classifier. Thus, designing a good classifier by using such techniques increases the time complexity of the system. In order to resolve such issues, the proposed work implements the deep convolutional neural network (DCNN)‐based models for classifying malignant versus normal WBCs. The proposed system is validated on 108 images of ALL‐IDB 1. Due to limited number of training samples, data augmentation is used to create a similar type of virtual image. In this work, experimentation is carried out for discrimination between normal and infected WBC using DCNN with four different activation functions. By using this method, a set of 6000 samples are generated and used for proper training of the DL model for all activation functions. The performance of each trained model is evaluated in terms of accuracy, recall, precision and F‐measure with the maximum values of 98.1%, 98.3%, 98.3% and 98.3% are achieved, respectively. Finally, it has been concluded that the defined DCNN model and ReLu activation function yield outstanding performance for lymphoblast characterization using microscopic blood images.

The abnormal growth of leukocytes causes hematologic malignancies such as leukemia. The clinical assessment methods for the diagnosis of the disease are labor-intensive and time-consuming. Image-based automated diagnostic systems can be of great help in the decision-making process for leukemia detection. A feature-dependent, intrinsic, reliable classifier is a critical component in building such a diagnostic system. However, the identification of vital and relevant features is a challenging task in the classification workflow. The proposed work presents a novel two-step methodology for the robust classification of leukocytes for leukemia diagnosis by building a VGG16-adapted fine-tuned feature-extractor model, termed as "LeuFeatx," which plays a critical role in the accurate classification of leukocytes. LeuFeatx was found to be capable of extracting notable leukocyte features using microscopic single-cell leukocyte images. The filters and learned features are visualized and compared with base VGG16 model features. Independent classification experiments using three public benchmark leukocyte datasets were conducted to assess the effectiveness of extracted features with the proposed LeuFeatx model. Multiclass classifiers trained using LeuFeatx deep features achieved higher precision and sensitivity for seven leukocyte subtypes compared to the latest research on the AML Morphological dataset, and it achieved higher sensitivity for all cell types vis-à-vis recent work on peripheral blood cells dataset from the Hospital Clinic of Barcelona. In a binary classification experiment using the ALL_IDB2 dataset, classifiers trained using LeuFeatx deep features achieved an accuracy of 96.15%, which is better than the other state-of-the-art methods reported in the literature. Thus, the higher performance of the classifiers across observed comparison metrics establishes the relevance of the extracted features and the overall robustness of the proposed model.

Background: The Sysmex XN series (XN-1000 and XN-9100, Sysmex Corporation, Kobe, Japan) represents a latest-generation automated hematology platform integrating fluorescence-based technologies and multi-channel analysis (WDF and WPC) to improve leukocyte characterization. This study aimed to evaluate the performance of the Sysmex XN series in detecting leukocyte abnormalities flagged during routine complete blood count analysis in a large cohort of healthy donors, using morphological assessment and flow cytometry as confirmatory methods. Methods: Approximately 8000 healthy blood donors from the AOU Meyer Transfusion Centre were evaluated between 2021 and 2024. All samples underwent CBC analysis using the XN-1000 and XN-9100 analyzers with the WDF channel. Samples showing WBC-related flags were subjected to reflex testing with the WPC channel, followed by digital blood smear review using the DI-60 system (CellaVision, Lund, Sweden) and flow cytometric immunophenotyping. Results: WDF flags for “blasts/abnormal lymphocytes” were identified in 23 samples. Two samples were negative on WPC analysis as well as on morphological and flow cytometric evaluation. Among the remaining cases, WPC analysis identified flags for abnormal lymphocytes, atypical lymphocytes, or blasts, which were variably associated with reactive changes, transient immune activation, or clonal lymphoproliferative conditions. In one donor, monoclonal B-cell lymphocytosis was diagnosed by flow cytometry. Overall, reactive morphological features confirmed by flow cytometry were observed in approximately 50% of flagged cases. Conclusions: WPC analysis provides relevant additional diagnostic information and demonstrates higher specificity compared with the WDF channel alone; however, it does not fully resolve all instrument-generated flags, confirming the essential role of morphological assessment. Interestingly, the frequent occurrence of inflammatory profiles in recently vaccinated donors suggests that transient immune activation may influence leukocyte flagging. Larger studies are warranted to further investigate this association and to optimize the diagnostic performance of the WPC channel in donor screening.

No abstract available

Acute Lymphoblastic Leukemia (ALL) and Acute Myeloid Leukemia (AML) are aggressive hematologic malignancies that necessitate early and accurate detection for improved patient outcomes. Traditional diagnostic methods, relying on manual microscopic examination of blood smears, are time-consuming, labor-intensive, and prone to inter-observer variability. This paper introduces a novel convolutional deep learning (CNN) framework designed for automated blood cancer detection and classification from peripheral blood smear (PBS) images. Our approach leverages advanced CNN architectures, including customized models like ALLNET and DeepLeuk, and state-of-the-art models such as ConvNeXt, to achieve high diagnostic accuracy. We detail the preprocessing techniques, model training strategies, and rigorous evaluation metrics employed. The proposed system demonstrates superior performance compared to traditional methods and existing AI models, achieving accuracies up to 99.5% for leukemia staging and 98.00% for leukocyte subtyping. Furthermore, we incorporate explainability techniques like Grad-CAM to provide biological interpretability, bridging the gap between AI-driven diagnostics and clinical decision-making. This research underscores the potential of deep learning to revolutionize hematopathology, offering a scalable, efficient, and reliable solution for early blood cancer detection and classification, ultimately aiming to minimize diagnostic errors and improve patient survival rates

Leukemia is a type of cancer which is caused by malignant neoplasms in leukocyte cells. Leukemia disease which can cause death quickly enough for the sufferer is a type of acute lymphocyte leukemia (ALL). In this study, we propose automatic detection of lymphocyte leukemia through classification of lymphocyte cell images obtained from peripheral blood smear single cell. There are two main objectives in this study. The first is to extract featuring cells. The second objective is to classify the lymphocyte cells into two classes, namely normal and abnormal lymphocytes. In conducting this study, we use combination of shape feature and histogram feature, and the classification algorithm is k-nearest Neighbour with k variation is 1, 3, 5, 7, 9, 11, 13, and 15. The best level of accuracy, sensitivity, and specificity in this study are 90%, 90%, and 90%, and they were obtained from combined features of area-perimeter-mean-standard deviation with k=7.

No abstract available

Background: Peripheral blood smear examination (HDT) is a test to observe the morphology of blood cells microscopically. A peripheral blood smear is a simple method, and its examination is widely available in laboratories. The advantage of peripheral blood smear examination is its ability to assess various components of peripheral blood cells, such as cell morphology (erythrocytes, leukocytes, and platelets). One of the commonly used methods for peripheral blood smear examination is Giemsa staining. The process of Giemsa staining fixation aims to preserve the cell structure and its components before the staining is performed. The precise fixation time is key to obtaining optimal staining results, which allows for accurate observation of cell morphology. The aim of this study is to analyze the morphology of leukocyte profiles in peripheral blood smears stained with Giemsa based on variations in fixation time to determine the optimal time. Methods: This research was conducted in August-September 2024. The sample in this study was whole blood taken from the academic community of the Department of Medical Laboratory Technology and examined with a peripheral blood smear using Giemsa staining, and analyzed using descriptive statistical tests with the SPSS for Windows 22 software. Results: The research results showed that at a fixation time of 3 minutes, 60% were good and 40% were less good; at 5 minutes, 100% were good; at 10 minutes, 100% were good; at 15 minutes, 40% were good and 60% were less good; and at 20 minutes, 20% were good and 80% were less good. Conclusion: The best time variation in Giemsa staining using phosphate buffer is fixation for 5 minutes and 10 minutes, where all preparations (5 preparations) are 100% in the good category.

This paper investigates the integration of deep learning (DL) architectures with Recurrent Neural Networks (RNNs) for automated leukemia detection in Peripheral Blood Smear (PBS) images. Models such as DenseNet201, EfficientNetB3, Inception v3, InceptionResNetV2, MobileNetV2, ResNet50, VGG16, and Xception, each enhanced with a Gated Recurrent Unit (GRU) layer, are employed to capture spatial and temporal dependencies in image data. Through extensive experimentation on a standard leukemia dataset, the performance of each model is evaluated based on accuracy and computational efficiency. Among these, the Xception + GRU model achieves the highest classification accuracy, attaining an impressive 99.69%. This exceptional result underscores the efficacy of combining Convolutional Neural Networks (CNNs) with RNNs, particularly GRUs, in accurately detecting Leukemia from PBS images. The findings offer valuable contributions to medical image analysis, demonstrating the potential of DL techniques to enhance automated disease diagnosis. By advancing the precision of leukemia detection, this study provides promising implications for improving patient care and treatment outcomes.

Leukemia is a cancer of white blood cells that originates in the bone marrow due to dysregulated blood cell production. Manual microscopic diagnosis of bone marrow blood samples is time-consuming, error-prone, and subject to inter-observer variability, as differences in pathologists' expertise may lead to inconsistent diagnoses. Integrating artificial intelligence (AI) can improve diagnostic accuracy and efficiency. This study proposes a deep learning-based method using a convolutional neural network (CNN) to automate the classification of acute lymphoblastic leukemia (ALL) from microscopic images. A dataset of 770 images was collected from HUSM (Hospital Universiti Sains Malaysia). The images underwent preprocessing (resizing, augmentation) and were evaluated using three CNN architectures: AlexNet, VGG-16, and ResNet-18. The best-performing model was selected through hyperparameter tuning, optimizing the optimizer, initial learning rate, mini-batch size, and learning rate schedule. Performance was assessed using accuracy, precision, sensitivity (recall), and F1-score. Finally, the performance metrices for classification such as accuracy, precision, sensitivity and F1-score are used to evaluate the model effectiveness. ResNet-18 achieved the highest performance, with scores of 99.56% accuracy, 99.71% precision, 99.60% sensitivity, and 99.56% F1-score. This model will be integrated into a computer-aided diagnosis (CAD) system to assist hematologists in classifying ALL from microscopic blood samples.

Acute lymphoblastic leukemia (ALL), a hematologic malignancy characterized by the overproduction of immature lymphocytes, a type of white blood cell. Accurate and timely diagnosis of ALL is crucial for effective management. This article introduces a novel multi-task advanced convolutional neural network (MTA-CNN) framework for ALL detection in medical imaging data by simultaneously performing, expression classification, and disease detection. The MTA-CNN is based on a deep learning architecture that can handle multiple tasks simultaneously, allowing it to learn more comprehensive and generalizable features. With, expression classification, and disease detection tasks, the MTA-CNN effectively leverages the complementary information from each task to improve overall performance. The proposed framework employs CNNs to extract informative features from medical images. These features capture the spatial and temporal characteristics of the data, which are essential for accurate ALL diagnosis. The cascaded structure of the MTA-CNN allows the model to learn features at different levels of abstraction, from low-level to high-level, enabling it to capture both fine-grained and coarse-grained information. To ensure the reliability of the detection results, non-maximum suppression is employed to eliminate redundant detections, focusing only on the most likely candidates. Additionally, the MTA-CNN’s ability to accurately localize key facial landmarks provides valuable information for further analysis, including identifying abnormal structures or changes in anatomical features associated with ALL. Experimental results on a comprehensive dataset of medical images demonstrate the superiority of the MTA-CNN over other learning methods. The proposed framework achieved an accuracy of 0.978, precision of 0.979, recall of 0.967, F1-score of 0.973, specificity of 0.991, Cohen’s kappa of 0.979, and negative predictive value (NPV) of 0.990. These metrics significantly outperform baseline models, highlighting the MTA-CNN’s ability to accurately identify and classify ALL cases. The MTA-CNN offers a promising approach for improving the efficiency and accuracy of ALL diagnosis.

Leukemia diagnosis traditionally depends on time-intensive examination of blood cell morphology, a process prone to human error. To address these challenges, this study explores the use of convolutional neural networks (CNNs) optimized with the Tversky loss function for automated, multilevel image classification in leukemia diagnostics. The model was designed to tackle binary classification for distinguishing normal from abnormal cells, and multiclass classification for identifying leukemia subtypes, while addressing the challenges of imbalanced datasets inherent in medical imaging. Trained on publicly available leukemia image datasets, the CNN achieved high accuracy in both tasks, effectively capturing subtle morphological variations critical for precise diagnosis. By incorporating performance metrics such as accuracy, precision, and recall, the study highlights the model’s reliability and robustness across classification tasks. The findings underscore the potential of CNN-based tools in enhancing diagnostic accuracy and efficiency, paving the way for future innovations in leukemia diagnostics and broader medical imaging applications.

. Microscopic analysis of peripheral blood smears remains a critical and complex step in leukemia diagnosis, which could greatly benefit from automation using deep learning. In this paper, we compare three different deep learning models for automated classification of leukemia cells: a simple CNN, a ResNet, and a hybrid vision transformer. The Kaggle leukemia image dataset, which includes 15,135 blood smear images, was used. The blood smear images were preprocessed using denoising, normalization, upscaling, and upscaling. Training was performed on high-performance GPUs and evaluated on multiple complex metrics such as F-score, precision, recall, and accuracy. The expected outcomes include identifying the most robust and accurate deep learning model for leukemia classification, providing insights into the strengths and weaknesses of different leukemia subtypes, and demonstrating strategies and the effectiveness of image distortion handling. The results showed that ViT Hybrid models outperformed CNN and ResNet, achieving 89% of accuracy, 88% of precision, 90% of recall, and 89% of F-score.This suggests that hybrid structures hold great promise for improving computer-aided diagnosis in hematology. These findings are expected to contribute significantly to the field of medical image analysis, offering an accurate and scalable diagnostic tool with immediate clinical application.

Due to the disease's late discovery, the number of new cases of Acute lymphoblastic leukemia (ALL) is rising along with its high fatality rates. However, the traditional method of diagnosing Acute lymphoblastic leukemia often involves multiple tests and procedures that leads the patient into life-threatening stages while the interpretation of the results of physical examination, blood tests, and imaging studies is sometimes leads to less accuracy and can be subjective and vary between different healthcare professionals. With the advancements of AI and big data analytics, early diagnosis of Acute lymphoblastic leukemia can be used to aid the clinical decisions of physicians and radiologists. In this study, an automated blood sample classification system using CNN classifier, which is Xception, VGG16, AlexNet, SqueezeNet, MobileNetV2 and ResNet-18, united with the Stochastic Gradient Descent (SGD) optimization algorithm has been developed to predict and classify normal WBC and ALL subtypes from microscopic peripheral blood cell image. The evaluation metrics, confusion matrices, average confidence levels, ROC curves, and highest/lowest confidence levels for each class are analysed to assess the efficacy of each classifier. The result from comparing the performance of the classifier found that AlexNet demonstrated the high accuracy, precision, recall, and F1-score, with an accuracy of 98.9%. AlexNet also had high average confidence levels for each ALL subtype, indicating its strong understanding and discrimination of the different subtypes in classifying normal WBC and ALL subtypes.

The increasing demand for reliable and efficient diagnostic support in hematology has motivated the application of deep learning techniques to medical image analysis. Among blood disorders, leukemia remains a critical condition that requires timely and accurate classification of subtypes for effective treatment planning. In this study, we propose the design of a convolutional neural network architecture, referred to as ModifiedNet, aimed at classifying four leukemia subtypes plus healthy cells from peripheral blood smear images. The network is developed based on architectural principles drawn from AlexNet, GoogLeNet, and ResNet, incorporating depthwise separable convolutions, dilated convolutions, and residual connections to enhance feature extraction efficiency while maintaining considerable network depth with reduced parameter count. Training was performed on an augmented dataset using standardized hyperparameters to ensure fair comparison with baseline architectures. ModifiedNet achieved superior performance with 97.24% test accuracy, outperforming established models including AlexNet (94.22%), GoogLeNet (95.1%), and ResNet-18 (93.94%). The model demonstrated strong discriminative ability across all leukemia subtypes, with particularly robust performance in distinguishing healthy cells from malignant variants. These results validate the effectiveness of the proposed architectural modifications for automated leukemia classification tasks.

Acute myeloid leukemia (AML) is a heterogenous blood cancer with a dismal prognosis. It emanates from leukemic stem cells (LSCs) arising from the genetic transformation of hematopoietic stem cells (HSCs). LSCs hold prognostic value, but their molecular and immunophenotypic heterogeneity poses challenges: there is no single marker for identifying all LSCs across AML samples. We hypothesized that imaging flow cytometry (IFC) paired with artificial intelligence-driven image analysis could visually distinguish LSCs from HSCs based solely on morphology. Initially, a seven-color IFC panel was employed to immunophenotypically identify LSCs and HSCs in bone marrow samples from five AML patients and ten healthy donors, respectively. Next, we developed convolutional neural network (CNN) models for HSC-LSC discrimination using brightfield (BF), side scatter (SSC), and DNA images. Classification using only BF images achieved 86.96% accuracy, indicating significant morphological differences. Accuracy increased to 93.42% when combining BF with DNA images, highlighting differences in nuclear morphology, although DNA images alone were inadequate for accurate HSC-LSC discrimination. Model development using SSC images revealed minor granularity differences. Performance metrics varied substantially between AML patients, indicating considerable morphologic variations among LSCs. Overall, we demonstrate proof-of-concept results for accurate CNN-based HSC-LSC differentiation, instigating the development of a novel technique within AML monitoring.

Leukemia is a health condition in which the body produces too many abnormal white blood cells or leukocytes. Leukemia can affect both children and adults. Early diagnosis of leukemia faces significant challenges, as diagnostic methods are time consuming, require experienced medical experts, and are expensive. Previous studies have been conducted using deep learning approaches, but it is still rare to find a model that shows the best performance and uses optimization methods to classify leukemia diseases. Therefore, a Convolutional Neural Network (CNN) model with EfficientNetV2M architecture and Bayesian Optimization is proposed as the main method assisted by ImageDataGenerator in preprocessing. This study shows a significant impact of Bayesian optimization with good Accuracy, Precision, Recall and F1-Score results of 91.37%, 93.00%, 87.00%, 89.00%, respectively, which are expected to improve the performance of the model in previous studies in classifying leukemia diseases.

Among the most pressing issues in the field of illness diagnostics is identifying and diagnosing leukemia at its earliest stages, which requires accurate distinction of malignant leukocytes at a low cost. Leukemia is quite common, yet laboratory diagnostic centres often lack the necessary technology to diagnose the disease properly, and the available procedures take a long time. They are considering the efficacy of machine learning (ML) in illness diagnostics and that deep learning as a machine learning method is becoming critical. This study proposes a convolutional neural network (CNN) deep learning model for leukemia diagnosis utilizing the AML (acute myeloid leukemia) dataset. The classification using the proposed method achieved results that exceeded 98% accuracy, the sensitivity of 94.73% and specificity of 98.87%.

No abstract available

Acute lymphoblastic leukemia (ALL) is a life‐threatening disease that commonly affects children and is classified into three subtypes: L1, L2, and L3. Traditionally, ALL is diagnosed through morphological analysis, involving the examination of blood and bone marrow smears by pathologists. However, this manual process is time‐consuming, laborious, and prone to errors. Moreover, the significant morphological similarity between ALL and various lymphocyte subtypes, such as normal, atypic, and reactive lymphocytes, further complicates the feature extraction and detection process. The aim of this study is to develop an accurate and efficient automatic system to distinguish ALL cells from these similar lymphocyte subtypes without the need for direct feature extraction. First, the contrast of microscopic images is enhanced using histogram equalization, which improves the visibility of important features. Next, a fuzzy C‐means clustering algorithm is employed to segment cell nuclei, as they play a crucial role in ALL diagnosis. Finally, a novel convolutional neural network (CNN) with three convolutional layers is utilized to classify the segmented nuclei into six distinct classes. The CNN is trained on a labeled dataset, allowing it to learn the distinguishing features of each class. To evaluate the performance of the proposed model, quantitative metrics are employed, and a comparison is made with three well‐known deep networks: VGG‐16, DenseNet, and Xception. The results demonstrate that the proposed model outperforms these networks, achieving an approximate accuracy of 97%. Moreover, the model's performance surpasses that of other studies focused on 6‐class classification in the context of ALL diagnosis.

No abstract available

Leukemia is a kind of blood cancer that influences people of all ages and is one of the leading causes of death worldwide. Acute lymphoblastic leukemia (ALL) is the most widely recognized type of leukemia found in the bone marrow of the human body. Traditional disease diagnostic techniques like blood and bone marrow examinations are slow and painful, resulting in the demand for non-invasive and fast methods. This work presents a non-invasive, convolutional neural network (CNN) based approach that utilizes medical images to perform the diagnosis task. The proposed solution consisting of a CNN-based model uses an attention module called Efficient Channel Attention (ECA) with the visual geometry group from oxford (VGG16) to extract better quality deep features from the image dataset, leading to better feature representation and better classification results. The proposed method shows that the ECA module helps to overcome morphological similarities between ALL cancer and healthy cell images. Various augmentation techniques are also employed to increase the quality and quantity of training data. We used the classification of normal vs. malignant cells (C-NMC) dataset and divided it into seven folds based on subject-level variability, which is usually ignored in previous methods. Experimental results show that our proposed CNN model can successfully extract deep features and achieved an accuracy of 91.1%. The obtained findings show that the proposed method may be utilized to diagnose ALL and would help pathologists.

No abstract available

One of the lethal diseases that has a high mortality rate in both adults and children is acute lymphoblastic leukemia (ALL). A clinical pathologist examines the microscopic pictures of white blood cells to get the traditional diagnosis of this illness. Nevertheless, this method is based on manual observation and frequently yields unreliable outcomes. This study suggests combining a convolutional neural network (CNN) approach with a vision transformer (ViT) to create an automated system for diagnosing acute lymphoblastic leukemia. This study was conducted using the ALL IDB databases for this purpose. However, to address the overfitting issue in the model, data augmentation approaches have been used to create images. Using 8 heads and 8 transformer layers, the Concatenated CNN-ViT model produced encouraging results with a projection dimension of 32. Nonetheless, the outcomes demonstrated that, in the diagnosis of ALL, the proposed model achieved $\mathbf{9 8. 6 2 \%}$ accuracy, $\mathbf{9 8. 6 6 \%}$ precision, $\mathbf{9 8. 4 1 \%}$ recall, and $\mathbf{9 8. 5 3 \%}$ F1-score. The high accuracy indicates that compared to other studies published in the same field, it offers a more efficient method of diagnosing acute lymphoblastic leukemia.

No abstract available

Recent advancements in artificial intelligence (AI) have led to numerous medical discoveries. The field of computer vision (CV) for medical diagnosis has received particular attention. Using images of peripheral blood (PB) smears, CV has been utilized in hematology to detect acute leukemia (AL). Significant research has been undertaken in the area of AL diagnosis automation in order to deliver an accurate diagnosis. This study addresses the morphological classification of atypical white blood cells (WBCs), including immature WBCs and atypical lymphocytes, in acute myeloid leukemia (AML), as observed in peripheral blood (PB) smear images. The purpose of this work is to build a classification model for atypical AML WBCs based on their distinctive features. Using a hybrid model based on geometric transformation (GT) and a deep convolutional autoencoder (DCAE), this work provides a novel technique in the field of AI for resolving the issue of imbalanced distribution of WBCs in blood samples, nicknamed the “GT-DCAE WBC augmentation model”. In addition, to extract context-free atypical WBC features, this study develops a stable learning paradigm by incorporating WBC segmentation into deep learning. In order to classify atypical WBCs into eight distinct subgroups, a hybrid multiclassification model termed the “two-stage DCAE-CNN atypical WBC classification model” (DCAE-CNN) was developed. The model achieved an average accuracy of 97%, a sensitivity of 97%, and a precision of 98%. Overall and by class, the model’s discriminating abilities were exceptional, with an AUC of 99.7% and a class-wise range of 80% to 100%.

The image analysis of peripheral blood smear pictures has received considerable attention in the support of early leukemia diagnosis; nevertheless, most current research works in the domain of deep learning are dedicated to binary classification or the use of transfer learning without a task-specific construction. This article contains a tailored convolutional neural network (CNN) architecture to multi-classify microscopic blood cell images to four clinically-specific classes; benign, malignant pre-B, malignant pro-B, and malignant early pre-B. It is specifically designed to address morphological finer differences between leukemia subtypes with a structured convolutional block, batch normalization and regularization techniques to meet moderatesized medical datasets. It used a publicly available dataset consisting of 3,242 labeled images, where all the splits in the dataset were stratified: training, validation, and testing. The Adam optimizer was used to train the model with categorical crossentropy loss, and the evaluation of the results was done through accuracy, precision, recall, $\mathbf{F 1}$-score and a confusion matrix. The accuracy of the experiment outcomes is 97.9 percent, and the performance of each of the subtypes is balanced. The originality of this work consists in the architectural specificity of the tasks of discrimination of four-class leukemia subtypes, the emphasis on the complete reproducibility of the results, and the analysis of the results in relation to the classes. The results suggest that the suggested framework can be used as a valuable decision-support system to perform automated screening of leukemia in clinics.

Acute Lymphoblastic Leukemia (ALL) is cancer in which bone marrow overproduces undeveloped lymphocytes. Over 6500 cases of ALL are diagnosed every year in the United States in both adults and children, accounting for around 25% of pediatric cancers, and the trend continues to rise. With the advancements of AI and big data analytics, early diagnosis of ALL can be used to aid the clinical decisions of physicians and radiologists. This research proposes a deep neural network-based (ALNett) model that employs depth-wise convolution with different dilation rates to classify microscopic white blood cell images. Specifically, the cluster layers encompass convolution and max-pooling followed by a normalization process that provides enriched structural and contextual details to extract robust local and global features from the microscopic images for the accurate prediction of ALL. The performance of the model was compared with various pre-trained models, including VGG16, ResNet-50, GoogleNet, and AlexNet, based on precision, recall, accuracy, F1 score, loss accuracy, and receiver operating characteristic (ROC) curves. Experimental results showed that the proposed ALNett model yielded the highest classification accuracy of 91.13% and an F1 score of 0.96 with less computational complexity. ALNett demonstrated promising ALL categorization and outperformed the other pre-trained models.

Cancer occurs when healthy cells in the body grow abnormally and out of control. Leukemia is a type of cancer that affects White Blood Cells (WBCs) and can cause a lethal infection and early death. Identification and classification of different types of leukemia are performed manually and automatically. The doctors analyze blood samples under a microscope and consider any changes in the number and structure of WBCs as a sign of cancer in the manual method. It is a time-consuming, inaccuracy-prone process that depends on the expertise and skill of the physician and the type of laboratory equipment. In recent years, more automated methods of identifying and classifying leukemia have been developed with the help of Artificial Intelligence (AI) and Computer Vision (CV), with the aim of overcoming the challenges of manual approaches. This paper introduces two types of attention blocks, Parallel Cognitive Attention Block (PCAB) and Sequential Cognitive Attention Block (SCAB), to integrate into the architecture of any Convolutional Neural Network (CNN). Each of the proposed attention blocks is composed of the channel and spatial attention sub-blocks. They extract the structure and location of WBCs in the feature maps, similar to the ventral and dorsal streams in the human brain. The PCAB and SCAB were embedded in the architecture of the ResNet18 and MobileNetv4. The baseline and attention-based networks are trained, validated, and tested by two types of data splitting on the four leukemia datasets, including ALL, ALL-IDB2, C-NMC, and Mixture-Leukemi (ALL-IDB2+Munich AML Morphology), with the same experimental conditions for 30 epochs. The classification results demonstrate that the proposed model (MobileNetv4PCAB) achieved better performance metrics than others on all datasets in the test steps. It showed that the suggested model achieved the accuracy values of 100%, 100%, 93.61%, and 99.4%, and the F1-score values of 100%, 100%, 95.64%, and 99.3% with ALL, ALL-IDB2, C-NMC, and Mixture-Leukemia datasets, respectively. We confirmed that the proposed model outperforms existing state-of-the-art methods.

White blood cell (WBC) identification performance is highly correlated with the quality of the extracted features, with radiomic features having greater resolution and more detailed information and deep features having more robust semantic information. This research integrates these two aspects to creatively suggest a WBC classification model based on radiomics and deep learning. This research suggests a brand-new method for extracting radiomic features from WBC images as well as a dual-branch feature fusion network RCTNet based on CNN and Transformer for extracting deep features. The radiomic feature extraction method not only has a simple segmentation algorithm but also solves the problem of cell adhesion, can obtain higher quality shape, color, and texture features without segmenting intact cells, and is more generalizable. RCTNet can extract more critical and recognizable deep features from WBC nuclei, avoiding the influence of too much redundant and invalid information on the results, and has better performance than several existing CNN models. We compared the classification outcomes based on radiomics with those based on deep learning to confirm the efficacy of the WBC classification model based on radiomics and deep learning suggested in this research. The experimental results demonstrated that combining radiomic features and deep features significantly improved the classification accuracy, with an AUC exceeding 0.9995, accuracy, sensitivity, specificity, precision, and F1-score reaching 0.9880, 0.9823, 0.9883, 0.9968, and 0.9881, respectively. The model has significant research significance in clinical applications and aids physicians in improving diagnosis and screening for diseases of WBCs.

Acute Lymphoblastic Leukemia (ALL) is a highly aggressive form of cancer that originates in the bone marrow and is characterized by uncontrolled growth of immature white blood cells. The proposed work utilizes the various public single-cell datasets and employs Generative Adversarial Networks to generate synthetic images (DCGAN) as well as GAN for semantic segmentation achieving an accuracy of 98.79% for healthy cell segmentation and 99.59% for blast cell segmentation. The work further explores the impact of segmentation on classification accuracy, integrating three classification models (RegNet-Y, ConvNeXt, Vision Transformer). Without segmentation, these models achieve accuracies of 67.16%, 82.19%, and 86.76%, respectively. Using the GAN segmentation outputs, notable improvements in accuracy are observed, with results reaching 72.63%, 93.79%, and 90.77%. This highlights the benefits of combining precise segmentation and robust classification models.

No abstract available

Segmentation of white blood cells is critical for a wide range of applications. It aims to identify and isolate individual white blood cells from medical images, enabling accurate diagnosis and monitoring of diseases. In the last decade, many researchers have focused on this task using U-Net, one of the most used deep learning architectures. To further enhance segmentation accuracy and robustness, recent advances have explored the combination of U-Net with other techniques, such as attention mechanisms and aggregation techniques. However, a common challenge in white blood cell image segmentation is the similarity between the cells’ cytoplasm and other surrounding blood components, which often leads to inaccurate or incomplete segmentation due to difficulties in distinguishing low-contrast or subtle boundaries, leaving a significant gap for improvement. In this paper, we propose GAAD-U-Net, a novel architecture that integrates attention-augmented convolutions to better capture ambiguous boundaries and complex structures such as overlapping cells and low-contrast regions, followed by a gating mechanism to further suppress irrelevant feature information. These two key components are integrated in the Double U-Net base architecture. Our model achieves state-of-the-art performance on white blood cell benchmark datasets, with a 3.4% Dice score coefficient (DSC) improvement specifically on the SegPC-2021 dataset. The proposed model achieves superior performance as measured by mean the intersection over union (IoU) and DSC, with notably strong segmentation performance even for difficult images.

This work utilizes a U-Net convolutional neural network for the segmentation of white blood cell (WBC) components, specifically targeting the nucleus and cytoplasm. This work utilizes a U-Net convolutional neural network for the segmentation of white blood cell (WBC) components, specifically targeting the nucleus and cytoplasm. Accurate WBC segmentation is challenging due to differences in cell shape, size, and staining quality. The segmented regions are further used to compute the cytoplasm-to-nucleus (C/N) ratio, which plays a vital role in medical diagnostics. Input images are pre-processed through normalization and resized to a standard dimension of 256 × 256 pixels. Batch normalization is applied to enhance model stability and convergence. The model is trained and deployed using Google Colab, achieving an accuracy of 80%. The proposed framework provides an effective solution for automated analysis of WBC images.

White blood cell image segmentation plays a vital role in the accurate analysis and diagnosis of blood‐related diseases, facilitating the identification and quantification of white blood cells (WBCs) in microscopic images. This process is essential for early disease detection, treatment monitoring, and immune response studies, ultimately supporting clinical decision‐making. In this paper, we propose an enhanced approach based on the segment anything model (SAM). First, contrast limited adaptive histogram equalisation is applied for preprocessing to enhance the features of WBCs. Then, SAM is utilised for segmentation. Experimental results demonstrate that our method achieves state‐of‐the‐art performance on cross‐domain datasets, providing accurate and reliable segmentation of WBCs.

White Blood Cells (WBC) play a crucial role in the immune system, and abnormalities in their count, shape, or distribution can indicate various diseases, including infections, autoimmune disorders, and leukemia. Accurate WBC segmentation is essential for quantifying various features of WBCs, such as size, shape, texture, and distribution, which are used as diagnostic indicators by computer-aided systems. However, accurate WBC segmentation from microscopic blood smear images is tedious, even for contemporary deep-learning models. This manuscript proposes an encoder-decoder-based model named SE-UNet, which combines the power of the squeeze-and-excitation (SE) module with the baseline UNet. Adding Squeeze and Excitation modules in the encoder section of U-Net enhances segmentation by incorporating channel-wise attention, extracting more prominent features from whole blood microscopic images, and aiding generalization without significantly increasing computational complexity. The proposed model achieves a 96.5% (0.965) precision value, 93.9% (0.939) recall value, and 95.2% (0.952) F1-score value for the WBC segmentation task on a private dataset obtained from AIIMS Patna. The SE-UNet outperforms the baseline U-Net by about 1.2% in terms of F1 score.

The accurate segmentation and classification of white blood cells (WBCs) is of utmost importance in diagnosing and monitoring different hematological disorders. In this paper, we present a unifying deep learning framework that takes care of the two tasks concurrently with the help of a Convolutional Neural Network (CNN) based architecture. A shared feature extractor under the idea of VGG was designed into the model to learn rich hierarchical representations from peripheral blood smear images. It lets the model branch into two task-specific heads: a segmentation head that produces pixel-level masks through deconvolutional layers, and a classification head that identifies WBC types using global average pooling followed by fully connected layers. A composite loss function for joint optimization of the model combining Dice loss, cross-entropy and Intersection over Union (IoU) loss for segmentation, and categorical cross-entropy for classification is applied for the optimization of both tasks. Results from experiments show how effective the proposed method is by exhibiting high accuracy in segmentation and classification. A reliable one-stop solution for quick and efficient automated WBC analysis in clinical settings.

To develop a methodology for rapid white blood cells (WBCs) classification of whole slide images (WSIs) acquired using an optical microscope. This approach resolves multiple issues faced by current classification methods, including poor robustness, low accuracy, and long processing times, by using a segmentation method based on stain separation and a classification method based on a lightweight convolution neural network (CNN).The suspected WBC region is obtained from WSIs using Otsu's threshold segmentation technique. An included color deconvolution applies stain vectors to digitally separated eosin stains from suspected WBC regions and identifies potential WBCs, which are then classified by the proposed lightweight CNN model using fine-grained feature extraction.WBC regions were accurately segmentated from WSIs, using the proposed color deconvolution and segmentation technique. The Classification algorithm achieves an average accuracy of 98.5%. The model required the use of 1.498×106 parameters and achieved a prediction speed of 3.45 ms/image, which is a significant improvement over four other state-of-the-art WBC classification models.This approach can be directly applied to unprocessed WSIs, which could improve diagnostic speed while maintaining high accuracy.

Medical image segmentation has numerous applications in diagnosing different diseases. Various types of diseases are found in white blood and Red blood cells. This paper represents the segmentation of WBCs from blood smear images. It is a complex and challenging task due to the frequent overlapping and variants in size and shape of WBCs with each other and RBCs. This overlapping is due to the rough border of the immature cells. The paper describes a new approach to WBC segmentation using UNet++, the marker watershed algorithm, and Neural Ordinary Differential Equations (ODE). This technique uses UNet++ for pre-segmentation, followed by the marker watershed method, which has been integrated using ODE to deepen the segmentation process. This novel integration enhances clinical applications in automated blood cell analysis, diagnostic imaging, and disease monitoring, improving accuracy and robustness. The ODE is used after the convolution operation to reduce the error at each step, preventing the massive propagation of error in the forward and the backpropagation. The White blood cells are segmented from the input smear images using ALL_IDB1 and ALL_IDB2 datasets, which are further used in the experiment section. UNet ++ is used to generate the pre-segmented probabilistic grayscale images. Some white blood cells are connected and make groups appearing in the grayscale images. These groups of WBCs are separated using a technique called the marker watershed, which gives us the final segmented result. The experimentation results show that the mean intersection over union (Jaccard method), the Dice similarity coefficient, and the mean pixel accuracy are 97.73%, 98.36%, and 98.97%, respectively. The structure and size of the white blood cells vary from red blood cells and platelets, which makes this work different from others. Furthermore, the combination of UNet++, marker watershed, and Neural Ordinary Differential Equation makes the proposed system unique from existing systems. This work can be further investigated to reduce computational complexity and memory space for optimizing deployment on low-resource devices, such as smart healthcare systems. Techniques like model pruning, quantization, or learned information distillation might be explored to create a lightweight version of the model without much loss in accuracy. Such developments would make possible mass uses of automated white blood cell segmentation in portable, low-cost health devices for point-of-care remote diagnostics and monitoring.

White blood cell (WBC) segmentation is a critical step in automated hematological analysis, aiding in disease diagnosis and treatment planning. Unlike usual traditional works, this work proposed a deep learning framework for classifying white blood cells, which can be one of the major impact factors in the field of morphological cell identification in various classed cells. The proposed approach involves pre-processing of images and mapping with their respective labels. With the help of masking techniques, we first extract the white blood cell from original image. The segmented image is passed to VGG16 model, which analyses the image, gives a clear analysis of the feature of the particular image, like high level features, spatial arrangement etc, On the other hand, Handcrafted features like Area, Perimeter, Solidity, Aspect Ratio and Extent are also calculated making it a hybrid dual-feature strategy. With LightGBM, we classify the processed image into particular cell class. Classification metrics have been developed and the model is evaluated by the accuracy score and classification report. By this proposed framework, identification and classification of cell architecture has become more efficient and easier. This work aims to assist researchers and pathologists in the efficient and precise identification of white blood cells for fast diagnostic and clean medical workflow.

No abstract available

No abstract available

No abstract available

Accurate and timely classification of white blood cells (WBCs) is crucial for diagnosing a myriad of hematological disorders, including leukemia. While deep learning models, particularly Convolutional Neural Networks (CNNs), have shown promise in automating this task from microscopic blood smear images, their performance can be hindered by complex backgrounds and intra-class variations. This paper proposes a novel segmentation-enhanced classification framework that synergistically combines classical image processing with deep learning. Our approach first employs a fixed-parameter Canny edge detection and contour-based algorithm to segment the WBC foreground. Subsequently, a learnable blending layer intelligently fuses the segmented foreground with the original image, allowing the downstream CNN to leverage both focused object information and contextual cues. We meticulously document our experimental journey, including initial attempts to train Canny parameters which proved unstable. The proposed model, featuring fixed segmentation and a learnable blending factor ( α ), was evaluated on a public blood cell image dataset for cancer detection. It achieved a test accuracy of 99.93%, outperforming a baseline CNN (99.07%). The blending factor α converged to approximately 0.63, indicating an optimal balance between the segmented and original image content. Furthermore, the segmentation-enhanced model demonstrated faster inference times. This work underscores the potential of hybrid approaches, particularly the utility of a learnable blending mechanism to effectively integrate classical segmentation outputs into neural networks for improved classification.

Microscopic image examination is essential for medical diagnostics to identify anomalies using cell counts based on morphology. Sickle Cell Disease (SCD) is an inherited blood condition characterized by defective hemoglobin, leading to severe anemia and complications. Detecting sickle cells in blood smears is essential, but the presence of White blood cells (WBCs) and platelets often leads to miscounting as they are classified incorrectly as red blood cells (RBCs). This study proposed an approach for segmenting WBCs and platelets by resembling the human color recognition process to differentiate the regions for accurate identification. First, the RGB color space is converted to RG chromaticity to locate WBCs and platelets with high pixel chromatic variance. Parametric segmentation is applied to the RG chromaticity images to identify the appropriate chromaticity channel for segmentation based on probability distribution values. The optimal threshold values have been determined using Particle Swarm Optimization (PSO) by dynamically narrowing the search space using values obtained through manual experimentation ranging from 0.001 to 1. This systematic process effectively identifies and segments platelets and WBCs, ensuring that overlapping platelets and WBCs are accurately segmented. Compared to state-of-the-art techniques, the proposed approach achieved an accuracy of 96.32 %, 96.97% for sensitivity, 96.96 % for precision and 97.46% for F- score in the pixel-wise segmentation of WBCs and platelets.

White blood cell (WBC) analysis plays a crucial role in diagnosing and treating various diseases. The research utilizes the HSV color space technique along with conventional machine learning approaches, including the Random Forest Classifier and Support Vector Machine (SVM), to differentiate various kinds of white blood cells (WBCs) and measure both their quantity and area. The segmentation results of these three methods are compared, and it is observed that the HSV color space model efficiently segments the region of interest of blood cells. In the context of traditional machine learning, the segmentation of white blood cells (WBCs) is evaluated using metrics such as the Dice Coefficient and the Jaccard Index. According to the findings, the Random Forest model attained a Dice Coefficient score of 90.4% and a Jaccard Index score of 82.5%. In contrast, the Support Vector Machine (SVM) model produced a Dice Coefficient score of 79.7% and a Jaccard Index score of 66.3%. The proposed methods demonstrate high accuracy and effectively segment diverse types and shapes of white blood cells. Furthermore, they facilitate the analysis of a significant numbers of white blood cells within a brief timeframe.

Qualitative analysis of blood disorders is the most challenging job in current medical image. Diagnosing disorders through qualitative analysis relies heavily on approximation. Disorders related to white blood cells (WBC) are notably prevalent in medical practice. Identifying blood disorders leads to classification of various blood related conditions. Hence, the automated identification of blood related disorder allows to bypass the existing complex environment and focus on the complex image insight provided by the images. In this study, the automated segmentation of neutrophil nucleus in white blood cells microscopic through Kapur’s entropy multilevel thresholding technique. In the process of an efficient segmentation through Kapur’s entropy, the threshold optimal values are required to change the class variance or entropy criteria. To achieve these optimizations, Improved Grey Wolf (IGWO) nature-inspired metaheuristic algorithms are employed for this study. The experimental test has conducted on standard images using various levels of threshold (2, 3, 4, 5 and 6). The efficacy of Kapur’s entropy based improved grey wolf optimization approach has been evaluated and performed with established optimization methods contains Genetic Algorithm(GA), Moth Fly Optimization (MFO), Firefly Optimization, Fire Hawk Optimization and Particle Swarm Optimization (PSO). Results demonstrated that Kapur’s entropy based IGWO outperformed for neutrophil nucleus segmentation in microscopic images when compared to other optimization techniques in both quality and consistency.

Childhood leukaemia demands meticulous blood cell analysis for diagnosis, focusing on morphological irregularities like asymmetry and abnormal cell counts. Traditional manual diagnosis via microscopic blood smear images suffers from reduced reliability, time intensiveness, and observer variability. Computer-aided diagnostic (CAD) systems address these challenges. Integrating real-time image pre-processing and segmentation ensures swift operation, reducing the CAD system processing time. This enhances its overall effectiveness, enabling timely medical intervention and better patient outcomes. This study aims to simplify the algorithmic complexity of pre-processing steps, including bilateral filtering and Contrast-Limited Adaptive Histogram Equalization (CLAHE), alongside the segmentation stage involving morphological operations and the watershed algorithm. This work proposes a parallel implementation utilizing OpenMP and CUDA, evaluating its performance using accuracy and Intersection over Union (IoU) metrics along with computing time and algorithmic complexity. It highlights the benefits of parallel processing in enhancing efficiency and and accuracy in blood cell analysis.

No abstract available

Accurate white blood cell (WBC) classification from stained microscopy images is critical for hematological diagnosis but is impeded by stain variability and class imbalance. This work aims to deliver a lightweight, stain-robust, and explainable WBC classifier with a publicly available demo to support reproducible evaluation. We normalize input images using Vahadane stain normalization to reduce color heterogeneity and apply targeted data augmentation to address class imbalance. An EfficientV2-B1 backbone (≈ 7 million parameters) is trained on the augmented dataset, and model decisions are interpreted using Grad-CAM heatmaps. Statistical robustness is quantified through bootstrap resampling with 1,500 samples to generate 95% confidence intervals for accuracy. The resulting model attains 98.69% test accuracy, with bootstrap analysis indicating a 95% confidence interval of 98.22-99.14%, and Grad-CAM visualizations corroborate that predictions rely on biologically plausible regions. We deploy an interactive web application that displays the original image, the Vahadane-normalized image, Grad-CAM overlays, and per-class confidence scores to promote transparency and practical use. Overall, the proposed pipeline combines stain normalization, augmentation, an efficient architecture, rigorous validation, and explainability to produce a robust, highperforming WBC classification system suitable for research and clinical translation.

Accurate segmentation and classification of white blood cell (WBC) are essential for clinical hematology, yet remain challenging due to staining variability, complex backgrounds, and class imbalance. This study introduces an explainable, saliency-enhanced deep learning framework designed to achieve stain-robust leukocyte analysis. The framework integrates a saliency-driven preprocessing module, a lightweight EfficientSwin hybrid backbone, and a ResNeXt-CC–inspired cross-layer feature fusion block to capture complementary fine-grained and global features. A multi-task head jointly performs WBC segmentation and subtype classification, while a saliency-alignment loss enforces consistency between learned attention and saliency priors, providing training-time interpretability rather than post-hoc visualization alone. SG-CLDFF was evaluated on three public datasets (BCCD, LISC, ALL-IDB) and further tested under cross-stain and cross-dataset conditions. The framework achieved 95.8% accuracy, 0.94 F1-score, and 0.82 IoU, improving over strong CNN and transformer baselines. Ablation studies confirmed that both saliency preprocessing and cross-layer fusion contribute independently to performance, with saliency alignment yielding ≥2 IoU improvement in cross-stain scenarios (p < 0.05). Qualitative results using saliency maps and Grad-CAM demonstrate focused attention on diagnostically meaningful regions. These findings validate SG-CLDFF as a robust, interpretable, and stain-resilient solution for automated WBC analysis, offering a practical foundation for deployment in digital hematology workflows.

Blood cell classification is a crucial task in medical diagnostics. This work can identify different blood cell types that play essential roles in the immune system and overall body function. This research introduces a custom Convolutional Neural Network (CNN) model with SHAP for explainable multiclass classification of blood cell images. We classify eight types of blood cells—basophil, eosinophil, erythroblast, immunoglobulin (Ig), lymphocyte, monocyte, neutrophil, and platelet. An ablation study was conducted to optimize the CNN architecture. We compared its performance with eleven transfer learning models, including VGG16, ResNet101, and EfficientNet variants. Our model achieved 92.60% accuracy while maintaining lower complexity. The use of SHAP provides transparency by highlighting crucial features in the model’s decision-making process.

Motivation: In the present times, several classification frameworks have been discovered with the involvement of advanced learning technologies. However, the existing models needed an improved version to tackle the complexity related to slow convergence and poor generalization. Hence, it is crucial to mitigate the limitations that take place in the conventional blood cell categorization framework. Methodology: So, in this work, a novel explainable AI-based blood cell categorization framework is suggested. In beginning, necessary data’s are gathered in the conventional resource. Futher, the gathered data’s are used to execute a blood cell classification procedure. Here, an Adaptive Explainable Pyramid Dilated-Convolutional Neural Network (AEPD-CNN) is used to carry out catrgorization. Moreover, the parameters in AEPD-CNN are tuned through American Zebra Optimization (AZO). Finally, the blood cell classified results are obtained from AEPD-CNN, and then various experimental validations are accomplished in an implemented scheme over an existing mechanism by considering various performance measures. Major Results: The developed AZO-AEPD-CNN model significantly outperformed the comparative methods, demonstrating superior accuracy by achieving a 6.6% improvement over CSA-AEPD-CNN, 5.5% over DO-AEPD-CNN, 4.4% over DFA-AEPD-CNN, and 3.3% over FFO-AEPD-CNN.

Microscopic analysis of blood smears remains a gold standard in hematological diagnosis, yet it is traditionally manual, slow, and prone to inter-observer variability. This work introduces an automated blood cell classification model leveraging the ResNet50 architecture, combined with explainable artificial intelligence (XAI) via SHAP (SHapley Additive exPlanations) to enhance diagnostic transparency and reliability. The dataset used, the Blood Cells Image Dataset, consists of $\mathbf{1 7, 0 9 2}$ highquality annotated microscopic images, covering six cell types: basophils, eosinophils, erythroblasts, lymphocytes, monocytes, and platelets. Data preprocessing included resizing, ImageNet normalization, and controlled augmentation to preserve morphological characteristics. The model, adapted for multiclass classification via transfer learning, incorporated regularization strategies such as L1/L2 penalties, dropout, early stopping, and ReduceLROnPlateau to optimize training and prevent overfitting. Performance metrics, including global accuracy, class-specific precision, recall, $\mathbf{F 1}$-score, and confusion matrices, were utilized. The model achieved an impressive accuracy of 97.33 %, with F1-scores exceeding 0.95 across all classes. SHAP visualizations aligned with classic hematological criteria, reinforcing the clinical reliability of the system.

White blood cell (WBC) classification and morphological attribute prediction are critical for automated hematological analyses. To provide detailed and interpretable predictions, this paper proposes a multimodal visual-language embedding learning approach based on the contrastive language image pretraining (CLIP) model for WBC classification and attribute prediction. First, structured natural language prompts are created around WBC types and morphological attributes to offer rich semantic context that enhances the processing of multimodal inputs. Moreover, a joint-task optimization strategy is introduced to align the generated encodings from the WBC images with their corresponding structured text prompts in a shared semantic space, thus improving interpretability and prediction accuracy. Furthermore, a multi-task loss function with an adaptive weighting mechanism is employed to address class imbalance, effectively balancing classification and attribute prediction tasks to boost model performance. Experimental evaluations on publicly available datasets demonstrate that the proposed method achieves state-of-the-art performance in both WBC classification and attribute prediction.

Finding and monitoring many diseases, including infections and leukaemia, depend on the classification of white blood cells (WBCs). Conventional techniques mostly rely on seeing samples under a microscope and calling for an expert to evaluate the outcomes. This method takes a lot of time and human mistake might create errors. This work uses a lightweight convolutional neural network (CNN) to provide a basic model to enhance the automated categorisation of white blood cells (WBCs) via deep learning. The proposed paradigm maximises computational efficiency, hence it is perfect for real-time applications and environments with limited resources. We kept the model correct while still designing a simplified form that would make using it easier. Trained and validated on a whole dataset including diverse microscopy images of WBCs, the model guarantees stability and generalizability throughout many imaging environments. Our findings reveal that, in terms of sorting accuracy, speed of processing, and operational efficiency, we have achieved significant progress over conventional techniques. The minimal weight of the device makes it simple to use in internet services and mobile health applications, therefore enabling more individuals to get necessary medical attention. By means of concepts on building efficient deep learning models for healthcare, this work provides a beneficial tool for better categorising white blood cells and also helps enhance medical photographs.

Accurate classification of white blood cells (WBCs) from peripheral blood smears is essential for early detection of hematological disorders. We propose a lightweight convolutional neural network (CNN) that achieves high efficiency and interpretability, tailored for use in resource-constrained clinical environments. The model was trained and evaluated on a publicly available dataset of four major WBC types: neutrophils, eosinophils, monocytes, and lymphocytes. Despite its compact architecture, it attained an overall accuracy of 84%, with strong performance for lymphocytes (F1 = 0.96) and monocytes (F1 = 0.89). To enhance clinical transparency, we incorporated Grad-CAM and Grad-CAM++ visualizations, producing class-discriminative attention maps that align closely with expert morphological interpretations. Compared to deeper CNNs such as VGG16 and DenseNet121 (8–25 M parameters), the proposed lightweight model contains only 0.81 M parameters while achieving 84% classification accuracy and providing explicit interpretability through integrated Grad-CAM and Grad-CAM++ visualizations. This balance between efficiency and explainability represents a practical step toward real-time, interpretable AI-assisted hematology.

White blood cells (WBCs) are a vital component of the immune system. The efficient and precise classification of WBCs is crucial for medical professionals to diagnose diseases accurately. This study presents an enhanced convolutional neural network (CNN) for detecting blood cells with the help of various image pre-processing techniques. Various image pre-processing techniques, such as padding, thresholding, erosion, dilation, and masking, are utilized to minimize noise and improve feature enhancement. Additionally, performance is further enhanced by experimenting with various architectural structures and hyperparameters to optimize the proposed model. A comparative evaluation is conducted to compare the performance of the proposed model with three transfer learning models, including Inception V3, MobileNetV2, and DenseNet201.The results indicate that the proposed model outperforms existing models, achieving a testing accuracy of 99.12%, precision of 99%, and F1-score of 99%. In addition, We utilized SHAP (Shapley Additive explanations) and LIME (Local Interpretable Model-agnostic Explanations) techniques in our study to improve the interpretability of the proposed model, providing valuable insights into how the model makes decisions. Furthermore, the proposed model has been further explained using the Grad-CAM and Grad-CAM++ techniques, which is a class-discriminative localization approach, to improve trust and transparency. Grad-CAM++ performed slightly better than Grad-CAM in identifying the predicted area's location. Finally, the most efficient model has been integrated into an end-to-end (E2E) system, accessible through both web and Android platforms for medical professionals to classify blood cell.

No abstract available

No abstract available

Malaria diagnosis based on microscopic examination of blood smears is time-consuming and highly dependent on skilled laboratory personnel, which limits its scalability in resource-constrained environments. This study investigated whether an explainable deep learning approach could provide reliable and interpretable malaria blood cell classification using a convolutional neural network based on the DenseNet121 architecture combined with Gradient-weighted Class Activation Mapping to visualize the image regions influencing model predictions. Five-fold cross-validation was applied to ensure a stable and unbiased performance evaluation. The model achieved a mean classification accuracy of 0.8285 with low variation across folds, and the precision, recall, and F1-score values were balanced between the parasitized and uninfected classes. Visual explanations consistently highlighted intracellular regions associated with parasite presence in infected cells and more uniform cytoplasmic regions in uninfected samples, indicating that the network learned the biologically meaningful features of the cells. The results demonstrated that DenseNet121 provided a stable and interpretable solution for malaria blood cell classification when supported by a visual explanation, thereby enabling transparent automated screening. The proposed framework is suitable for integration into smart healthcare and medical informatics systems, where both predictive reliability and interpretability are required.

Automated blood cell classification from microscopic images is an important task in biomedical image analysis, particularly for developing efficient and accessible decision-support systems. However, achieving high classification performance on low-resolution biomedical datasets remains challenging due to limited spatial details, morphological similarities among cell types, and the computational cost of large-scale deep learning models. In this study, a lightweight and parameter-efficient convolutional neural network architecture, named UP-CBAM-CNN, is proposed for the classification of peripheral blood cell images using the BloodMNIST dataset. The proposed architecture combines multi-scale parallel convolutional branches, residual learning, and Convolutional Block Attention Module (CBAM) based channel and spatial attention mechanisms. Ten candidate CNN designs were collected into a search space and ranked against one another via Harris Hawks Optimization (HHO), so that the configuration delivering the strongest results could be isolated. In addition, a Ranking Stabilization Epoch (RSE) strategy based on Spearman rank correlation was introduced to analyze when the model ranking became stable during training. By conducting a systematic parameter optimization across a search space of ten distinct CNN architectures, the most efficient model, named UP-CBAM-CNN, was identified as the optimal configuration. This model was trained from scratch using AdamW optimization, OneCycleLR scheduling, label smoothing, gradient clipping, and five-view Test-Time Augmentation (TTA). Experimental results on the BloodMNIST test set show that UP-CBAM-CNN achieved 98.04% accuracy, $97.91 \%$ precision, $97.85 \%$ recall, $97.88 \%$ F1-score, and $99.89 \%$ ROC AUC with only approximately 370 K parameters. These findings indicate that the proposed CNN architecture provides a highly competitive and computationally efficient alternative for low-resolution BloodMNIST classification.

White blood cell (WBC) classification is a clinically critical task in hematology, underpinning the diagnosis of blood malignancies, including leukemia, as well as infections and immune disorders. Accurate automated WBC differential counting is essential for early detection of conditions such as chronic lymphocytic leukemia (CLL) and acute myeloid leukemia (AML), where abnormal leukocyte counts and morphology are primary diagnostic indicators. Manual differential counting is time-consuming, subjective, and prone to inter-observer variability. This paper proposes a weighted ensemble framework combining three lightweight pretrained convolutional neural networks (CNNs): EfficientNet-B0, MobileNetV3-Small, and MobileNetV3-Large for automated four-class WBC classification on Paul Mooney’s Blood Cell Images dataset (BCCD, Kaggle). A standardized preprocessing pipeline was employed to enhance nuclear contrast and normalize photometric variability. Each backbone was finetuned under a two-phase transfer-learning protocol with AdamW optimization and cosine annealing scheduling. Model evaluation was conducted via stratified 3-fold cross-validation, and ensemble predictions were aggregated using accuracy-weighted soft voting. The proposed ensemble achieved a classification accuracy of 98.57%, a weighted F1-score of 0.9857, a Matthews Correlation Coefficient (MCC) of 0.9810, and a mean AUC of 0.9994. These results demonstrate the effectiveness of ensemble-based transfer learning combined with targeted preprocessing for robust, well-calibrated WBC classification, with direct applicability to clinical laboratory automation.

Automated white blood cell (WBC) classification is important for supporting early screening of haematological disorders, but many deep learning approaches depend on large backbones and are insufficiently evaluated under variations in staining protocols, microscope settings, and dataset curation. To address this limitation, this study investigates cross-dataset robustness using two widely used WBC datasets, PBC-Normal and Raabin-WBC, together with their combined distribution, where heterogeneous imaging conditions and class imbalance make five-class WBC recognition challenging. The aim is to develop a computationally efficient and reproducible classification framework while systematically examining the role of preprocessing and backbone design for Basophil, Eosinophil, Lymphocyte, Monocyte, and Neutrophil classification. We present ECA-CNN-Lite, a 2.8 M-parameter residual CNN that integrates the established Efficient Channel Attention mechanism for channel-wise feature recalibration, and evaluate it under four preprocessing conditions: not cropped and unprocessed, not cropped and processed, cropped and unprocessed, and cropped and processed. The model is trained on 224 × 224 RGB images using AdamW optimization and cosine learning-rate decay, and is benchmarked against CNN and CNN Transformer baselines, including ConcatNeXt, VGG16-ViT, ResNet50-ViT, and VGG19. On individual datasets, ECA-CNN-Lite achieves 0.99 accuracy on PBC-Normal and 0.97 accuracy on Raabin-WBC, outperforming the strongest baseline by up to 0.04 absolute accuracy. On the combined dataset, it obtains 0.97 accuracy under the cropped and processed setting, with class-wise F1-scores of at least 0.95 across all five WBC types. Training-curve and F1-heatmap analyses indicate faster convergence, reduced sensitivity to preprocessing variation, and more balanced performance across rare and frequent classes compared with the evaluated baselines. These findings suggest that ECA-CNN-Lite, when combined with controlled cropping and stain-normalisation, provides a practical lightweight reference framework for cross-dataset WBC classification under the evaluated benchmark conditions.

Proper classification of the types of blood cells is an important part of automated hematological diagnostic procedures and is an essential step towards early diagnosis and recognition of blood related disorders. This paper presents a new hybrid Convolutional Vision Transformer model named CBC-ViT which is expected to classify five types of blood cells (basophil, erythroblast, monocyte, myeloblast, neutrophil). The proposed to be designed architecture incorporates a CNN backbone that implements hierarchical features extraction and the trainable panel of Vision Transformer (ViT) encoder that relates on multi-head self attention to obtain the dependencies and contextual features that are long-range. Adam optimizer with finely tuned learning parameters is used to optimize the model to have a stable convergence. The experimental outcomes indicate that CBC-ViT works better with a test accuracy of 99.20 and all the measures (precision, recall, and F1-scores) are above 99 percent and this indicates the strength and dependability of the proposed solution. To further guarantee transparency and Trustworthiness, we implemented Explainable AI (XAI) methods, such as Grad-CAM and Integrated Gradients and SmoothGrad, which offer explanations of the prediction at the model level in visual form and will hence help medical practitioners to justify automated predictions. The findings highlight the perspective of CBC-ViT as a predictable and elucidable decision-making instrument in the hematological diagnostics.