脑电在风味研究中的y33

对面孔情绪的识别是人们日常社交活动的一项重要能力。联觉是感觉传递理论中的一种，即一种刺激在引起该通道感觉的同时引起了另一种通道的感觉。在大众的意识中，大脑对面部的识别是基于面部的视觉信息，但其实来自味觉、嗅觉、听觉等的信息也可能对面孔情绪识别产生影响。前人研究结果表明，气味可以影响面孔信息的加工和情绪的识别，如好闻的气味与积极情绪相连，难闻的气味与消极情绪相连。相较于其他感觉的跨通道研究，对味觉的研究起步较晚，但结果有限，主要集中在行为学和脑成像方面，因此对味觉的跨通道研究需进一步深入。文章从视觉、嗅觉、听觉与面孔情绪识别的联系到味觉影响情绪面孔识别来论证分析，最后再提出已知研究的不足和未来展望。

No abstract available

Excessive salt intake is a pressing food health issue, and odor-induced saltiness enhancement (OISE) is a novel strategy for targeted salt reduction. Understanding the neural mechanisms of OISE is essential for salt reduction. In this study, the mechanism of saltiness enhancement induced by three volatile organic compounds (VOCs) identified in Jinhua dry ham was investigated in 20 panelists using electroencephalography (EEG). The study demonstrated that VOCs enhanced salty taste perception, primarily through low-frequency brain waves. Source localization revealed occipital lobe activation during salty taste recognition, while OISE stimuli enhanced activity in the primary and secondary gustatory cortices. Additionally, VOCs enhanced phase synchronization among activated brain regions, as indicated by functional connectivity. This study enhances the understanding of olfactory-gustatory interactions and provides a neurological basis for the effects of OISE.

Purpose Eating time and sleep habits are important modifiable behaviors that affect metabolic health, but the relationship between food intake and sleep remains incompletely understood. Observational data suggest that late food intake is associated with impaired sleep quality. We examined the effect of routine dinner (RD, 5 hours before bedtime) vs late dinner (LD, 1 hour before bedtime) on sleep architecture in healthy volunteers. Participants and Methods This was a post hoc analysis of a randomized crossover study of RD vs LD with a fixed sleep opportunity in a laboratory setting. On each of the two visits, 20 healthy adult volunteers (10 women) received an isocaloric meal followed by overnight polysomnography. Sleep architecture over the course of the night was assessed using visual sleep staging and EEG spectral power analysis and was compared between RD and LD. We modeled the proportions of spectral power in alpha, beta, delta, and theta bands as functions of dinner timing, time of night, and their interaction with mixed-effect spline regression. Results Conventional sleep stages were similar between the 2 visits. LD caused a 2.5% initial increase in delta power and a reciprocal 2.7% decrease in combined alpha and beta power (p<0.0001). These effects diminished as sleep continued with a reversal of these patterns in the latter part of the night. Conclusion Contrary to the existing literature, shifting dinner timing from 5 hours before sleep to 1 hour before sleep in healthy volunteers did not result in significant adverse changes in overnight sleep architecture. In fact, LD was associated with deeper sleep in the beginning of the night and lighter sleep in the latter part of the night in healthy volunteers. This novel manifestation of postprandial hypersomnia may have therapeutic potential in patients with sleep disorders.

Abstract Visual attention refers to the cognitive process of selectively targeting certain visual stimuli in the environment, while others are ignored. In the case of food, it also has a strong influence on the taste, perception of the experience, and emotions related to food. The primary aim of this paper is to investigate how the visual presentation of food (burger) prepared under two conditions using identical raw materials, and a third condition utilizing a more nutritionally balanced composition (vegetarian), with varying visibility of vegetables (colored food elements), affects consumer preferences. Alongside traditional Tablet Assisted Personal Interviewing (TAPI), biometric techniques including Eye Tracking, Facereading, and Electroencephalography (EEG) were employed to capture unconscious emotional responses. We found statistically significant differences in emotional response due to the visualization of food. We also found that how food looks has a direct effect on how people think it smells and tastes. A significant discovery is that these disparities were validated not only at a declarative level but also at a subconscious level, offering a novel perspective on the perception of visual design and food arrangement, as well as enhancing targeting and application in the marketing management of foodservice establishments.

There are several methods available in measuring food taste. The sensory evaluation, for instance, is a typical method for panels to test of taste and recognize smell with their nose by measuring the degree of taste characteristic, intensity, and pleasure. There are many issues entailed in the traditional sensory evaluation method such as forming a panel and evaluation cost; moreover, it is only localized in particular areas. Accordingly, this paper aimed to select food in one particular area, and compare and review the content between sensory evaluations using a taste biological sensor, as well as presenting an analysis of brainwaves using EEG and finally a proposal of a new method for sensory evaluation. In this paper, the researchers have conducted a sensory evaluation whereas a maximum of nine points were accumulated by purchasing eight types of rice wine. These eight types of Makgeolli were generalized by generating multidimensional data with the use of TS-5000z, thus learning mapping points and scaling them. The contribution of this paper, therefore, is to overcome the disadvantages of the sensory evaluation with the usage of the suggested taste biological sensor system.

Specific attributes of products have been studied about consumer characteristics and how these factors directly impact the food production chain. Due to the barriers to penetration in a market segment, the knowledge of consumer preferences is important to enable adaptation of effective commercial strategies. Utilizing neuromarketing, electroencephalograms (EEGs), and principal component analysis (PCA), this study investigates the cognitive responses of 20 to 29-year-olds to samples of functional foods and correlates this information with purchase decisions. Brain responses were monitored using an EEG while participants tasted eight functional foods. Analysis demonstrated that important factors in deciding whether to buy or not included, after PCA processing, low beta and low gamma frequencies as well as attention and meditation percentages. The results showed that 0.96 of consumers were accurately identified as favoring a specific food and flavors, were accurately classified as bitter, sour, sweet, or salty with over 0.91 accuracy. These findings indicate that the key to effective strategies in the marketing of functional foods lies in aligning the sensory, emotional and brain attributes with consumer expectations.

Activation of adenosine A2a receptors in cerebral neurons induces sleep in various mammals. It was previously found that Japanese sake yeast enriched in adenosine analogues activates A2a receptors in vitro and induces sleep in mice. Here it is reported that sake yeast activated A2a receptors in a cultured human cell line and improved human sleep quality in a clinical trial. Sake yeast activated A2a receptors in HEK cells in a dose‐dependent manner with an EC50 of 40 μg mL−1, and the activation was attenuated almost completely by the A2a receptor antagonist ZM241385 with an IC50 of 73 nm. In a double‐blind placebo‐controlled crossover clinical study, 68 healthy participants ingested tablets containing either 500 mg of sake yeast powder or a placebo (cellulose) 1 h before sleep for 4 days. Electroencephalograms were recorded during sleep at home with a portable device for 4 week days. Electroencephalogram analyses revealed that sake yeast supplementation significantly (P = 0.03) increased delta power during the first cycle of slow‐wave sleep by 110%, without changing other sleep parameters. Sake yeast supplementation also significantly increased growth hormone secretion in the urine on awakening by 137% from 3.17 ± 0.41 (placebo) to 4.33 ± 0.62 (sake yeast) pg mg−1 creatinine (P = 0.03). Subjective sleepiness (P = 0.02) and fatigue (P = 0.06) in the morning were improved by sake yeast. Given these benefits and the absence of adverse effects during the study period, it was concluded that sake yeast supplementation is an effective and safe way to support daily high‐quality, deep sleep.

The categorization of food via sensing nutrients or toxins is crucial to the survival of any organism. On ingestion, rapid responses within the gustatory system are required to identify the oral stimulus to guide immediate behavior (swallowing or expulsion). The way in which the human brain accomplishes this task has so far remained unclear. Using multivariate analysis of 64-channel scalp EEG recordings obtained from 16 volunteers during tasting salty, sweet, sour, or bitter solutions, we found that activity in the delta-frequency range (1-4 Hz; delta power and phase) has information about taste identity in the human brain, with discriminable response patterns at the single-trial level within 130 ms of tasting. Importantly, the latencies of these response patterns predicted the point in time at which participants indicated detection of a taste by pressing a button. Furthermore, taste pattern discrimination was independent of motor-related activation and encoded taste identity rather than other taste features such as intensity and valence. On comparison with our previous findings from a delayed taste-discrimination task (Crouzet et al., 2015), taste-specific neural representations emerged earlier during this speeded taste-detection task, suggesting a goal-dependent flexibility in gustatory response coding. Together, these findings provide the first evidence of a role of delta activity in taste-information coding in humans. Crucially, these neuronal response patterns can be linked to the speed of simple gustatory perceptual decisions - a vital performance index of nutrient sensing.

Long-term effects of learned associations between an image and a taste have not been studied with electromagnetic brain scanning techniques. The possibility that taste conditioning may change sensory image processing was investigated in young adult subjects. EEG-responses evoked by images were recorded before and after a training session using an image as conditioned stimulus and a pleasant taste as unconditioned stimulus. The results showed that in posterior electrodes placed over visual cortex areas, the following changes occurred after conditioning: (1) the amplitude and duration of the N2-P3 waves in the visual evoked potentials were enhanced; (2) the N2 and P3 peak delays were shortened; (3) power induced by image presentation was enhanced in the delta and theta frequency bands; (4) cross-hemispheric delta and theta coherences among the posterior electrodes were enhanced; (5) calculations of the underlying whole brain distribution of currents using swLORETA showed elevated current densities in posterior voxels. None of the above changes occurred in a sham-trained control group. In electrodes placed over the prefrontal cortex, delta and theta power also rose significantly. It is suggested that the appetitive taste conditioning potentiated synaptic activity in visual cortex networks and that this led to an increased speed of image processing.

We investigated brain electric field signatures of subjective feelings after chewing regular gum or gum base without flavor. 19-channel eyes-closed EEG from 20 healthy males before and after 5 minutes of chewing the two gum types in random sequence was source modeled in the frequency domain using the FFT-Dipole-Approximation. 3-dimensional brain locations and strengths (Global Field Power, GFP) of the equivalent sources of five frequency bands were computed as changes from pre-chewing baseline. Gum types differed (ANOVA) in pre-post changes of source locations for the alpha-2 band (to anterior and right after regular gum, opposite after gum base) and beta-2 band (to anterior and inferior after regular gum, opposite after gum base), and of GFP for delta-theta, alpha-2 and beta-1 (regular gum: increase. gum base: decrease). Subjective feeling changed to more positive values after regular gum than gum base (ANOVA).--Thus, chewing gum with and without taste-smell activates different brain neuronal populations.

Understanding neural pathways and cognitive processes involved in the transformation of dietary fats into sensory experiences has profound implications for nutritional well-being. This study presents an efficient approach to comprehending the neural perception of fat taste using electroencephalogram (EEG). Through the examination of neural responses to different types of fatty acids (FAs) in 45 participants, we discerned distinct neural activation patterns associated with saturated versus unsaturated fatty acids. The spectrum analysis of averaged EEG signals revealed notable variations in δ and α-frequency bands across FA types. The topographical distribution and source localization results suggested that the brain encodes fat taste with specific activation timings in primary and secondary gustatory cortices. Saturated FAs elicited higher activation in cortical associated with emotion and reward processing. This electrophysiological evidence enhances our understanding of fundamental mechanisms behind fat perception, which is helpful for guiding strategies to manage hedonic eating and promote balanced fat consumption.

Sweetness is a key dimension of sensory experience in food, and variations in the type and concentration of sweeteners can elicit distinct brain responses. In this study, electroencephalography (EEG) was employed to systematically evaluate neural activity elicited by different concentrations of sucrose solutions (1%, 3%, 5%, and 7%) and by non-nutritive sweeteners matched in perceived sweetness to a 7% sucrose solution (10% erythritol, 0.0133% sucralose, and 0.0368% stevioside). The results revealed that an increased sucrose concentration was associated with progressively weaker EEG signal intensity, suggesting that the brain can effectively distinguish sweetness intensity. Under iso-sweet conditions, different types of sweeteners induced significantly distinct EEG patterns, indicating that the nature of the sweetener modulates flavor perception at the neural level. Further analysis showed increases in both δ- and α-band power following sweet taste stimulation, with prominent activations observed in the frontal, parietal, and right temporal regions. These findings demonstrate the utility of EEG in detecting subtle differences in brain responses to sweeteners, offering new insights into the neural mechanisms underlying sweet taste perception.

We evaluated the modifications of electroencephalographic (EEG) power spectra and EEG connectivity in overweight and obese patients with elevated food addiction (FA) symptoms. Fourteen overweight and obese patients (3 men and 11 women) with three or more FA symptoms and fourteen overweight and obese patients (3 men and 11 women) with two or less FA symptoms were included in the study. EEG was recorded during three different conditions: 1) five minutes resting state (RS), 2) five minutes resting state after a single taste of a chocolate milkshake (ML-RS), and 3) five minutes resting state after a single taste of control neutral solution (N-RS). EEG analyses were conducted by means of the exact Low Resolution Electric Tomography software (eLORETA). Significant modification was observed only in the ML-RS condition. Compared to controls, patients with three or more FA symptoms showed an increase of delta power in the right middle frontal gyrus (Brodmann Area [BA] 8) and in the right precentral gyrus (BA 9), and theta power in the right insula (BA 13) and in the right inferior frontal gyrus (BA 47). Furthermore, compared to controls, patients with three or more FA symptoms showed an increase of functional connectivity in fronto-parietal areas in both the theta and alpha band. The increase of functional connectivity was also positively associated with the number of FA symptoms. Taken together, our results show that FA has similar neurophysiological correlates of other forms of substance-related and addictive disorders suggesting similar psychopathological mechanisms.

Geographical Indication (GI) certifications enable producers to set production standards and create competitive advantage based on product's origin. In a coffee tasting experiment, brain responses to origin information of 40 participants, grouped equally by gender and involvement level, were collected by electroencephalography to verify: the impact of the GI cue in four brain waves (alpha, beta, delta and theta) and two brain lobes (frontal and temporal); preference; gender and involvement moderations. Results show that women presented power differences in both hemispheres, more channels/waves, which indicates greater sensitivity to the origin cue. Men presented power differences in fewer channels/waves. It is observed that involvement has a tenuous moderation effect when compared to gender. As for preference, the analysis of delta and theta waves indicated that men preferred coffee with GI; while women preferred coffee without GI, even though most of them indicated the opposite when verbally asked at the end of the tasting section.

This study investigates the flavor perception of strong-aroma Baijiu through physiological electrical signals, focusing on electroencephalography (EEG) and electromyography (EMG) during olfactory and gustatory evaluations. It examines how sensory qualities, especially mellowness, influence brain and muscle responses. Results showed significant differences in EEG δ and β wavebands, mainly in the frontal and temporal lobes, reflecting varying brain activities across Baijiu types. Mellower Baijiu triggered fewer activations in the digastric muscle, indicating a smoother swallowing experience. Baijiu with higher sensory scores reduced corrugator muscle activity and increased zygomatic major muscle activation, indicating a more pleasant taste. The findings highlight that Baijiu flavor perception is shaped by memory, emotion, and sensory inputs, demonstrating its complexity. This study offers valuable insights into flavor perception's multimodal nature and suggests novel ways to refine Baijiu evaluation using physiological signals to understand taste quality.

Alcoholic beverages flavour is complex and unique with different alcohol content, and the application of flavour perception could improve the objectivity of flavour evaluation. This study utilized electroencephalogram (EEG) to assess brain reactions to alcohol percentages (5 %-53 %) and Baijiu's complex flavours. The findings demonstrate the brain's proficiency in discerning between alcohol concentrations, evidenced by increasing physiological signal strength in tandem with alcohol content. When contrasted with alcohol solutions of equivalent concentrations, Baijiu prompts a more significant activation of brain signals, underscoring EEG's capability to detect subtleties due to flavour complexity. Additionally, the study reveals notable correlations, with δ and α wave intensities escalating in response to alcohol stimulation, coupled with substantial activation in the frontal, parietal, and right temporal regions. These insights verify the efficacy of EEG in charting the brain's engagement with alcoholic flavours, setting the stage for more detailed exploration into the neural encoding of these sensory experiences.

This study is the first to utilize electroencephalography (EEG) technology to characterize the full-bodied mouthfeel of Baijiu and to construct a parameter-controlled drinking experiment design. By employing spectral and time-frequency analysis methods, the study systematically investigated the impact of key experimental parameters, such as consumption volume (0.5-1.0 mL) and stimulus duration (0-60 s), on the neural representation of sensory evaluation of Baijiu. The results revealed that the power response in the delta and theta frequency bands (1-8 Hz) in the parietal and occipital regions is specifically correlated with the sensory grading of Baijiu, enabling the differentiation of Baijiu of different quality grades (p < 0.05). Additionally, alpha frequency band (8-13 Hz) activity showed a potential correlation with post-consumption emotional ratings. This methodological innovation lays a neuroscientific foundation for future flavor profile prediction combined with metabolomics and provides significant reference value for the standardized development of food sensory evaluation technologies.

The commercial utilization of jumbo squid (Dosidicus gigas) faces significant constraints due to its pronounced sour off-flavor. This investigation employed a multimodal approach combining electroencephalography (EEG), electronic sensing technology, and conventional sensory evaluation to characterize the perceptual attributes and neural correlates of varying off-flavor intensities, aiming to explore the underlying mechanisms. Sensory profiling indicated that sour off-flavor of jumbo squid was mainly attributed to the combined effects of sourness, bitterness, and persistent aftertaste. Further analysis using the electronic tongue and electronic nose clarified the distinct characteristics of the squid sour off-flavor at varying intensities. The spectrum analysis of averaged EEG signals revealed robust neural activation patterns, with particularly prominent responses in the δ and α frequency bands. Distinct levels of sour off-flavor elicited clearly different temporal dynamics, with perceptual distinctions emerging within the 0-200 and 950-1500 ms following stimulus onset. Notably, the topographical distribution and source localization results suggested that increasing off-flavor intensity elicited progressively stronger neural activation, with the prefrontal cortex and right temporal lobe showing the most pronounced responses. By employing EEG for comparative analysis of sour off-flavor intensities in jumbo squid, this study provides novel insights into the neurophysiological mechanisms underlying off-flavor perception. Therefore, EEG is proved as a promising objective assessment tool for quantifying subtle flavor variations often missed by conventional sensory panels in seafood quality evaluation.

This study used electroencephalography to characterize brain responses to the umami peptide RPIEK (0.75, 1.50, and 3.00 mg/mL) and matched monosodium glutamate solutions (0.10, 0.50, and 1.00 mg/mL), focusing on oscillatory dynamics, cortical regions, and spatiotemporal response profiles. The results demonstrated that RPIEK significantly increased (P < 0.05) area under the curve of the power spectral density, with delta and theta band activity principally encoding umami information. Brain can distinguish between umami stimuli matched for perceptual intensity based on differences in the temporal sequence of regional activation, and implicated the frontal lobe as a critical contributor to umami perception. Significant differences in rhythmic oscillatory activity, temporal activation windows, and spatial distribution across brain regions were observed between RPIEK and monosodium glutamate. Umami peptides elicited stronger activation in cortical regions associated with higher-order cognitive, emotional, and reward processing. This study is expected to elucidate the neurophysiological mechanisms of umami peptide perception.

Citrus flavors are globally popular in food industry, yet research on the perceptual preferences of various citrus flavors is limited. Based on the subjective sensory evaluation, this study introduces a novel sensory analysis approach, using electroencephalography (EEG), to objectively measure the sensory and cognitive responses to nine citrus flavors, including d-limonene, concentrated (H-) and original essential oils of sweet orange (SEO), bergamot EO (BEO), lemon EO (LEO), and grapefruit EO (GEO). Results revealed that δ (0.5-4 Hz) and α (8-13 Hz) waves activity predominated in brain responses to citrus flavor, with greater activity observed in frontal and central regions compared to other areas. Sniffing citrus EOs triggered more complex and dynamic electrical activity than d-limonene, indicated by higher power density across all frequency bands (0.1-30 Hz). Interestingly, while the original citrus EOs were associated with higher self-reported acceptability, the concentrated forms elicited greater brain responses. Specifically, H-SEO and L-LEO eliciting significantly greater δ and α wave activity in the prefrontal region than their original forms (P < 0.05). A preliminary correlation was observed between brain laterality in α waves power and acceptability scores of citrus flavor, with δ waves power in the prefrontal region further demonstrating an effective reflection of self-reported acceptability scores for SEO and LEO stimuli. This is the first EEG-based study to compare brain responses to different citrus flavors, providing important implications for the food industry in optimizing product formulations and enhancing consumer experiences.

Previous attempts to uncover a relation between taste processing and weight status have yielded inconclusive results leaving it unclear whether lean and obese individuals process taste differently, and whether group differences reflect differential sensory encoding or evaluative and reward processing. Here, we present the first comparison of dynamic neural processing as assessed by gustatory evoked potentials in obese and lean individuals. Two supra-threshold concentrations of sweet and salty tastants as well as two sizes of blue and green squares were presented to 30 lean (BMI 18.5-25) and 25 obese (BMI > 30) individuals while recording head-surface electroencephalogram (EEG). Multivariate pattern analyses (MVPA) revealed differential taste quality representations from 130 ms until after stimulus offset. Notably, taste representations faded earlier and exhibited a reduced strength in the obese compared to the lean group; temporal generalization analysis indicated otherwise similar taste processing. Differences in later gustatory response patterns even allowed decoding of group membership. Importantly, group differences were absent for visual processing thereby excluding confounding effects from anatomy or signal-to-noise ratio alone. The latency of observed effects is consistent with memory maintenance rather than sensory encoding of taste, thereby suggesting that later evaluative aspects of taste processing are altered in obesity.

The quick and reliable detection and identification of a tastant in the mouth regulate nutrient uptake and toxin expulsion. Consistent with the pivotal role of the gustatory system, taste category information (e.g., sweet, salty) is represented during the earliest phase of the taste-evoked cortical response (Crouzet et al., 2015), and different tastes are perceived and responded to within only a few hundred milliseconds, in rodents (Perez et al., 2013) and humans (Bujas, 1935). Currently, it is unknown whether taste detection and discrimination are sequential or parallel processes, i.e., whether you know what it is as soon as you taste it. To investigate the sequence of processing steps involved in taste perceptual decisions, participants tasted sour, salty, bitter, and sweet solutions and performed a taste-detection and a taste-discrimination task. We measured response times (RTs) and 64-channel scalp electrophysiological recordings and tested the link between the timing of behavioral decisions and the timing of neural taste representations determined with multivariate pattern analyses. Irrespective of taste and task, neural decoding onset and behavioral RTs were strongly related, demonstrating that differences between taste judgments are reflected early during chemosensory encoding. Neural and behavioral detection times were faster for the iso-hedonic salty and sour tastes than their discrimination time. No such latency difference was observed for sweet and bitter, which differ hedonically. Together, these results indicate that the human gustatory system detects a taste faster than it discriminates between tastes, yet hedonic computations may run in parallel (Perez et al., 2013) and facilitate taste identification.

For humans, taste is essential for perceiving food's nutrient content or harmful components. The current sensory evaluation of taste mainly relies on artificial sensory evaluation and electronic tongue, but the former has strong subjectivity and poor repeatability, and the latter is not flexible enough. This work proposed a strategy for acquiring and recognizing taste electroencephalogram (EEG), aiming to decode people's objective perception of taste through taste EEG. Firstly, according to the proposed experimental paradigm, the taste EEG of subjects under different taste stimulation was collected. Secondly, to avoid insufficient training of the model due to the small number of taste EEG samples, a Temporal and Spatial Reconstruction Data Augmentation (TSRDA) method was proposed, which effectively augmented the taste EEG by reconstructing the taste EEG's important features in temporal and spatial dimensions. Thirdly, a multi-view channel attention module was introduced into a designed convolutional neural network to extract the important features of the augmented taste EEG. The proposed method has accuracy of 99.56%, F1-score of 99.48%, and kappa of 99.38%, proving the method's ability to distinguish the taste EEG evoked by different taste stimuli successfully. In summary, combining TSRDA with taste EEG technology provides an objective and effective method for sensory evaluation of food taste.

The taste electroencephalogram (EEG) evoked by the taste stimulation can reflect different brain patterns and be used in applications such as sensory evaluation of food. However, considering the computational cost and efficiency, EEG data with many channels has to face the critical issue of channel selection. This paper proposed a channel selection method called class activation mapping with attention (CAM-Attention). The CAM-Attention method combined a convolutional neural network with channel and spatial attention (CNN-CSA) model with a gradient-weighted class activation mapping (Grad-CAM) model. The CNN-CSA model exploited key features in EEG data by attention mechanism, and the Grad-CAM model effectively realized the visualization of feature regions. Then, channel selection was effectively implemented based on feature regions. Finally, the CAM-Attention method reduced the computational burden of taste EEG recognition and effectively distinguished the four tastes. In short, it has excellent recognition performance and provides effective technical support for taste sensory evaluation.

The classification of harmful brain activities, such as seizures and periodic discharges, play a vital role in neurocritical care, enabling timely diagnosis and intervention. Electroencephalography (EEG) provides a non-invasive method for monitoring brain activity, but the manual interpretation of EEG signals are time-consuming and rely heavily on expert judgment. This study presents a comparative analysis of deep learning architectures, including Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and EEGNet, applied to the classification of harmful brain activities using both raw EEG data and time-frequency representations generated through Continuous Wavelet Transform (CWT). We evaluate the performance of these models use multimodal data representations, including high-resolution spectrograms and waveform data, and introduce a multi-stage training strategy to improve model robustness. Our results show that training strategies, data preprocessing, and augmentation techniques are as critical to model success as architecture choice, with multi-stage TinyViT and EfficientNet demonstrating superior performance. The findings underscore the importance of robust training regimes in achieving accurate and efficient EEG classification, providing valuable insights for deploying AI models in clinical practice.

The recent availability of large datasets in bio-medicine has inspired the development of representation learning methods for multiple healthcare applications. Despite advances in predictive performance, the clinical utility of such methods is limited when exposed to real-world data. This study develops model diagnostic measures to detect potential pitfalls before deployment without assuming access to external data. Specifically, we focus on modeling realistic data shifts in electrophysiological signals (EEGs) via data transforms and extend the conventional task-based evaluations with analyses of a) the model's latent space and b) predictive uncertainty under these transforms. We conduct experiments on multiple EEG feature encoders and two clinically relevant downstream tasks using publicly available large-scale clinical EEGs. Within this experimental setting, our results suggest that measures of latent space integrity and model uncertainty under the proposed data shifts may help anticipate performance degradation during deployment.

Detecting single-trial P300 from EEG is difficult when only a few labeled trials are available. When attempting to boost a small target set with a large source dataset through transfer learning, cross-dataset shift arises. To address this challenge, we study transfer between two public visual-oddball ERP datasets using five shared electrodes (Fz, Pz, P3, P4, Oz) under a strict small-sample regime (target: 10 trials/subject; source: 80 trials/subject). We introduce Adaptive Split Maximum Mean Discrepancy Training (AS-MMD), which combines (i) a target-weighted loss with warm-up tied to the square root of the source/target size ratio, (ii) Split Batch Normalization (Split-BN) with shared affine parameters and per-domain running statistics, and (iii) a parameter-free logit-level Radial Basis Function kernel Maximum Mean Discrepancy (RBF-MMD) term using the median-bandwidth heuristic. Implemented on an EEG Conformer, AS-MMD is backbone-agnostic and leaves the inference-time model unchanged. Across both transfer directions, it outperforms target-only and pooled training (Active Visual Oddball: accuracy/AUC 0.66/0.74; ERP CORE P3: 0.61/0.65), with gains over pooling significant under corrected paired t-tests. Ablations attribute improvements to all three components.

A major shortcoming of medical practice is the lack of an objective measure of conscious level. Impairment of consciousness is common, e.g. following brain injury and seizures, which can also interfere with sensory processing and volitional responses. This is also an important pitfall in neurophysiological methods that infer awareness via command following, e.g. using functional MRI or electroencephalography (EEG). Transcranial electrical stimulation (TES) can be employed to non-invasively stimulate the brain, bypassing sensory inputs, and has already showed promising results in providing reliable indicators of brain state. However, current non-invasive solutions have been limited to magnetic stimulation, which is not easily translatable to clinical settings. Our long-term vision is to develop an objective measure of brain state that can be used at the bedside, without requiring patients to understand commands or initiate motor responses. In this study, we demonstrated the feasibility of a framework using Deep Learning algorithms to classify EEG brain responses evoked by a defined multi-dimensional pattern of TES. We collected EEG-TES data from 11 participants and found that delivering transcranial direct current stimulation (tDCS) to posterior cortical areas targeting the angular gyrus elicited an exceptionally reliable brain response. For this paradigm, our best Convolutional Neural Network model reached a 92% classification F1-score on Holdout data from participants never seen during training, significantly surpassing human-level performance at 60-70% accuracy. These findings establish a framework for robust consciousness measurement for clinical use. In this spirit, we documented and open-sourced our datasets and codebase in full, to be used freely by the neuroscience and AI research communities, who may replicate our results with free tools like GitHub, Kaggle, and Colab.

High temporal resolution measurements of human brain activity can be performed by recording the electric potentials on the scalp surface (electroencephalography, EEG), or by recording the magnetic fields near the surface of the head (magnetoencephalography, MEG). The analysis of the data is problematic due to the fact that multiple neural generators may be simultaneously active and the potentials and magnetic fields from these sources are superimposed on the detectors. It is highly desirable to un-mix the data into signals representing the behaviors of the original individual generators. This general problem is called blind source separation and several recent techniques utilizing maximum entropy, minimum mutual information, and maximum likelihood estimation have been applied. These techniques have had much success in separating signals such as natural sounds or speech, but appear to be ineffective when applied to EEG or MEG signals. Many of these techniques implicitly assume that the source distributions have a large kurtosis, whereas an analysis of EEG/MEG signals reveals that the distributions are multimodal. This suggests that more effective separation techniques could be designed for EEG and MEG signals.

Researchers have reported high decoding accuracy (>95%) using non-invasive Electroencephalogram (EEG) signals for brain-computer interface (BCI) decoding tasks like image decoding, emotion recognition, auditory spatial attention detection, etc. Since these EEG data were usually collected with well-designed paradigms in labs, the reliability and robustness of the corresponding decoding methods were doubted by some researchers, and they argued that such decoding accuracy was overestimated due to the inherent temporal autocorrelation of EEG signals. However, the coupling between the stimulus-driven neural responses and the EEG temporal autocorrelations makes it difficult to confirm whether this overestimation exists in truth. Furthermore, the underlying pitfalls behind overestimated decoding accuracy have not been fully explained due to a lack of appropriate formulation. In this work, we formulate the pitfall in various EEG decoding tasks in a unified framework. EEG data were recorded from watermelons to remove stimulus-driven neural responses. Labels were assigned to continuous EEG according to the experimental design for EEG recording of several typical datasets, and then the decoding methods were conducted. The results showed the label can be successfully decoded as long as continuous EEG data with the same label were split into training and test sets. Further analysis indicated that high accuracy of various BCI decoding tasks could be achieved by associating labels with EEG intrinsic temporal autocorrelation features. These results underscore the importance of choosing the right experimental designs and data splits in BCI decoding tasks to prevent inflated accuracies due to EEG temporal autocorrelation.

EEG decoding systems based on deep neural networks have been widely used in decision making of brain computer interfaces (BCI). Their predictions, however, can be unreliable given the significant variance and noise in EEG signals. Previous works on EEG analysis mainly focus on the exploration of noise pattern in the source signal, while the uncertainty during the decoding process is largely unexplored. Automatically detecting and reducing such decoding uncertainty is important for BCI motor imagery applications such as robotic arm control etc. In this work, we proposed an uncertainty estimation and reduction model (UNCER) to quantify and mitigate the uncertainty during the EEG decoding process. It utilized a combination of dropout oriented method and Bayesian neural network for uncertainty estimation to incorporate both the uncertainty in the input signal and the uncertainty in the model parameters. We further proposed a data augmentation based approach for uncertainty reduction. The model can be integrated into current widely used EEG neural decoders without change of architecture. We performed extensive experiments for uncertainty estimation and its reduction in both intra-subject EEG decoding and cross-subject EEG decoding on two public motor imagery datasets, where the proposed model achieves significant improvement both on the quality of estimated uncertainty and the effectiveness of uncertainty reduction.

Recognition of electroencephalographic (EEG) signals highly affect the efficiency of non-invasive brain-computer interfaces (BCIs). While recent advances of deep-learning (DL)-based EEG decoders offer improved performances, the development of geometric learning (GL) has attracted much attention for offering exceptional robustness in decoding noisy EEG data. However, there is a lack of studies on the merged use of deep neural networks (DNNs) and geometric learning for EEG decoding. We herein propose a manifold attention network (mAtt), a novel geometric deep learning (GDL)-based model, featuring a manifold attention mechanism that characterizes spatiotemporal representations of EEG data fully on a Riemannian symmetric positive definite (SPD) manifold. The evaluation of the proposed MAtt on both time-synchronous and -asyncronous EEG datasets suggests its superiority over other leading DL methods for general EEG decoding. Furthermore, analysis of model interpretation reveals the capability of MAtt in capturing informative EEG features and handling the non-stationarity of brain dynamics.

Transformers are groundbreaking architectures that have changed a flow of deep learning, and many high-performance models are developing based on transformer architectures. Transformers implemented only with attention with encoder-decoder structure following seq2seq without using RNN, but had better performance than RNN. Herein, we investigate the decoding technique for electroencephalography (EEG) composed of self-attention module from transformer architecture during imagined speech and overt speech. We performed classification of nine subjects using convolutional neural network based on EEGNet that captures temporal-spectral-spatial features from EEG of imagined speech and overt speech. Furthermore, we applied the self-attention module to decoding EEG to improve the performance and lower the number of parameters. Our results demonstrate the possibility of decoding brain activities of imagined speech and overt speech using attention modules. Also, only single channel EEG or ear-EEG can be used to decode the imagined speech for practical BCIs.

Electroencephalography (EEG) is a neuroimaging technique that records brain neural activity with high temporal resolution. Unlike other methods, EEG does not require prohibitively expensive equipment and can be easily set up using commercially available portable EEG caps, making it an ideal candidate for brain-computer interfaces. However, EEG signals are characterised by poor spatial resolution and high noise levels, complicating their decoding. In this study, we employ a contrastive learning framework to align encoded EEG features with pretrained CLIP features, achieving a 7% improvement over the state-of-the-art in EEG decoding of object categories. This enhancement is equally attributed to (1) a novel online sampling method that boosts the signal-to-noise ratio and (2) multimodal representations leveraging visual and language features to enhance the alignment space. Our analysis reveals a systematic interaction between the architecture and dataset of pretrained features and their alignment efficacy for EEG signal decoding. This interaction correlates with the generalisation power of the pretrained features on ImageNet-O/A datasets ($r=.5$). These findings extend beyond EEG signal alignment, offering potential for broader applications in neuroimaging decoding and generic feature alignments.

Decoding EEG signals is crucial for unraveling human brain and advancing brain-computer interfaces. Traditional machine learning algorithms have been hindered by the high noise levels and inherent inter-person variations in EEG signals. Recent advances in deep neural networks (DNNs) have shown promise, owing to their advanced nonlinear modeling capabilities. However, DNN still faces challenge in decoding EEG samples of unseen individuals. To address this, this paper introduces a novel approach by incorporating the conditional identification information of each individual into the neural network, thereby enhancing model representation through the synergistic interaction of EEG and personal traits. We test our model on the WithMe dataset and demonstrated that the inclusion of these identifiers substantially boosts accuracy for both subjects in the training set and unseen subjects. This enhancement suggests promising potential for improving for EEG interpretability and understanding of relevant identification features.

The human brain is constantly processing and integrating information in order to make decisions and interact with the world, for tasks from recognizing a familiar face to playing a game of tennis. These complex cognitive processes require communication between large populations of neurons. The non-invasive neuroimaging methods of electroencephalography (EEG) and magnetoencephalography (MEG) provide population measures of neural activity with millisecond precision that allow us to study the temporal dynamics of cognitive processes. However, multi-sensor M/EEG data is inherently high dimensional, making it difficult to parse important signal from noise. Multivariate pattern analysis (MVPA) or "decoding" methods offer vast potential for understanding high-dimensional M/EEG neural data. MVPA can be used to distinguish between different conditions and map the time courses of various neural processes, from basic sensory processing to high-level cognitive processes. In this chapter, we discuss the practical aspects of performing decoding analyses on M/EEG data as well as the limitations of the method, and then we discuss some applications for understanding representational dynamics in the human brain.

Accurate electroencephalogram (EEG) pattern decoding for specific mental tasks is one of the key steps for the development of brain-computer interface (BCI), which is quite challenging due to the considerably low signal-to-noise ratio of EEG collected at the brain scalp. Machine learning provides a promising technique to optimize EEG patterns toward better decoding accuracy. However, existing algorithms do not effectively explore the underlying data structure capturing the true EEG sample distribution, and hence can only yield a suboptimal decoding accuracy. To uncover the intrinsic distribution structure of EEG data, we propose a clustering-based multi-task feature learning algorithm for improved EEG pattern decoding. Specifically, we perform affinity propagation-based clustering to explore the subclasses (i.e., clusters) in each of the original classes, and then assign each subclass a unique label based on a one-versus-all encoding strategy. With the encoded label matrix, we devise a novel multi-task learning algorithm by exploiting the subclass relationship to jointly optimize the EEG pattern features from the uncovered subclasses. We then train a linear support vector machine with the optimized features for EEG pattern decoding. Extensive experimental studies are conducted on three EEG datasets to validate the effectiveness of our algorithm in comparison with other state-of-the-art approaches. The improved experimental results demonstrate the outstanding superiority of our algorithm, suggesting its prominent performance for EEG pattern decoding in BCI applications.

State-of-the-art performance in electroencephalography (EEG) decoding tasks is currently often achieved with either Deep-Learning (DL) or Riemannian-Geometry-based decoders (RBDs). Recently, there is growing interest in Deep Riemannian Networks (DRNs) possibly combining the advantages of both previous classes of methods. However, there are still a range of topics where additional insight is needed to pave the way for a more widespread application of DRNs in EEG. These include architecture design questions such as network size and end-to-end ability. How these factors affect model performance has not been explored. Additionally, it is not clear how the data within these networks is transformed, and whether this would correlate with traditional EEG decoding. Our study aims to lay the groundwork in the area of these topics through the analysis of DRNs for EEG with a wide range of hyperparameters. Networks were tested on five public EEG datasets and compared with state-of-the-art ConvNets. Here we propose EE(G)-SPDNet, and we show that this wide, end-to-end DRN can outperform the ConvNets, and in doing so use physiologically plausible frequency regions. We also show that the end-to-end approach learns more complex filters than traditional band-pass filters targeting the classical alpha, beta, and gamma frequency bands of the EEG, and that performance can benefit from channel specific filtering approaches. Additionally, architectural analysis revealed areas for further improvement due to the possible under utilisation of Riemannian specific information throughout the network. Our study thus shows how to design and train DRNs to infer task-related information from the raw EEG without the need of handcrafted filterbanks and highlights the potential of end-to-end DRNs such as EE(G)-SPDNet for high-performance EEG decoding.

The significant inter-subject variability in electroen-cephalogram (EEG) signals often results in substantial changes to neural network weights as data distributions shift. This variability frequently causes catastrophic forgetting in continual EEG decoding tasks, where previously acquired knowledge is overwritten as new subjects are introduced. While retraining the entire dataset for each new subject can mitigate forgetting, this approach imposes significant computational costs, rendering it impractical for real-world applications. Therefore, an ideal brain-computer interface (BCI) model should incrementally learn new information without requiring complete retraining, thereby reducing computational overhead. Existing EEG decoding meth-ods typically rely on large, centralized source-domain datasets for pre-training to improve model generalization. However, in practical scenarios, data availability is often constrained by privacy concerns. Furthermore, these methods are susceptible to catastrophic forgetting in continual EEG decoding tasks, significantly limiting their utility in long-term learning scenarios. To address these issues, we propose the Personalized Continual EEG Decoding (PCED) framework for continual EEG decoding. The framework uses Euclidean Alignment for fast domain adap-tation, reducing inter-subject variability. To retain knowledge and prevent forgetting, it includes an exemplar replay mechanism that preserves key information from past tasks. A reservoir sampling-based memory management strategy optimizes exemplar storage to handle memory constraints in long-term learning. Experiments on the OpenBMI dataset with 54 subjects show that PCED balances knowledge retention and classification performance, providing an efficient solution for real-world BCI applications.

Electroencephalography (EEG)-based brain-computer interfaces (BCIs) transform spontaneous/evoked neural activity into control commands for external communication. While convolutional neural networks (CNNs) remain the mainstream backbone for EEG decoding, their inherently short receptive field makes it difficult to capture long-range temporal dependencies and global inter-channel relationships. Recent CNN-Transformer (Conformer) hybrids partially address this issue, but most adopt a serial design, resulting in suboptimal integration of local and global features, and often overlook explicit channel-wise modeling. To address these limitations, we propose DBConformer, a dual-branch convolutional Transformer network tailored for EEG decoding. It integrates a temporal Conformer to model long-range temporal dependencies and a spatial Conformer to extract inter-channel interactions, capturing both temporal dynamics and spatial patterns in EEG signals. A lightweight channel attention module further refines spatial representations by assigning data-driven importance to EEG channels. Extensive experiments under four evaluation settings on three paradigms, including motor imagery, seizure detection, and steady-state visual evoked potential, demonstrated that DBConformer consistently outperformed 13 competitive baseline models, with over an eight-fold reduction in parameters than current high-capacity EEG Conformer architecture. Furthermore, the visualization results confirmed that the features extracted by DBConformer are physiologically interpretable and aligned with prior knowledge. The superior performance and interpretability of DBConformer make it reliable for accurate, robust, and explainable EEG decoding. Code is publicized at https://github.com/wzwvv/DBConformer.

Decoding speech from non-invasive brain signals, such as electroencephalography (EEG), has the potential to advance brain-computer interfaces (BCIs), with applications in silent communication and assistive technologies for individuals with speech impairments. However, EEG-based speech decoding faces major challenges, such as noisy data, limited datasets, and poor performance on complex tasks like speech perception. This study attempts to address these challenges by employing variational autoencoders (VAEs) for EEG data augmentation to improve data quality and applying a state-of-the-art (SOTA) sequence-to-sequence deep learning architecture, originally successful in electromyography (EMG) tasks, to EEG-based speech decoding. Additionally, we adapt this architecture for word classification tasks. Using the Brennan dataset, which contains EEG recordings of subjects listening to narrated speech, we preprocess the data and evaluate both classification and sequence-to-sequence models for EEG-to-words/sentences tasks. Our experiments show that VAEs have the potential to reconstruct artificial EEG data for augmentation. Meanwhile, our sequence-to-sequence model achieves more promising performance in generating sentences compared to our classification model, though both remain challenging tasks. These findings lay the groundwork for future research on EEG speech perception decoding, with possible extensions to speech production tasks such as silent or imagined speech.

Feature extraction for automatic classification of EEG signals typically relies on time frequency representations of the signal. Techniques such as cepstral-based filter banks or wavelets are popular analysis techniques in many signal processing applications including EEG classification. In this paper, we present a comparison of a variety of approaches to estimating and postprocessing features. To further aid in discrimination of periodic signals from aperiodic signals, we add a differential energy term. We evaluate our approaches on the TUH EEG Corpus, which is the largest publicly available EEG corpus and an exceedingly challenging task due to the clinical nature of the data. We demonstrate that a variant of a standard filter bank-based approach, coupled with first and second derivatives, provides a substantial reduction in the overall error rate. The combination of differential energy and derivatives produces a 24% absolute reduction in the error rate and improves our ability to discriminate between signal events and background noise. This relatively simple approach proves to be comparable to other popular feature extraction approaches such as wavelets, but is much more computationally efficient.