Traditional EEG Research Articles

Modeling the two-dimensional variations in EEG signals to analyze the impact of music on sleep states

Variations in music intensity and pitch can significantly impact sleep and alleviate insomnia through relaxing music. Traditional EEG analysis methods struggle with capturing temporal features and managing time-series data complexity. To address these challenges, we propose an improved Temporal Network model integrated with a Temporal Self-Attention (TSA) mechanism. This model enhances time-step feature capture in EEG signals, transforming one-dimensional EEG data into two-dimensional tensors to capture multi-scale temporal features. Experiments on the DEAP dataset demonstrate significant performance improvements, with 84.26% accuracy, 90.06% precision, and 97.82% recall, outperforming current models. Our model effectively analyzes the impact of relaxing music on different sleep stages, providing insights for improving sleep quality. Future work will optimize the model structure, incorporate additional deep learning techniques, and expand the scope to include multiple physiological signals, aiming to enhance the clinical value for sleep disorder diagnosis and personalized music therapy.

Multiband EEG signature decoded using machine learning for predicting rTMS treatment response in major depression.

Major depressive disorder (MDD) is a global health challenge with high prevalence. Further, many diagnosed with MDD are treatment resistant to traditional antidepressants. Repetitive transcranial magnetic stimulation (rTMS) offers promise as an alternative solution, but identifying objective biomarkers for predicting treatment response remains underexplored. Electroencephalographic (EEG) recordings are a cost-effective neuroimaging approach, but traditional EEG analysis methods often do not consider patient-specific variations and fail to capture complex neuronal dynamics. To address this, we propose a data-driven approach combining iterated masking empirical mode decomposition (itEMD) and sparse Bayesian learning (SBL). Our results demonstrated significant prediction of rTMS outcomes using this approach (Protocol 1: r=0.40, p<0.01; Protocol 2: r=0.26, p<0.05). From the decomposition, we obtained three key oscillations: IMF-Alpha, IMF-Beta, and the remaining residue. We also identified key spatial patterns associated with treatment outcomes for two rTMS protocols: for Protocol 1 (10Hz left DLPFC), important areas include the left frontal and parietal regions, while for Protocol 2 (1Hz right DLPFC), the left and frontal, left parietal regions are crucial. Additionally, our exploratory analysis found few significant correlations between oscillation specific predictive features and personality measures. This study highlights the potential of machine learning-driven EEG analysis for personalized MDD treatment prediction, offering a pathway for improved patient outcomes.

Stability of infraslow correlation structure in time-shifted intracranial EEG signals.

It is increasingly understood that the epilepsies are characterized by network pathology that can span multiple spatial and temporal scales. Recent work indicates that infraslow (<0.2Hz) envelope correlations may form a basis for distant spatial coupling in the brain. We speculated that infraslow correlation structure may be preserved even with some time lag between signals. To this end, we studied intracranial EEG (icEEG) data collected from 22 medically refractory epilepsy patients. For each patient, we selected hour-long background, awake icEEG epochs before and after antiseizure medication (ASM) taper. For each epoch, we selected 5,000 random electrode contact pairs and estimated magnitude-squared coherence (MSC) below 0.15Hz of band power time-series in the traditional EEG frequency bands. Using these same contact pairs, we shifted one signal of the pair by random durations in 15-s increments between 0 and 300s. We aggregated these data across all patients to determine how infraslow MSC varies with duration of lag. We further examined the effect of ASM taper on infraslow correlation structure. We also used surrogate data to empirically characterize MSC estimator and to set optimal parameters for estimation specifically for the study of infraslow activity. Our empirical analysis of the MSC estimator showed that hour-long segments with MSC computed using 3-min windows with 50% overlap was sufficient to capture infraslow envelope correlations while minimizing estimator bias and variance. The mean MSC decreased monotonically with increasing time lag until 105s of lag, then plateaued between 106 and 300s. Significantly nonzero infraslow envelope MSC was preserved in all frequency bands until about 1min of time lag, both pre- and post-ASM taper. We also saw a slight, but significant increase in infraslow MSC post-ASM taper, consistent with prior work. These results provide evidence for the feasibility of examining infraslow activity via its modulation of higher-frequency activity in the absence of DC-coupled recordings. The use of surrogate data also provides a general methodology for benchmarking measures used in network neuroscience studies. Finally, our study points to the clinical relevance of infraslow activity in assessing seizure risk.

Adaptive channel-weight dual-constrained semi-supervised EEG clustering

When processing Electroencephalography(EEG) for applications such as brain disease diagnosis, rehabilitation, and brain–computer interfaces, we encounter challenges related to missing labels or incorrect annotations. These challenges may arise from subjective factors, technical limitations, or data quality issues. In traditional EEG processing, unsupervised clustering algorithms are commonly used; however, due to the lack of labels, these methods may struggle to fully leverage latent information, resulting in uncertainty and instability in clustering outcomes. To address this issue, we propose a semi-supervised clustering method tailored for partially labeled EEG signals, named Adaptive Channel-weight Dual-constrained Semi-supervised EEG Clustering (ACDSEEGc). This method incorporates dual constraints: a connectivity constraint for the interconnections between EEG points and a neighborhood constraint for the spatial neighborhood relationships among EEG data points. Additionally, it integrates channel weights to emphasize information from important channels. The objective function of ACDSEEGc integrates cross-entropy evaluation, channel weight learning, pseudo-label learning, connectivity constraint, and least squares error minimization. Our aim is to maximize the use of limited prior knowledge to achieve better EEG clustering results. Experimental validation demonstrates that ACDSEEGc effectively produces superior clustering outcomes compared to state-of-the-art standard unsupervised and semi-supervised EEG/time-series clustering algorithms. Furthermore, we conducted sensitivity experiments and ablation studies to verify the performance of ACDSEEGc under different settings.

Multi-Scale Feature and Multi-Channel Selection toward Parkinson's Disease Diagnosis with EEG.

Motivated by Health Care 4.0, this study aims to reducing the dimensionality of traditional EEG features based on manual extracted features, including statistical features in the time and frequency domains. A total of 22 multi-scale features were extracted from the UNM and Iowa datasets using a 4th order Butterworth filter and wavelet packet transform. Based on single-channel validation, 29 channels with the highest R2 scores were selected from a pool of 59 common channels. The proposed channel selection scheme was validated on the UNM dataset and tested on the Iowa dataset to compare its generalizability against models trained without channel selection. The experimental results demonstrate that the proposed model achieves an optimal classification accuracy of 100%. Additionally, the generalization capability of the channel selection method is validated through out-of-sample testing based on the Iowa dataset Conclusions: Using single-channel validation, we proposed a channel selection scheme based on traditional statistical features, resulting in a selection of 29 channels. This scheme significantly reduced the dimensionality of EEG feature vectors related to Parkinson's disease by 50%. Remarkably, this approach demonstrated considerable classification performance on both the UNM and Iowa datasets. For the closed-eye state, the highest classification accuracy achieved was 100%, while for the open-eye state, the highest accuracy reached 93.75%.

Neural correlates for word-frequency effect in Chinese natural reading.

Word frequency effect has always been of interest for reading research because of its critical role in exploring mental processing underlying reading behaviors. Access to word frequency information has long been considered an indicator of the beginning of lexical processing and the most sensitive marker for studying when the brain begins to extract semantic information Sereno & Rayner, Brain and Cognition, 42, 78-81, (2000), Trends in Cognitive Sciences, 7, 489-493, (2003). While the word frequency effect has been extensively studied in numerous eye-tracking and traditional EEG research using the RSVP paradigm, there is a lack of corresponding evidence in studies of natural reading. To find the neural correlates of the word frequency effect, we conducted a study of Chinese natural reading using EEG and eye-tracking coregistration to examine the time course of lexical processing. Our results reliably showed that the word frequency effect first appeared in the N200 time window and the bilateral occipitotemporal regions. Additionally, the word frequency effect was reflected in the N400 time window, spreading from the occipital region to the central parietal and frontal regions. Our current study provides the first neural correlates for word-frequency effect in natural Chinese reading so far, shedding new light on understanding lexical processing in natural reading and could serve as an important basis for further reading study when considering neural correlates in a realistic manner.

Epilepsy lesion localization method based on brain function network.

In the past, the localization of seizure onset zone (SOZ) primarily relied on traditional EEG signal analysis methods. However, due to their limited spatial and temporal resolution, accurately pinpointing neural activity was challenging, thereby restricting their clinical applicability. Compared with traditional EEG signals, SEEG signals have superior spatial and temporal resolution, and can more accurately record neural activity near epileptic foci, making them better suited for studying SOZ. In addition, the traditional EEG signal analysis methods still have limitations, mainly focusing on the analysis of local signal features, while ignoring the complexity and interconnection of the overall brain network. How to more accurately locate SOZ is still not well resolved. The purpose of this study is to develop an effective positioning method for more accurate positioning. To overcome these limitations, this study proposed a model integrating brain functional network analysis with nonlinear dynamics. We utilized weighted phase lag index (WPLI) to construct brain functional network, epilepic network connectivity strength (ENCS) as the feature, and introduced persistence entropy (PE) for feature fusion, subsequently employing support vector machine (SVM) classification. The proposed method was verified on the HUP-iEEG dataset, our solution identified the SOZ with 0.9440 accuracy, 0.9848 precision, 0.8974 recall rate, 0.9340 F1 score and 0.9697 area under the ROC curve across patients, which outperforms the existing approaches. It exhibits a 2.30 percentage point enhancement in localisation accuracy along with a 2.97 percentage points in AUC compared to others. Our method consider the interactions between nodes in brain network connections, as well as the inherent nonlinear and non-stationary properties of neural signals, to be more robust.

Global Field Time-Frequency Representation-Based Discriminative Similarity Analysis of Passive Auditory ERPs for Diagnosis of Disorders of Consciousness.

Behavioural diagnosis of patients with disorders of consciousness (DOC) is challenging and prone to inaccuracies. Consequently, there have been increased efforts to develop bedside assessment based on EEG and event-related potentials (ERPs) that are more sensitive to the neural factors supporting conscious awareness. However, individual detection of residual consciousness using these techniques is less established. Here, we hypothesize that the cross-state similarity (defined as the similarity between healthy and impaired conscious states) of passive brain responses to auditory stimuli can index the level of awareness in individual DOC patients. To this end, we introduce the global field time-frequency representation-based discriminative similarity analysis (GFTFR-DSA). This method quantifies the average cross-state similarity index between an individual patient and our constructed healthy templates using the GFTFR as an EEG feature. We demonstrate that the proposed GFTFR feature exhibits superior within-group consistency in 34 healthy controls over traditional EEG features such as temporal waveforms. Second, we observed the GFTFR-based similarity index was significantly higher in patients with a minimally conscious state (MCS, 40 patients) than those with unresponsive wakefulness syndrome (UWS, 54 patients), supporting our hypothesis. Finally, applying a linear support vector machine classifier for individual MCS/UWS classification, the model achieved a balanced accuracy and F1 score of 0.77. Overall, our findings indicate that combining discriminative and interpretable markers, along with automatic machine learning algorithms, is effective for the differential diagnosis in patients with DOC. Importantly, this approach can, in principle, be transferred into any ERP of interest to better inform DOC diagnoses.

Decoding Subject-Driven Cognitive States from EEG Signals for Cognitive Brain–Computer Interface

In this study, we investigated the feasibility of using electroencephalogram (EEG) signals to differentiate between four distinct subject-driven cognitive states: resting state, narrative memory, music, and subtraction tasks. EEG data were collected from seven healthy male participants while performing these cognitive tasks, and the raw EEG signals were transformed into time–frequency maps using continuous wavelet transform. Based on these time–frequency maps, we developed a convolutional neural network model (TF-CNN-CFA) with a channel and frequency attention mechanism to automatically distinguish between these cognitive states. The experimental results demonstrated that the model achieved an average classification accuracy of 76.14% in identifying these four cognitive states, significantly outperforming traditional EEG signal processing methods and other classical image classification algorithms. Furthermore, we investigated the impact of varying lengths of EEG signals on classification performance and found that TF-CNN-CFA demonstrates consistent performance across different window lengths, indicating its strong generalization capability. This study validates the ability of EEG to differentiate higher cognitive states, which could potentially offer a novel BCI paradigm.

Exploring Contrastive Learning Neural-Congruency on EEG Recording of Children with Dyslexia

Electroencephalogram (EEG) recordings of children are often used to study the underlying neural basis of causal factors of reading disorders and dyslexia. However, the inter-subject variability in EEG and the unconstrained nature of reading experiments used to elicit these factors made it challenging for traditional EEG analysis methods to extract neural components of these factors. In this work, we aim to explore the use of novel deep neural network architectures and contrastive learning methods to overcome the methodological limitations of traditional techniques and enhance the extraction process of neural components during reading tasks. Notably, we formulate a neural network architecture to extract EEG embedding using contrastive loss that maximizes the neural congruency in non-dyslexic children compared to children with dyslexia. We plan to evaluate our approach on three EEG datasets involving children with dyslexia performing Rapid Automatized Naming (RAN) and Phonological Processing (PA) tasks. The proposed contrastive learning framework will provide an enhanced tool to facilitate studying the neural underpinnings of naming speed and their association with reading performance and related difficulties.

Fast reconstruction of EEG signal compression sensing based on deep learning

When traditional EEG signals are collected based on the Nyquist theorem, long-time recordings of EEG signals will produce a large amount of data. At the same time, limited bandwidth, end-to-end delay, and memory space will bring great pressure on the effective transmission of data. The birth of compressed sensing alleviates this transmission pressure. However, using an iterative compressed sensing reconstruction algorithm for EEG signal reconstruction faces complex calculation problems and slow data processing speed, limiting the application of compressed sensing in EEG signal rapid monitoring systems. As such, this paper presents a non-iterative and fast algorithm for reconstructing EEG signals using compressed sensing and deep learning techniques. This algorithm uses the improved residual network model, extracts the feature information of the EEG signal by one-dimensional dilated convolution, directly learns the nonlinear mapping relationship between the measured value and the original signal, and can quickly and accurately reconstruct the EEG signal. The method proposed in this paper has been verified by simulation on the open BCI contest dataset. Overall, it is proved that the proposed method has higher reconstruction accuracy and faster reconstruction speed than the traditional CS reconstruction algorithm and the existing deep learning reconstruction algorithm. In addition, it can realize the rapid reconstruction of EEG signals.

A neurophysiological basis for aperiodic EEG and the background spectral trend

Electroencephalograms (EEGs) display a mixture of rhythmic and broadband fluctuations, the latter manifesting as an apparent 1/f spectral trend. While network oscillations are known to generate rhythmic EEG, the neural basis of broadband EEG remains unexplained. Here, we use biophysical modelling to show that aperiodic neural activity can generate detectable scalp potentials and shape broadband EEG features, but that these aperiodic signals do not significantly perturb brain rhythm quantification. Further model analysis demonstrated that rhythmic EEG signals are profoundly corrupted by shifts in synapse properties. To examine this scenario, we recorded EEGs of human subjects being administered propofol, a general anesthetic and GABA receptor agonist. Drug administration caused broadband EEG changes that quantitatively matched propofol’s known effects on GABA receptors. We used our model to correct for these confounding broadband changes, which revealed that delta power, uniquely, increased within seconds of individuals losing consciousness. Altogether, this work details how EEG signals are shaped by neurophysiological factors other than brain rhythms and elucidates how these signals can undermine traditional EEG interpretation.

Deep learning approach for detection of unfavorable driving state based on multiple phase synchronization between multi-channel EEG signals

In recent years, electroencephalogram (EEG) signals have gained significant popularity in the detection and recognition of drivers' unfavorable driving states (UDS) and related aspects. However, the methods that rely on traditional EEG features and evaluate different electrodes or brain regions separately have somewhat limited the further development of recognition effectiveness. To address this, this study analyzes the phase relationships between electrodes in EEG signals to reveal the functional networks and connectivity patterns within the brain. Three phase synchronization (PS) methods, namely phase locking value (PLV), corrected imaginary phase locking value (CIPLV), and weighted phase lag index (WPLI), are used to construct connectivity matrices and networks in six classic sub-rhythmic bands. Additionally, a 2D convolutional neural network (CNN) with fewer learning parameters is designed to automatically learn high-level abstractions and the most discriminative features, thereby avoiding the laborious and subjective process of manual feature extraction. The proposed method has been compared with traditional classification algorithms using a leave-one-out cross-validation strategy. The comparative results indicate that our proposed approach, which combines WPLI phase synchronization in the gamma1 sub-rhythmic band with a pyramid-structured CNN model, achieves significantly higher accuracy in recognizing drivers' UDS than other conventional methods through a one-way analysis of variance (ANOVA) followed by post hoc multiple comparisons with Tukey analysis. Furthermore, our proposed 2D-CNN model contains fewer parameters, making it more competitive for further practical applications in real-world driving scenarios.

Portable Headband Electroencephalogram in the Detection of Absence Epilepsy.

The accuracy of headband electroencephalogram (EEG) was compared to traditional EEG in pediatric patients with absence epilepsy. This study enrolled 10 patients with previously diagnosed absence epilepsy and examined the concordance of headband EEG and traditional EEG in the follow-up EEG of treated absence epilepsy. The study found a concordant result in 80% of cases providing a signal that absence epilepsy is an effective target for headband EEG. The study showcases a need for further research in headband EEG technology and continued improvements in technology.

Advanced framework for epilepsy detection through image-based EEG signal analysis.

Recurrent and unpredictable seizures characterize epilepsy, a neurological disorder affecting millions worldwide. Epilepsy diagnosis is crucial for timely treatment and better outcomes. Electroencephalography (EEG) time-series data analysis is essential for epilepsy diagnosis and surveillance. Complex signal processing methods used in traditional EEG analysis are computationally demanding and difficult to generalize across patients. Researchers are using machine learning to improve epilepsy detection, particularly visual feature extraction from EEG time-series data. This study examines the application of a Gramian Angular Summation Field (GASF) approach for the analysis of EEG signals. Additionally, it explores the utilization of image features, specifically the Scale-Invariant Feature Transform (SIFT) and Oriented FAST and Rotated BRIEF (ORB) techniques, for the purpose of epilepsy detection in EEG data. The proposed methodology encompasses the transformation of EEG signals into images based on GASF, followed by the extraction of features utilizing SIFT and ORB techniques, and ultimately, the selection of relevant features. A state-of-the-art machine learning classifier is employed to classify GASF images into two categories: normal EEG patterns and focal EEG patterns. Bern-Barcelona EEG recordings were used to test the proposed method. This method classifies EEG signals with 96% accuracy using SIFT features and 94% using ORB features. The Random Forest (RF) classifier surpasses state-of-the-art approaches in precision, recall, F1-score, specificity, and Area Under Curve (AUC). The Receiver Operating Characteristic (ROC) curve shows that Random Forest outperforms Support Vector Machine (SVM) and k-Nearest Neighbors (k-NN) classifiers. The suggested method has many advantages over time-series EEG data analysis and machine learning classifiers used in epilepsy detection studies. A novel image-based preprocessing pipeline using GASF for robust image synthesis and SIFT and ORB for feature extraction is presented here. The study found that the suggested method can accurately discriminate between normal and focal EEG signals, improving patient outcomes through early and accurate epilepsy diagnosis.

The Value of Long-term Video EEG Monitoring to Diagnose and Track Childhood Epilepsy.

Long-term video-EEG monitoring (LTM) is a new technique to assess and track fluctuations, classify seizures, identify epileptic syndromes, and determine the number of seizures and epilepsy-simulating disorders. The present study aims to evaluate the concordance of traditional EEG and LTM in assessing childhood epilepsy. This cross-sectional before-after study was performed on 120 children with epilepsy who were referred to the Epilepsy Monitoring Unit (EMU) at the Children's Medical Center between September 2021 and September 2022 and were monitored for at least eight hours in this unit. The source of the study information collection was the patients' recorded files. A neurologist reviewed the primary EEGs, and two experts blindly reviewed and interpreted the patients' LTMs under a clinical neurophysiologist's supervision. The diagnoses changed after employing LTM in most children with epilepsy. Based on the diagnostic agreement analysis between EEG and LTM, the coefficient value for LTM was calculated at -0.37 (p= 0.229), showing that LTM has significantly expanded patients' diagnoses and care plans. The use of LTM improves the diagnosis, classification, and monitoring of epilepsy in affected children and can be a reliable supplement to EEG in some instances.

Differential diagnosis of Lewy body dementias using multivariate EEG classifiers

AbstractBackgroundLewy body dementia (LBD) results from pathology in ɑ‐synuclein and is the second most common cause of dementia. LBD has two manifestations: Parkinson’s disease dementia (PDD) and dementia with Lewy bodies (DLB). There is currently no non‐invasive, fast, reliable way to distinguish between early‐stage PDD, DLB, and Alzheimer’s disease (AD). Although previous traditional EEG visual rating methods have not been useful in delineating dementia subtypes, resting‐state electroencephalography (rsEEG) is a promising method for differential diagnosis of dementia, because it is inexpensive, non‐invasive, and widely available for clinical use. Here, we demonstrate how a multivariate classifier using rsEEG as input can differentiate PDD, DLB, and Alzheimer’s disease.MethodWe analyzed rsEEG data from older adults (N = 133, age 59‐89) with AD (n = 33), DLB (n = 27) or PDD (n = 45). We computed the aperiodic and periodic parameterization of brain activity, along with metrics of variation in fractal dimension within specific frequency bands. We then used these features, along with age and sex, to develop XGBoost machine‐learning models to predict patient diagnoses. Models were evaluated using cross‐validation with balanced accuracy, weighted sensitivity, and weighted specificity.ResultThe first task was a two‐way classification: Control vs Dementia (accuracy = 86%, sensitivity = 93%, specificity = 79%, p &lt; .001). The second task was a three‐way classification: AD vs DLB vs PDD (accuracy = 66%, sensitivity = 70%, specificity = 82%, p &lt; .001). We also evaluated two‐way models distinguishing between AD vs DLB (accuracy = 86%, sensitivity = 87%, specificity = 86%, p &lt; .001), AD vs PDD (accuracy = 84%, sensitivity = 85%, specificity = 83%, p &lt; .001), AD vs LBD (combined DLB and PDD; accuracy = 82%, sensitivity = 85%, specificity = 78%, p &lt; .01), and DLB vs PDD (accuracy = 68%, sensitivity = 71%, specificity = 65%, p &lt; .05).ConclusionMachine‐learning and rsEEG are capable of distinguishing between AD and LBD. Moreover, machine‐learning combined with spectral and fractal features can distinguish between DLB and PDD, despite both forms of dementia arising from ɑ‐synuclein pathology.

Characterizing EEG signals of meditative states using persistent homology and Hodge spectral entropy

We present a novel topological characterization of EEG time series signals associated with two meditative states: autobiographical self-reflection and mantra-based meditation. The results demonstrate significant information geometric differences among the two groups of meditators and the control group for two brain regions, namely anterior frontal (AF) and temporoparietal (TP). The analysis involves global topological invariants and persistent homology features of the point cloud data constructed from the EEG time series using Takens’s embedding. While traditional EEG analyses face the well-known issues of noise and artefacts, the persistent homology approach is robust to them due to its emphasis on the overall topological changes rather than the transient structures. We analyse the characteristics of the Betti areas, peak location of Betti curves, persistent entropy, persistent amplitude and Hodge spectral entropy of the meditative states for up to the second homology dimension. We find that the mantra- based meditative state is rich in persistent features compared to autobiographical self-reflection. Also, the entropic measures indicate reduced chaotic dynamics and enhanced topological structure formation for the former. Moreover, the correlation of Hodge spectral entropy with persistent entropy or Betti area in a group is a key feature in differentiating the TP and AF channels. These topological features can be used to classify the meditative states with more than 95% accuracy.

Deep generative adversarial networks with marine predators algorithm for classification of Alzheimer’s disease using electroencephalogram

Alzheimer's disease (AD) is a neurological disorder characterized by cognitive decline and memory loss. An early and precise diagnosis of Alzheimer's disease is critical for effective therapy and management. The electroencephalogram (EEG) has shown promise as a non-invasive and cost-effective tool for Alzheimer's disease (AD) categorization. The capacity to diagnose Alzheimer's disease at an early stage is one of the benefits of EEG (electroencephalogram) over other methodologies in Alzheimer's disease research. Traditional EEG analysis approaches, such as estimating coherence between various pairs of electrodes, necessitate a significant amount of human labor. We introduce a novel strategy for AD classification based on EEG by combining deep generative adversarial networks (GANs) and the Marine Predators Algorithm (MPA). GANs are powerful deep-learning models that can be trained to generate realistic samples via negative generator and discriminator training. MPA is a nature-inspired optimization method well-known for its ability to solve complicated optimization problems. The proposed system generates synthetic EEG samples using GANs' ability to learn meaningful representations from raw EEG data. The MPA is then employed to enhance classification performance by optimizing the discriminative features extracted by GANs. The MPA simulates the hunting behavior of marine predators, facilitating the exploration of the feature space and identifying significant characteristics for AD categorization. We evaluated the efficacy of the proposed strategy using a publicly accessible EEG database of AD patients and healthy controls. The results of classification precision, sensitivity, and specificity demonstrate that deep GANs with MPA are superior to state-of-the-art techniques.

Aided diagnosis of cervical spondylotic myelopathy using deep learning methods based on electroencephalography

Cervical spondylotic myelopathy (CSM) is the most severe type of cervical spondylosis. It is challenging to achieve early diagnosis with current clinical diagnostic tools. In this paper, we propose an end-to-end deep learning approach for early diagnosis of CSM. Electroencephalography (EEG) experiments were conducted with patients having spinal cord cervical spondylosis and age-matched normal subjects. A Convolutional Neural Network with Long Short-Term Memory Networks (CNN-LSTM) model was employed for the classification of patients versus normal individuals. In contrast, a Convolutional Neural Network with Bidirectional Long Short-Term Memory Networks and attention mechanism (CNN-BiLSTM-attention) model was used to classify regular, mild, and severe patients. The models were trained using focal Loss instead of traditional cross-entropy Loss, and cross-validation was performed. Our method achieved a classification accuracy of 92.5 % for the two-class classification among 40 subjects and 72.2 % for the three-class classification among 36 subjects. Furthermore, we observed that the proposed model outperformed traditional EEG decoding models. This paper presents an effective computer-aided diagnosis method that eliminates the need for manual extraction of EEG features and holds potential for future auxiliary diagnosis of spinal cord-type cervical spondylosis.