Chun-Shu Wei

Electroencephalography (EEG) is a promising tool for automated detection of mild cognitive impairment (MCI) and dementia, but comparisons across studies are limited by inconsistent datasets and evaluation protocols. This study benchmarks ten deep learning models across four resting-state EEG datasets and eight binary classification tasks using a unified preprocessing pipeline and five-fold subject-wise cross-validation. Each experiment was repeated ten times. SCCNet obtained the highest mean subject-level accuracy, sensitivity, and F1 score, while ShallowConvNet achieved the highest mean segment-level accuracy, specificity, and precision. Subject-level aggregation improved mean accuracy for all evaluated models, and performance varied substantially across datasets and diagnostic tasks. Higher computational cost did not consistently correspond to better classification performance, with several compact architectures remaining competitive with substantially larger models. The results provide a reproducible reference for comparing EEG-based dementia classification models under consistent subject-independent evaluation conditions.

The YOTO (You Only Think Once) dataset presents a human electroencephalog- raphy (EEG) resource for exploring multisensory perception and mental imagery. The study enrolled 26 participants who performed tasks involving both unimodal and multimodal stimuli. Researchers collected high-resolution EEG signals at a 1000 Hz sampling rate to capture high-temporal-resolution neural activity related to internal mental representations. The protocol incorporated visual, auditory, and combined cues to investigate the integration of multiple sensory modalities, and participants provided self-reported vividness ratings that indicate subjec- tive perceptual strength. Technical validation involved event-related potentials (ERPs) and power spectral density (PSD) analyses, which demonstrated the reli- ability of the data and confirmed distinct neural responses across stimuli. This dataset aims to foster studies on neural decoding, perception, and cognitive mod- eling, and it is publicly accessible for researchers who seek to advance multimodal mental imagery research and related applications.

Schizophrenia poses diagnostic challenges due to a lack of objective assessment. We propose MBSzEEGNet, a multi-branch deep-learning (DL) model for robust and interpretable EEG-based schizophrenia classification. Its specialized branches capture oscillatory and spatial-spectral features, enhancing generalization across two resting-state schizophrenia EEG datasets. MBSzEEGNet consistently outperforms leading DL architectures, achieving up to 85.71% subject-wise accuracy on one dataset and 75.64% on the other. Saliency-based explanations highlight potential biomarkers in the delta (0.5-4 Hz) and alpha (8-12 Hz) bands and the temporal and right parietal region. Our findings suggest that integrating explainable multi-branch DL architecture with EEG can enhance schizophrenia diagnosis and provide deeper insights into schizophrenia-related neural markers.

Closed-loop neuromodulation is a promising personalized treatment for various neuropsychiatric disorders, delivering precise stimuli based on real-time brain signals. However, its clinical potential is currently limited by technical challenges inherent in its real-time nature. This article addresses the primary technical challenges in EEG-Phase-Synchronized Transcranial Magnetic Stimulation (TMS), including poor stimulation accuracy and inefficient biomarker detection (correlated with deadlock). These challenges arise from the vulnerability of existing algorithms to filter edge effects. Inspired by predictive coding theory in neuroscience, we propose a novel signal padding method (forecast padding) to mitigate the filter edge effect. To properly quantify the improvements that forecast padding brings about in real-world systems, we introduce a novel delay-relevant validation framework and demonstrate its reliability using experimental data from a real system. Through this framework, we demonstrate that forecast padding significantly improves both stimulation accuracy and deadlock rate. Given the pervasive impact of filter edge effects in closed-loop neuromodulation and other signal processing domains, forecast padding shows broad application potential across various fields.

The YOTO (You Only Think Once) dataset presents a human electroencephalography (EEG) resource for exploring multisensory perception and mental imagery. The study enrolled 26 participants who performed tasks involving both unimodal and multimodal stimuli. Researchers collected high-resolution EEG signals at a 1000 Hz sampling rate to capture high-temporal-resolution neural activity related to internal mental representations. The protocol incorporated visual, auditory, and combined cues to investigate the integration of multiple sensory modalities, and participants provided self-reported vividness ratings that indicate subjective perceptual strength. Technical validation involved event-related potentials (ERPs) and power spectral density (PSD) analyses, which demonstrated the reliability of the data and confirmed distinct neural responses across stimuli. This dataset aims to foster studies on neural decoding, perception, and cognitive modeling, and it is publicly accessible for researchers who seek to advance multimodal mental imagery research and related applications.

Closed-loop neuromodulation is a promising personalized treatment for various neuropsychiatric disor-ders, delivering precise stimuli based on real-time brain signals. However, its clinical potential is currently limited by technical challenges inherent in its real-time nature. This article addresses the primary technical challenges in EEG-Phase-Synchronized Transcranial Magnetic Stimulation (TMS), including poor stimulation accuracy and inefficient biomarker detection (correlated with deadlock). These challenges arise from the vulnerability of existing algorithms to filter edge effects. Inspired by predictive coding theory in neuroscience, we propose a novel signal padding method (forecast padding) to mitigate the filter edge effect. To properly quantify the improvements that forecast padding brings about in real-world systems, we introduce a novel delay-relevant validation framework and demonstrate its reliability using experimental data from a real system. Through this framework, we demonstrate that forecast padding significantly improves both stimulation accuracy and deadlock rate. Given the pervasive impact of filter edge effects in closed-loop neuromodulation and other signal processing domains, forecast padding shows broad application potential across various fields.

Mild cognitive impairment (MCI) is an early stage of non-age-related cognitive decline with an increased risk of progressing to dementia. Early detection of MCI is essential for implementing preventative strategies that can delay or prevent the onset of dementia, ultimately improving patient outcomes and reducing healthcare costs. electroencephalograms (EEGs) and event-related potentials (ERPs) have shown significant promise in detecting MCI due to their affordability, real-time monitoring capabilities, and non-invasiveness. EEG provides continuous brain activity data, while ERPs offer insights into specific cognitive processes by analyzing brain responses to stimuli. These methods can complement each other in MCI diagnosis by providing a comprehensive view of overall brain function and detailed information on specific cognitive processes. However, EEG and ERP are susceptible to noise and inter-individual variability, which can hinder their reliability. Additionally, applying machine learning models on EEG or ERP for MCI detection presents challenges such as the risk of overfitting and difficulties in interpreting the underlying decision-making process. This review emphasizes recent advancements in signal processing and feature extraction methods applied to EEG and ERP data and explores the use of machine learning and deep learning techniques to enhance diagnostic accuracy and interpretative depth. By integrating these methodologies, the review highlights how EEG and ERP can contribute to a more effective understanding and monitoring of cognitive changes associated with MCI, underscoring the importance of early diagnosis for timely intervention and improved patient care. Finally, the review focuses on future research directions, including the development of advanced analytical techniques and multimodal integration approaches involving EEG and ERP to further improve diagnostic accuracy and clinical application.

Steady-state visual-evoked potential (SSVEP)-based brain-computer interfaces (BCIs) offer a non-invasive means of communication through high-speed speller systems. However, their efficiency is highly dependent on individual training data acquired during time-consuming calibration sessions. To address the challenge of data insufficiency in SSVEP-based BCIs, we introduce SSVEP-DAN, the first dedicated neural network model designed to align SSVEP data across different domains, encompassing various sessions, subjects, or devices. Our experimental results demonstrate the ability of SSVEP-DAN to transform existing source SSVEP data into supplementary calibration data. This results in a significant improvement in SSVEP decoding accuracy while reducing the calibration time. We envision SSVEP-DAN playing a crucial role in future applications of high-performance SSVEP-based BCIs. The source code for this work is available at: https://github.com/CECNL/SSVEP-DAN.

Human electroencephalography (EEG) is a brain monitoring modality that senses cortical neuroelectrophysiological activity in high-temporal resolution. One of the greatest challenges posed in applications of EEG is the unstable signal quality susceptible to inevitable artifacts during recordings. To date, most existing techniques for EEG artifact removal and reconstruction are applicable to offline analysis solely, or require individualized training data to facilitate online reconstruction. We have proposed CLEEGN, a novel convolutional neural network for plug-and-play automatic EEG reconstruction. CLEEGN is based on a subject-independent pre-trained model using existing data and can operate on a new user without any further calibration. The performance of CLEEGN was validated using multiple evaluations including waveform observation, reconstruction error assessment, and decoding accuracy on well-studied labeled datasets. The results of simulated online validation suggest that, even without any calibration, CLEEGN can largely preserve inherent brain activity and outperforms leading online/offline artifact removal methods in the decoding accuracy of reconstructed EEG data. In addition, visualization of model parameters and latent features exhibit the model behavior and reveal explainable insights related to existing knowledge of neuroscience. We foresee pervasive applications of CLEEGN in prospective works of online plug-and-play EEG decoding and analysis.