Result Details
A Deep Learning Approach to EEG-Based Diagnosis of Cognitive Skills Impairment: Electrode-Level Analysis Insights
Early identification of cognitive skill impairments is crucial for timely clinical intervention. This study utilizes EEG data to classify cognitive states into three categories: no impairment, mild impairment, and severe impairment. EEG signals from 88 participants were segmented into 10 -second windows and analyzed using a modified EEGNet deep learning architecture, achieving a classification accuracy of 89 %. In addition to classification, statistical analyzes including Welch's t test and Benjamini-Hochberg's FDR correction were used to identify electrodes significantly affected, particularly F3, F4, C3, C4,T3,T4,P3,P4,O1,O2,Fz,Cz and Pz, implanted in memory, attention, and executive functions. Topographic brain activation maps highlighted these regional abnormalities, while spectral analysis revealed altered frequency band distributions across impairment levels. Connectivity analysis also showed decreased functional integration between brain regions in mild and severe cases. Combining deep learning, statistical inference, and EEG-based network features presents a robust framework for diagnosing and interpreting cognitive impairments.
EEG, Cognitive Impairment, EEGNet, Brain Connectivity, Spectral Analysis, Mild Cognitive Impairment, Dementias, EEG Classification
@inproceedings{BUT198330,
author="Yasir {Hussain} and {} and Aamir Saeed {Malik}",
title="A Deep Learning Approach to EEG-Based Diagnosis of Cognitive Skills Impairment: Electrode-Level Analysis Insights",
booktitle="2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI)",
year="2025",
pages="1470--1475",
publisher="IEEE",
address="Athens, Greece",
doi="10.1109/ICTAI66417.2025.00213",
isbn="979-8-3315-4919-0",
url="https://ieeexplore.ieee.org/document/11272536"
}