Working Memory Functional Connectivity Analysis for Dementia Classification using EEG
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911211151425536 |
|---|---|
| author | Ranjan, Shivani Jain, Anant Badal, Robin Kumar, Amit Shende, Harshal Joshi, Deepak Yadav, Pramod Kumar, Lalan |
| author_facet | Ranjan, Shivani Jain, Anant Badal, Robin Kumar, Amit Shende, Harshal Joshi, Deepak Yadav, Pramod Kumar, Lalan |
| contents | Background: Dementia, particularly Alzheimer's Disease (AD), is a progressive neurodegenerative disorder marked by cognitive decline. Early detection, especially at the Mild Cognitive Impairment (MCI) stage, is essential for timely intervention. Working Memory (WM) impairment is a key early indicator of neurodegeneration, affecting higher cognitive processes. Electroencephalography (EEG), with its high temporal resolution, offers a cost-effective method to assess brain dynamics. This study investigates WM-related EEG functional connectivity (FC) to identify brain network alterations across dementia stages. Methods: EEG signals were recorded from 24 participants (8 AD, 8 MCI, and 8 healthy controls) during WM tasks, including encoding, recall, and retrieval stages. Data preprocessing involved noise reduction and feature extraction using Spherical and Head Harmonic Decomposition (SHD, HHD). FC was quantified using Cross-Plot Transition Entropy (CPTE) and Phase Lag Index (PLI). Network metrics such as Degree and Eigenvector Centrality were analyzed using Support Vector Machine, Random Forest, and XGBoost classifiers. Results: The CPTE-based connectivity metrics outperformed the traditional PLI approach in differentiating dementia stages, attaining a peak classification accuracy of 97.53% during the retrieval phase with the Random Forest model. A connectivity threshold of 0.5 was optimal for network discrimination. SHD and HHD features also demonstrated strong discriminative potential. AD subjects exhibited higher synchronization patterns during WM tasks than healthy controls. Conclusions: The integration of WM tasks with EEG-based FC analysis provides a robust framework for dementia classification. The proposed CPTE-based approach offers a robust, scalable, non-invasive, and effective diagnostic tool for early detection and monitoring of neurodegenerative diseases. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_13399 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Working Memory Functional Connectivity Analysis for Dementia Classification using EEG Ranjan, Shivani Jain, Anant Badal, Robin Kumar, Amit Shende, Harshal Joshi, Deepak Yadav, Pramod Kumar, Lalan Signal Processing Background: Dementia, particularly Alzheimer's Disease (AD), is a progressive neurodegenerative disorder marked by cognitive decline. Early detection, especially at the Mild Cognitive Impairment (MCI) stage, is essential for timely intervention. Working Memory (WM) impairment is a key early indicator of neurodegeneration, affecting higher cognitive processes. Electroencephalography (EEG), with its high temporal resolution, offers a cost-effective method to assess brain dynamics. This study investigates WM-related EEG functional connectivity (FC) to identify brain network alterations across dementia stages. Methods: EEG signals were recorded from 24 participants (8 AD, 8 MCI, and 8 healthy controls) during WM tasks, including encoding, recall, and retrieval stages. Data preprocessing involved noise reduction and feature extraction using Spherical and Head Harmonic Decomposition (SHD, HHD). FC was quantified using Cross-Plot Transition Entropy (CPTE) and Phase Lag Index (PLI). Network metrics such as Degree and Eigenvector Centrality were analyzed using Support Vector Machine, Random Forest, and XGBoost classifiers. Results: The CPTE-based connectivity metrics outperformed the traditional PLI approach in differentiating dementia stages, attaining a peak classification accuracy of 97.53% during the retrieval phase with the Random Forest model. A connectivity threshold of 0.5 was optimal for network discrimination. SHD and HHD features also demonstrated strong discriminative potential. AD subjects exhibited higher synchronization patterns during WM tasks than healthy controls. Conclusions: The integration of WM tasks with EEG-based FC analysis provides a robust framework for dementia classification. The proposed CPTE-based approach offers a robust, scalable, non-invasive, and effective diagnostic tool for early detection and monitoring of neurodegenerative diseases. |
| title | Working Memory Functional Connectivity Analysis for Dementia Classification using EEG |
| topic | Signal Processing |
| url | https://arxiv.org/abs/2510.13399 |