Automated diagnosis of mild cognitive impairment through connectivity analysis of EEG signals and a DL scheme
Abstract There is a hypothesis that deep learning (DL) can enhance the accuracy of diagnosing Alzheimer’s disease (AD) through effectively classifying EEGs from people with AD, mild cognitive impairment (MCI), or healthy aging. To investigate the hypothesis, a new signal processing technique was uti...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Journal of Engineering and Applied Science |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s44147-025-00674-0 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract There is a hypothesis that deep learning (DL) can enhance the accuracy of diagnosing Alzheimer’s disease (AD) through effectively classifying EEGs from people with AD, mild cognitive impairment (MCI), or healthy aging. To investigate the hypothesis, a new signal processing technique was utilized to transform intricate EEGs into input images for a DL framework. For this purpose, an ensemble pre-trained convolutional neural network (CNN) is suggested to classify EEG signals of 66 AD patients, 61 MCI patients, and 60 healthy aging individuals. First, functional connectivity maps are constructed from one-dimensional EEGs through the synchronization likelihood method. Next, extracted images are utilized to retrain five CNNs: VGG-16, ResNet-18, Inception-v3, Inception-v1, and AlexNet. The voting technique is utilized to recognize the final class assignment. Finally, the stratified fivefold cross-validation algorithm is deployed to investigate the categorization efficacy of the suggested approach. The experiment showed that the best accuracy was yielded for the suggested ensemble technique with an average accuracy of 97.90%, sensitivity of 98.40%, and specificity of 97.29%. After the ensemble method, ResNet-18 achieved the best categorization performance with a mean accuracy of 97.38%, sensitivity of 97.81%, and specificity of 96.90%. The obtained results showed that our ensemble CNN model and functional connectivity technique are successfully applied to diagnose AD and MCI patients from resting-state EEG. The findings of this exploration have demonstrated encouraging outcomes; however, additional efforts are necessary to advance and expand this research. |
|---|---|
| ISSN: | 1110-1903 2536-9512 |