A data-driven machine learning framework for Alzheimer's disease diagnosis and staging using DRKPCA-ELM on MRI scans
Alzheimer's Disease (AD) is a progressive brain disorder that gradually impairs memory and cognitive abilities. Detecting it early is essential, as timely treatment can help slow its progression and improve the quality of life for those affected. To support early diagnosis, healthcare professio...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-10-01
|
| Series: | Ain Shams Engineering Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2090447925003417 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Alzheimer's Disease (AD) is a progressive brain disorder that gradually impairs memory and cognitive abilities. Detecting it early is essential, as timely treatment can help slow its progression and improve the quality of life for those affected. To support early diagnosis, healthcare professionals benefit from computer-aided diagnosis (CAD) systems, which assist in predicting the stage of the disease with greater accuracy and efficiency. In this paper, we propose a novel feature reduction approach, named Downsized Rank Kernel Principal Component Analysis (DRKPCA), and we present a scheme built using the proposed DRKPCA algorithm together with the Extreme Learning Machine (ELM) classifier, called DRKPCA-ELM. The main objective of the proposed algorithm is to diagnose and classify brain MRIs of Alzheimer's cases and predict the stages of dementia. In the initial phase, the scheme extracts features using deep learning techniques. In the next step, dimensionality is reduced using DRKPCA. The output of DRKPCA serves as input to the ELM classifier. The kernel parameters are tuned by the Tabu search optimization strategy. Cross-validation is implemented to improve the generalization of the DRKPCA-ELM classifier. To validate the proposed DRKPCA-ELM scheme, the method is tested on the ADNI database. This research yields three big conclusions. First, DRKPCA is indeed very effective for dimensionality reduction and has much better performance than other methods serving the same purpose. Second, the DRKPCA-ELM classifier is reliable in differentiating stages of dementia or Alzheimer's Disease from healthy individuals. Third, the proposed approach is robust and should hold when validated against other independent benchmark datasets. |
|---|---|
| ISSN: | 2090-4479 |