Alzheimer's disease diagnosis from MRI and SWI fused image using self adaptive differential evolutionary RVFL classifier

Abstract

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that involves gradual memory loss and eventually leads to severe cognitive decline at the final stage. Advanced neuroimaging modalities, including magnetic resonance imaging (MRI), prove advantageous in diagnosing the severity of the progression of AD. T1-W structural MRI and susceptibility-weighted imaging (SWI) are two of the most popular MRI sequences for medical diagnosis. The formation of tangles in the hippocampus is a major cause of AD development. Significant hippocampal loss and temporal lobe atrophy characterize the progression to AD, which can be visualized using T1-W structural MRI. Recent research has shown that persons with AD have higher levels of iron in their basal ganglia, which causes non-local changes in phase images due to variances in tissue susceptibility. SWI images use phase information to detect the magnetic disturbance and to visualize the iron lesions better. In this paper, we propose the fusion of MRI and SWI images to integrate the structural atrophies and iron lesions accumulation in AD patients. Features from the fused images will be retrieved by a pre-trained deep learning network and categorized using a random vector functional link network (RVFL). Furthermore, a self-adaptive differential evolutionary algorithm will be used to fine-tune the RVFL network's input weights and biases. In order to test the effectiveness of the suggested approach, experiments are done on the publicly available OASIS dataset. The source code of the proposed network is available at/github.com/triptigoel/SaDE-RVFL-for-AD-Diagnosis. © 2025 Elsevier B.V.