Weighted Kernel Ridge Regression based Randomized Network for Alzheimer's Disease Diagnosis using Susceptibility Weighted Images

Abstract

Alzheimer's disease (AD) is a neurological disorder that primarily affects the elderly and is characterized by cognitive decline and memory loss. Recent research has shown that susceptibility-weighted imaging (SWI) images are useful for diagnosing AD because they reveal abnormally high iron deposition in certain brain regions of people with the disease. Machine learning (ML) algorithms, particularly deep learning (DL) networks, are making incredible strides in AD diagnosis using imaging data to assist physicians in making decisions. The random-vector functional link network (RVFL) is an example of a single-hidden-layer feedforward network that uses a closed-form solution-based approach to offer a variety of feature mapping functions and kernels. In the proposed paper, SWI image features are extracted with a DL network, ResNet 50, and afterward classified with a kernel ridge regression-based RVFL network. To manage data with an unbalanced class distribution, we present a weighted kernel ridge regression-based RVFL network that is capable of generalizing to balanced data. We used SWI images from the publicly accessible OASIS dataset to evaluate the proposed methods for AD diagnosis. Experiment results show that the proposed model outperforms the state-of-the-art models.