Enhancing Alzheimer's Disease Diagnosis: Leveraging Convolutional Neural Networks for Feature Extraction and KNN Classification

Abstract

Detecting Alzheimer's disease (AD) accurately is crucial for early patient management, allowing for proactive steps to minimize irreversible brain damage. AD-related cognitive decline varies in severity, from mild cognitive impairment to mild, moderate, and severe dementia. Early AD detection serves various purposes, such as reducing healthcare costs, slowing brain degeneration, and enhancing treatment outcomes. Machine learning techniques, especially in MRI image analysis, have gained significance in AD diagnosis. In our research, we've employed a deep learning model with convolutional layers for extracting essential information from MRI scans. Additionally, we've utilized the gray wolf meta-heuristic algorithm to identify optimal features. These features are then used in a K-nearest neighbors (KNN) classifier to categorize AD into three distinct types (very mild, mild, and moderate) while distinguishing normal and healthy cases. Our approach achieved an impressive accuracy rate of 95.6% when tested.