A multi-layer perceptron network-based model for classifying stages of alzheimer's disease using clinical data

Abstract

Alzheimer's Disease is a neurodegenerative disorder that progressively impairs individuals' ability to perform daily activities. This irreversible condition cannot be halted once initiated, but early detection may allow for treatments to slow its progression. In this study, clinical data from the Alzheimer's Disease Neuroimaging Initiative dataset were utilized to identify different stages of Alzheimer's and predict the time required for conversion from mild cognitive impairment (MCI) to Alzheimer's Disease. Clinical indicators of Alzheimer's include age, education level, disease progression rate, and cognitive information. Machine learning techniques such as multi-layer perceptron networks, random forests, support vector machines, and decision tree classifiers were employed for binary and multi-class classification of Alzheimer's Disease (AD), Late Mild Cognitive Impairment (LMCI), Early Mild Cognitive Impairment (EMCI), and Cognitive Control (CN). Among these techniques, the multi-layer perceptron network demonstrated superior performance, achieving accuracies of 99.97% for AD vs LMCI, 99.57% for AD vs EMCI, 99.96% for AD vs CN, 95.05% for EMCI vs CN and LMCI vs CN, 99.97% for AD vs LMCI vs CN, 91.2% for EMCI vs LMCI vs AD, 86.25% for CN vs EMCI vs LMCI, 91.94% for CN vs LMCI vs AD, 85.14% for CN vs EMCI vs AD, and 77.5% for AD vs LMCI vs EMCI vs CN. The proposed model has the potential to facilitate early detection and prediction of Alzheimer's stages without the need for imaging scans, thus offering a valuable tool for clinical practice.