Structural biomarker-based Alzheimer's disease detection via ensemble learning techniques

Abstract

Alzheimer's disease (AD) is a degenerative neurological disorder with incurable characteristics. To identify the substantial solution, we used a structural biomarker (structural magnetic resonance imaging) to see the neurostructural changes in the different regions of the brain of AD, mild cognitive impairment, and cognitive normal subjects. In this study, we detected the AD and their subtypes by using the traditional machine learning and ensemble learning models. It is also identified the relative impact score of various cortical and subcortical regions of AD and their subtypes. Experimental study contains two levels of classification: binary and multiclass. The Ensemble_LR_SVM model in binary classification has 99% accuracy in detection. Random forest model in the multiclass has 82% of accuracy. In the cortical-subcortical analysis, the right hemisphere's parahippocampal and entorhinal regions were discovered to be the most influential. Similarly, the inferior temporal and isthmus cingulate regions had a significant influence in the left hemisphere.