Exploring asymmetrical white matter abnormalities in Alzheimer’s using deep learning

Abstract

Despite the availability of various diagnostic techniques and extensive research efforts aimed at understanding Alzheimer’s disease (AD), accurately and automatically diagnosing AD using biomarkers and comprehending the intricate structural changes in the Alzheimer’s brain using state-of-theart technologies remains a significant challenge. In particular, the asymmetrical white matter abnormalities in the Alzheimer’s brain’s structural connectivity have been poorly studied. To address this critical issue, this paper presents a novel approach that detects AD by feeding the structural hemispherical brain networks to a Convolutional Neural Network (CNN) based classification model and then pinpointing the discriminative asymmetrical in white matter connectivity changes through the interpretations of classification choices. This study found significant outcomes regarding asymmetrical intra-hemispheric connections in AD. These outcomes include distinct connectivity changes in the left and right hemispheres, significant changes primarily in the left hemisphere, discriminative changes involving more subcortical regions in both hemispheres, and increased temporal-subcortical and frontal-subcortical connectivity changes in the left hemisphere. This research has the potential to enhance diagnostic accuracy, improve understanding of the disease, and shed light on its asymmetric nature.