Deep Feature Selection and Causal Analysis of Alzheimer’s Disease

• Published in: Frontiers in neuroscience Vol. 13; p. 1198
• Main Authors: Liu, Yuanyuan, Li, Zhouxuan, Ge, Qiyang, Lin, Nan, Xiong, Momiao
• Format: Journal Article
• Language: English
• Published: Lausanne Frontiers Research Foundation 15.11.2019; Frontiers Media S.A
• Subjects: Algorithms; Alzheimer's disease; Biomarkers; Brain; causal inference; Deep learning; Dementia; Diabetic retinopathy; diffusion tensor imaging images; Discriminant analysis; feature selection; Genetic analysis; Genetic diversity; genetic-imaging data analysis; Magnetic resonance imaging; Medical imaging; Medical research; Neural networks; Neuroimaging; Neuroscience; NMR; Nuclear magnetic resonance; Registration; Ventricle; United States > US
• ISSN: 1662-453X; 1662-4548; 1662-453X
• DOI: 10.3389/fnins.2019.01198

Deep convolutional neural networks (DCNNs) have achieved great success for image classifi-cation in medical research. Deep learning with brain imaging is the imaging method of choice for diagnosis and prediction of Alzheimer’s disease (AD). However, it is also well known that DCNNs are ‘black boxes’ due to their low interpretability to humans. The lack of transparency of deep learning compromises its application to the prediction and mechanism investigation in AD. To overcome this limitation, we develop a novel general framework that integrates deep leaning, feature selection, causal inference and genetic-imaging data analysis for predicting and understanding AD. The proposed algorithm not only improves the prediction accuracy, but also can identify the brain regions underlying the development of AD and causal paths from genetic variants to AD via images mediation. The proposed algorithm is applied to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset with diffusion tensor imaging (DTI) in 151 subjects ( 51 AD and 100 non-AD) that were measured at 4 time points of baseline, 6 months, 12 months and 24 months. The algorithm achieved prediction accuracies of more than 92% at all four time points. The algorithm also identified brain regions underlying AD consisting of the temporal lobes (including the hippocampus) and the ventricular system.