Early Classification of Abnormal Health using Longitudinal Structural MRI Data

This paper investigates extensively Machine Learning (ML) algorithms for abnormal health conditions using two sessions of Structural Magnetic Resonance Imaging (sMRI) data. First abnormal health condition considered is classification of early MCI (EMCI) and Cognitively Normal (CN) subjects using benchmark OASIS-3 longitudinal neuroimaging dataset. The second abnormal health condition considered is classification of Autism Spectrum Disorders (ASDs) from Typical Development (TD) subjects using sMRI images from multisite ABIDE II longitudinal samples. Freesurfer longitudinal processing technique is used to process the longitudinal sMRI samples and extract the features. Our work is first of its kind and serves as baseline for future research in longitudinal analysis.