Skip to main content
Deep Learning-based Analysis of White Matter Variation in Schizophrenia Patients
Indranath Chatterjee (Tongmyong University, South Korea), Videsha Bansal (JK Lakshmipat University, India), Sunghyun Kim (Tongmyong University, South Korea)
Presenting author:
Indranath Chatterjee, Videsha Bansal, Sunghyun Kim
Schizophrenia is one of the most severe disorders in terms of mental health. Almost 1% of the world's population has been affected by this disorder. In recent times, clinicians began to encounter an increasing number of schizophrenia patients. The clinical manifestation of the disorder includes symptoms such as hallucination, delusion, and disorganized speech, as well as emotional and social isolation. Schizophrenia is diagnosed if any two symptoms persist for longer than six months. The presence of anxiety, depression, and frequent changes in the mood sometimes complicates the diagnosis of this severe disorder. According to the literature, schizophrenia causes both functional and anatomical alterations in the brain. In terms of anatomical changes, variations in the density of white matter in the brain, in addition to grey matter, might be a useful diagnostic tool for physicians to identify the course of the illness. Several approaches have been made to identify these variations using machine learning and deep learning techniques, however, we hardly find a few studies involving deep learning algorithms for white matter analysis of schizophrenia patients. In the study, we attempt to propose a novel deep learning-based framework for the identification of the white matter alterations in the brain. With the help of this computer-aided diagnosis tool, we can trace any small changes in the density of white matter across various parts of the brain. Here, we utilize T1-weighted MRI scans having 1.5T intensity, acquired from 44 schizophrenia patients and 39 healthy controls. After the preprocessing stages, we have extracted the white matter density maps for each subject’s data. Data were processed using a voxel-based method, which allows whole-brain analysis showing the traces of white matter. Furthermore, we propose a novel deep learning-based framework for understanding the variation of white matter changes in schizophrenia patients, while comparing them with healthy controls as a baseline. The result shows that there is a significant decrease in the white matter density in this patient group. This study may bring in new opinions about the pathophysiology of the disorder.