Application of deep learning in MRI classification of Schizophrenia

Abstract

In today’s world, when people are suffering from complex brain diseases, MRI has been playing a very significant part in understanding brain functionalities and its abnormalities. Deep learning has been recently used for the analysis of MRI, fMRI, structural MRI etc. and through this, we have achieved better performance than traditional computer-aided diagnosis for brain disorders. However, similar compo sition of brain diseases makes it hard to find out and differentiate the accuracy of exact disease from the acquired neuroimaging data. Accordingly, in this paper, a multi channel 2D CNN based architecture was implemented on COBRE dataset 1 which presents a significantly high accuracy over some models. Our modified multichannel 2D CNN architecture achieves around 97% accuracy which improves our classification performance. Furthermore, the paper discusses the boundaries of existing studies, the DL methods and present future possible directions.