Recognizing Brain Regions Of Schizophrenia Using Magnetitic Resonance Imaging And Machine Learning

Abstract

With the use of machine learning (ML) and neural images, one can find difference between schizophrenia (SZ) patients and normal controls (NCs).

The two useful techniques for conducting an inquiry into abnormalities in human brain are Functional MRI (fMRI) and structural MRI (sMRI) and with growing and fast development, the studies have revealed that these brain data can be a benchmark in the medical field for detecting such psychiatric disorder with high accuracy. There is no denying fact that Machine learning (ML) these days have been widely in a source of detecting neurobiological tendency and neuropsychiatry illness, with high degree of precision. To diagnose the affected area of brain in schizophrenia we can use the neural images which prove to be very beneficial and can be used to references in medical field. In this research article, sMRIs is used for detecting and comparing schizophrenia patient and normal controls. The techniques which are used here requires hyperparameter optimization of a neural network. This is called coarse-to-fine feature selection which is a Machine Learning substructure. For proposing this, it require two sample t-tests for pulling out the contrast feature within the group, after that inapplicable, irrelevant, and superfluous property is eliminated and lastly by considering the gray matter (GM) and white matter (WM) area selection, a decision model is built with the help of support vector machine (SVM). Here rather than focussing on group measure, this study is focussed on individual measure and have proposed a model that is widely used with high rate of interest unlike the previously modelled which reported differences of both group and individual measures leading to less involvement and application.  By closely studying and demonstrating the result of experiment, here it is seen that the proposed structure determines highest accuracy of classification between schizophrenia (SZ) patients and normal controls (NCs) over up to 80%. With this proposed method, one can also determines various other disease with accurate results.