Classification of schizophrenia from feature-model analysis of bilaterally correlated diagnosis, symptoms, and imaging findings pyramid

Document Type : Original Article


1 Deptartment of Computer Science, ITM (SLS) Baroda University, Vadodara, India

2 DANA Brain Health Institute, Iranian Neuroscience Society-Fars Chapter, Shiraz, Iran

3 Department of Neuroscience, School of Advanced Medical Sciences and Technologies, Shiraz University of Medical Sciences, Shiraz, Iran

4 Provost, Techno India JNR, Institute of Technology, Udaipur 313003, Rajasthan, India

5 Neuroscience Center, Instituto de Investigaciones Científicas y Servicios de Alta Tecnología (INDICASAT AIP), City of Knowledge, Panama City, Republic of Panama


Schizophrenia (SZ) is a mental illness that impairs a person's mental capacity, emotional
dispositions, and personal and social quality of life. Manual SZ patient screening is timeconsuming,
expensive, and prone to human mistakes. As a result, a autonomous, relatively
accurate, and reasonably economical system for diagnosing schizophrenia patients is
required. Machine learning methods are capable of learning subtle hidden patterns from
high dimensional imaging data and achieve significant correlations for the classification
of Schizophrenia. In this study, the diverse types of symptoms of the affected person are
selected which have the weights assigned by cross-correlations and the model classifies
the probability of schizophrenia in the person based on the highest weighted symptoms
present in the report of the patient using machine learning classifiers. The classification
is made by various classifiers in which the Support Vector Machine (SVM) gives the
best result. In the neuroscience domain, it has been one of the most popular machinelearning
tools. SVM with Radial Basis Function kernel helps to distinguish between
patients and healthy controls with significant accuracy of 76% without normalization and
Principal Component Analysis (PCA). The K nearest neighbor’s algorithm also with no
normalization and PCA showed an accuracy of 73% in predicting SZ which is remarkably
close to the SVM given the small size dataset.