IJCST Vol. 4, ISSue 1, Jan - MarCh 2013 ISSN : 0976-8491 (Online) | ISSN : 2229-4333 (Print) www.ijcst.com 594 InternatIonal Journal of Computer SCIenCe and teChnology Automated Classiication of Schizophrenia With Neural Networks Gore Ranjana Waman Dept. of Computer, Savitribai Phule Women’s Engineering College, Aurangabad, Maharashtra, India Abstract Schizophrenia is a complex mental disorder. So Identiication of Schizophrenic is very important in quantitative biological research. In this paper, I proposed a method of classiication of schizophrenia and healthy controls, using a neural network and ICA. A reliable technique for discriminating schizophrenia based upon Functional Magnetic Resonance Imaging (fMRI) would be a signiicant advance. fMRI technology enables medical doctors to observe brain activity patterns that represent the execution of subject tasks, both physical and mental. The scans were acquired on 1.5T Siemens scanner. The data was preprocessed Using SPM and then ICA is applied to fMRI data, that has been fruitful in grouping the data into meaningful spatially independent components. Work discussed in this paper speciically focuses on fMRI data collected from both healthy controls and patients diagnosed with schizophrenia. The neural networks are trained using the back propagation algorithm, in which the error signals are propagated backward through the network. In a three layer neural network the weights are updated. The output of neural network will be ‘yes’ or ‘no’ i.e. patient is schizophrenic or not. This is how I classify schizophrenic and healthy controls and got better results. Keywords Schizophrenia, ICA, FMRI, Backpropagation I. Introduction Schizophrenia is a mental disorder with an unknown physiology. Schizophrenia is a complex psychiatric disorder that has eluded a characterization in terms of local abnormalities of brain activity. The objective of this work is to identify biomarkers predictive of schizophrenia based on fMRI data collected for both schizophrenic and non-schizophrenic subjects performing a simple Sirp task in the scanner. Unlike some other brain disorders (e.g., stroke or Parkinson’s disease), schizophrenia appears to be “delocalized”, i.e. dificult to attribute to a dysfunction of some particular brain areas. Functional Magnetic Resonance Imaging technology enables medical doctors to observe brain activity patterns that represent the execution of subject tasks, both physical and mental. These include localizing regions of the brain activated by a task, determining distributed networks that correspond to brain function and making predictions about psychological or disease states. Each of these objectives can be approached through the application of suitable statistical methods. This role can range from determining the appropriate statistical method to apply to a data set, to the development of unique statistical methods geared speciically toward the analysis of fMRI data. My primary goal was to determine the degree to which the spatial maps for schizophrenic patient and healthy control maps discriminated the two groups. It is possible to diagnose schizophrenia using fMRI activation patterns in schizophrenia I present an approach for classifying patients and healthy controls based on fMRI brain activation maps generated using the statistical parametric mapping (SPM) approach [1]. fMRI analysis of schizophrenia was implemented using independent component analysis (ICA) to identify multiple temporally cohesive, spatially distributed regions of brain activity that represent functionally connected networks. With the activation maps, we get better scalability to very large subject pools (e.g., hundreds or thousands of subjects), and activation map has the potential to integrate data at the activation map level that would be technically dificult to combine at the raw data level. In schizophrenia, there are differences in fMRI activation in several brain regions, even in the absence of measurable differences in task performance [2]. Over the past decade, fMRI has emerged as a powerful instrument to collect vast quantities of data about activity in the human brain. A typical fMRI experiment can produce a three-dimensional image related to the human subject’s brain activity every half second, at a spatial resolution of a few millimeters. fMRI is a technique for obtaining three dimensional images related to neural activity in the brain through time. More precisely, fMRI measures the ratio of oxygenated hemoglobin to deoxygenated hemoglobin in the blood with respect to a control baseline, at many individual locations within the brain. It is widely believed that blood oxygen level is inluenced by local neural activity, and hence this blood oxygen level dependent (BOLD) response is generally taken as an indicator of neural activity. I present a method for using a training set of fMRI activation maps to build classiiers for speciic brain conditions. The classiiers evaluate activation maps according to their degree of similarity to the maps in the training set and assign them probabilities of being patient or non-patient. According to the National Institute of Mental Health (NIMH), Schizophrenia is the most chronic and disabling of the severe mental disorders. The disease affects some of the most highly evolved functions in humans such as perception, memory, attention, cognition, and emotion. The cognitive symptoms of schizophrenia, which include dificulties with attention, memory, and problem solving, can create signiicant barriers to a normal and productive life. Finding treatments for these symptoms has been hampered by a lack of scientiic consensus on which cognitive impairments should be targeted for research and what tools are best for measuring them. Although there has been much work showing difference in the fMRI of schizophrenic patients and healthy controls in resting state scans, this research focuses on task-related scans, in which the subject is performing SIRP task. Because of the high dimensionality of fMRI data, a data reduction scheme is typically applied prior to ICA. The dimensionality of the data in fMRI is determined by the repeat time (TR) parameter. This can be changed from scan to scan and has no relationship to the number of sources in the brain.ICA is a statistical and computational technique for revealing hidden factors that underlie sets of random variables, measurements, or signals. Feed forward neural networks were used to analyze spatial ICA components extracted from the fMRI data and to classify the subjects as either patients or healthy controls. Artiicial neural networks are computational systems whose architecture and operation are inspired from the knowledge about biological neural cells in the brain. In this paper, the neural networks are trained