The Research On Classification Based On Brain Functional Connectivity Of Schizophrenia

Functional magnetic resonance imaging (fMRI) is the main technical means tostudy the brain structure, function and their relationship with the pathological, and ithas been widely used in brain experiments and clinical research. How to extractmeaningful brain regions from the complex brain fMRI data is significance for thebrain research.In recent years, classification methods in the field of data mining have beenwidely applied to analyze and study the brain fMRI data. Brain functionalconnectivity data is an effective way for analysis of the brain fMRI data. For thisreanson, we proposed two methods to analyze the brain functional connectivity datawhich respectively come from schizophrenias, their healthy siblings and normalpeople. The main works are summarized as follows:In this paper, we proposed one new method for classifying the brain functionalconnectivity (NCBFC). We used an entropy-based discretization method forpreprocessing the brain functional connectivity data to choose the discriminatoryfeature patterns, and then we used those patterns as the input of the support vectormachine for classification on the three groups of brain functional connectivity data.After studying and analysis of multi-classification problem, we indirectly proved thegenetic characteristics of schizophrenia by the pattern classification analysis, and ourmethod has superior performance on the sensitivity, specificity and accuracy from theexperimental result.Strong Jumping Emerging Pattern (SJEP) has strong ability for patternclassification; it is aimed at work out low accuracy of pattern classification by JEPpattern because of its low support. For this reason, this paper proposed a classificationmethod C_SJEP which is based on SJEPs. C_SJEP used border algorithm to mineSJEP pattern, and predicted the test samples’ attributes by PCL method. Theexperimental results show that C_SJEP method has a higher accuracy when comparedwith the other classification methods.From the results of classification of brain functional connectivity data, we have amore profound understanding of how to utilize such data to analyze mechanism ofhuman brain on the physiological and pathological, this will provide effective way forthe detection and diagnosis of mental illness; and we also have a deeper understanding of the analysis of multi-class problems. This will promote the patternclassification methods to develop in the brain science and clinical research.