Discriminative And Predictive Analysis Of Schizophrenia Studied With Multi-modal Magnetic Resonance Imaging

Schizophrenia is a severe mental disorder with hallucination,delusion,apathy,social withdrawal,insight and working memory dysfunctions.The lack of explicit biomarkers in the diagnosis of schizophrenia increases chances of misdiagnosis rate and missed diagnosis rate.We performed discriminative and predictive analyses in schizophrenia based on multimodal magnetic resonance imaging(MRI)data.We made systematic comparisons among different classification models and regression models.In addition,we discussed the most discriminative features and features contributing to the prediction.The main content of this paper are:1.We recruited 44 first-episode schizophrenia(FESZ)patients,44 chronic schizophrenia(CSZ)patients,and 56 normal controls,and acquired their structural MRI and resting-state functional MRI data.We calculated gray matter volume(GMV),degree centrality(DC),amplitude of low frequency fluctuation(ALFF),and regional homogeneity(ReHo)of 90 brain regions,based on an automated anatomical labeling(AAL)atlas.We then applied these features and their combined features as input features of classifiers.Support vector machine(SVM),linear discriminative analysis(LDA),and random forest(RF)were used to classify the FESZ group,the NC group,and the CSZ group.We used nested cross validation to evaluate classification performance.We found that: 1)the SVM with recursive feature elimination(RFE)achieved the best classification performance(maximum accuracy of three groups’ classification is 0.828~0.895);2)LDA achieved less classification accuracy and area of curve(maximum accuracy of three groups’ classification is 0.778~0.832);and 3)RF achieved the worst classification performance(maximum accuracy of three groups’ classification is 0.659~0.768).When we applied the combined features as input features,its classification performance was the best.The most discriminative features for classifications were predominantly located in the emotion and visual systems.2.In the predictive analysis of Positive and Negative syndrome scales(PANSS)total scores,we applied the same features used in the discriminative analyses to compare the five pattern regression models,including support vector regression(SVR),partial least square(PLS)regression,principle component regression(PCR),Lasso regression,and ridge regression.We found that SVR achieved the best prediction accuracy(r=0.416)when we use ReHo as input feature and PLS regression and PCR achieved significant prediction results.The features contributing to the prediction were primarily located in the occipital lobe.This study achieved good classification performance in discriminative analysis of schizophrenia,and significant predicted PANSS total scores.The findings in the thesis may provide potential biomarkers for disease diagnosis and disease state evaluation of schizophrenia.