Discriminative analysis of schizophrenia using support vector machine and recursive feature elimination on structural MRI images

Structural abnormalities in schizophrenia (SZ) patients have been well documented with structural magnetic resonance imaging (MRI) data using voxel-basedmorphometry (VBM) and region of interest (ROI) analyses. However, these analyses can only detect group-wise differences and thus, have a poor predictive value for individuals. In the present study, we applied a machine learning method that combined support vectormachine (SVM)with recursive feature elimination (RFE) to discriminate SZ patients from normal controls (NCs) using their structural MRI data.We first employed both VBM and ROI analyses to compare gray matter volume (GMV) andwhite matter volume (WMV) between 41 SZ patients and 42 ageand sex-matched NCs. The method of SVM combined with RFE was used to discriminate SZ patients fromNCs using significant between-group differences in both GMV andWMV as input features.We found that SZpatients showedGMandWMabnormalities in several brain structuresprimarily involved in the emotion,memory, and visual systems. An SVMwith a RFE classifier using the significant structural abnormalities identified by the VBM analysis as input features achieved the best performance (an accuracy of 88.4%, a sensitivity of 91.9%, and a specificity of 84.4%) in the discriminative analyses of SZ patients. These results suggested that distinct neuroanatomical profiles associated with SZ patients might provide a potential biomarker for disease diagnosis, and machine-learning methods can reveal neurobiological mechanisms in psychiatric diseases. Abbreviations: ANOVA = analysis of variance, AUC = area under the curve, CAL = calcarine, CAU = caudate, CSF = cerebrospinal fluid, CUN = cuneus, FDR = false discovery rate, FFG = fusiform gyrus, FWHM = full-width at half-maximum, GLM = general linear model, GM= graymatter, GMV= gray matter volume, HIP= hippocampus, LING= lingual gyrus, MFG=middle frontal gyrus, MOG =middle occipital gyrus, MRI =magnetic resonance imaging, NCs = normal controls, PANSS = Positive and Negative Syndrome Scale, PCUN = precuneus, PHG = parahippocampal gyrus, PoCG = postcentral gyrus, RFE = recursive feature elimination, RGMV = regional gray matter volume, ROC = receiver operating characteristic, ROI = region of interest, RWMV = regional white matter volume, SCID = structured clinical interview according to the DSM-IV-TR, STG = superior temporal gyrus, SVM= support vector machine, SZ = schizophrenia, TBV = total brain volume, THM = thalamus, VBM = voxel-based morphometry, WM = white matter, WMV = white matter volume.