Biomarkers have been sought after in the field of schizophrenia research for decades. Diffusion magnetic resonance imaging (dMRI) has been widely used to investigate structural differences between patients with schizophrenia and healthy participants. However, there is no study showing the capability of individualized prediction of the schizophrenia based on the differences in the tract integrity over the whole brain. In this study, we used the tract-based automatic analysis (TBAA) method [1] to obtain whole brain tract integrity for individual subject. The information of integrities, named 2D connectogram, was tested for predicting adult patients with schizophrenia. Subjects were separated into males and females to remove the gender effect. We aim to investigate the capability of individualized prediction of schizophrenia by using the information of whole brain fiber tracts.One hundred and eight schizophrenia patients (males: 54, females: 54) and 144 age-matched healthy controls (males: 70, females: 74) were recruited in the analysis. Images were acquired on a 3T MRI system with a 32-channel head coil (Tim Trio, Siemens, Erlangen, Germany). Diffusion spectrum imaging (DSI) was acquired for 102 diffusion encoding gradients with b_max = 4000 s/mm² (TR/TE = 9600/130 ms, image matrix size = 80 x 80, spatial resolution = 2.5 x 2.5 mm², and slice thickness = 2.5 mm). TBAA method was applied to subjects to assess the whole-brain white matter properties. A 2D connectogram comprising generalized fractional anisotropy (GFA) along predefined 76 major tracts and 100 continuous steps for each tract was estimated as standardized information for each subject. Subjects of males/females were randomly separated into training group and predicting group for 200 permutations. For each permutation, whole brain difference (WD) between training patients and training controls was determined by comparing the 2D connectogram between the two groups. Series of masks were determined to represent the locations containing different effect size (ES) and cluster size (CS) in WD (ES: 0, 0.05, 0.1,…, 1; CS: 1, 2, 3,…, 15). For predicting group, the whole-brain tract information of each subject was assessed by performing TBAA first. A schizo-like index (SLI) was then estimated for each subject by using following steps. 1) Schizo-or-control maps (SOC) were estimated by comparing the connectogram with WD. Steps with GFA values closer to schizophrenia than to control were noted as schizo-liked, otherwise as control-liked. 2) Masked SOC (mSOC) was derived by applying a mask to SOC. Steps passing the criteria of the mask were reserved for mSOC. 3) SLI was defined as the number of steps which were schizo-liked in the mSOC. The performance of prediction was evaluated with receiver operating characteristic (ROC) curve analysis by comparing the SLI scores and clinical diagnostic results. Masks with different criteria of ES and CS were applied to evaluate what kind of difference between schizophrenia patients and healthy participants over the whole brain has the best capability in distinguishing the two groups.Figure 1 shows the maps of averaged area under ROC curves (AUC) of different masks from 200 permutations of males. The highest AUC of 0.87 was found for predicting males with the mask of ES ≥ 0.75 and CS ≥ 1 step (Black grid). Figure 2 shows the maps for female. The highest predicting performance (AUC = 0.77) was found with the mask of ES > 0.1 and CS > 1 step (Black grid). Figure 3 shows the accumulated masks of the masks with highest AUC from 200 permutations for males (3a) and females (3b).In this study, we examined the performance of predicting patients with schizophrenia based on the patterns of altered tract integrity over the whole brain. The whole-brain tract information was compared with predefined differences between schizophrenia patients and healthy participants to calculate an index of SLI indicating the similarity to schizophrenia. Our results showed that the prediction performance was high (AUC = 0.86 for males; AUC = 0.77 for females) when we compared the white matter integrities at specific segments on fiber pathways. The segments for predicting male subjects are located in several specific tract bundles with high ES as figure 3a showed while the contrast of the colors may imply the importance level for prediction. The fiber pathways for predicting female subjects are widespread over whole brain (figure 3b) which implies that there is higher variability of the tract impairment in female patients than in male patients.The information of the whole-brain tracts estimated by TBAA method is potentially useful for predicting adult patients with schizophrenia. Our study warrants a prospective study to validate the diagnostic accuracy of the method.