The Imaging And Genetic Study Of Schizophrenia Patients With Violent Criminal Behavior

Objectives:1.To investigate the clinical characteristics,the characteristics of prefrontal cortex functional connectivity in the resting state prefrontal cortex region and the genetic risk of schizophrenia patients with violent criminal behavior.2.To determine the correlation between indicators of prefrontal cortex functional connectivity in the resting state and factors associated with violence risk and polygenic risk score of schizophrenia patients with violent criminal behavior.3.To construct a violence classification model for patients with schizophrenia using machine learning methods,based on the data of clinical characteristics,prefrontal cortex functional connectivity and polygenic risk score.Methods:1.A case-control study was conducted consist of three groups of subjects,I)Violent group:violent schizophrenia patients（involved in murder and intentional injury crimes,identified as incapable of criminal responsibility）,II)Non-violent group:the schizophrenia patients who had never attacked others before,and III)Healthy Control group:consists of healthy people.Finally,a total of 722 subjects were included,including370 in the violent group,216 in the non-violent group,and 136 in the healthy control group.The social demographics,historical environmental data and clinically relevant data of the subjects were collected.Moreover,the PANSS and other symptom scales,PCL:SV,HCR-20,and RBANS were used to evaluate the psychiatric symptoms,psychopathy,violence risk,and neurocognitive function of the patient groups.Differences in these factors were compared among violent,non-violent patients and（or）the healthy control groups.2.Among the enrolled patients,resting-state near-infrared functional imaging data（f NIRS）complete data were collected for 401 subjects whereas,373 subjects’data were included in the analysis after data quality control.Based on the functional connectivity analysis method of regions of interest（ROI）,the difference in the average functional connectivity strength（FC）among the ROIs among the three groups of subjects was compared using the method of covariance analysis.Spearman’s correlation analysis was used to identify the correlation between FC and violence risk-related indicators with differences between the violent and non-violent groups.3.Among the enrolled patients,a total of 585 blood samples were collected followed by DNA extraction.After data quality control screening,ultimately,a total of 565 samples were included in the analysis.The Polygenic Risk Score（PRS）was calculated based on the GWAS results from the public database of schizophrenia in the East Asian population released by PGC and the SNP typing results of the samples in our study.The spearman correlation was used to analyze the correlation between the polygenic risk score for schizophrenia（SCZ-PRS）and violence risk-related indicators,as well as the prefrontal functional connectivity indicators that were different between the violent and non-violent groups.4.Based on the multidimensional factors,including clinical characteristics,prefrontal cortex functional connectivity and SCZ-PRS,three machine learning methods of logistic regression,random forest and naive Bayes were used to establish violence classification models of schizophrenia patients.Results:1.Violent patients with schizophrenia showed higher proportion of males（χ²=18.500,P<0.001）,lower education（χ²=108.832,P<0.001）,the rural registration（χ²=263.712,P<0.001）and substance use disorders（P<0.001）than the non-violent group and the healthy control group.The violent group also had a higher proportion of poor family environment in early years of life（χ²=21.618,P<0.001）,previous criminal history（χ²=11.708,P=0.001）and previous destruction history（χ²=40.437,P<0.001）than the non-violent group.2.Compared with the non-violent group,the age of onset in the violent group was older（t=-4.607,P<0.001）,and the untreated period of mental illness（Z=-3.204,P=0.001）and the hospitalization time was longer（t=-5.883,P<0.001）,and the proportion of previous psychiatric visits was lower（χ²=35.312,P<0.001）.The total score and each factor score of HCR-20（P<0.001）and PCL:SV（P<0.001）in the violent group were comparatively higher than those in the non-violent group.3.In terms of neurocognitive function,the total score and each subscale of the RBANS test in the violent group and the non-violent group were lower than those in the healthy control group（P<0.001）.The total score of whole test,immediate memory,visual span and delayed memory of the violent group were higher than the non-violent group（P<0.001）,but lower than the non-violent group in attention（P=0.007）.4.In terms of brain functional connectivity characteristics,compared with the non-violent group,the FC between the frontal pole and right pars triangularis Broca’s area（PFC-TRI.R）（P=0.041）,the right dorsolateral prefrontal cortexand the right pars triangularis Broca’s area（DLPFC.R-TRI.R）（P=0.028）,the right pars triangularis Broca’s area and the right Pre-Motor and Supplementary Motor Cortex（TRI.R-PMSMC.R）（P=0.016）and the right pars triangularis Broca’s area and the left Pre-Motor and Supplementary Motor Cortex（TRI.R-PMSMC.L）（P=0.048）were lower in the violent group,but the FC between the bilateral Pre-Motor and Supplementary Motor Cortex（P<0.001）was higher.In addition,the results of correlation analysis showed that in the violent group,FC_{PMSMC.R-PMSMC.L}were positively correlated with the antisocial factors of psychopathic（PCL:SV）（r=0.172,P=0.048）;in all patients,we found that the FC_FPC-TRI.Rwas negatively correlated with the total score of psychiatric disorders（PCL:SV）;the FC_{PMSMC.R-PMSMC.L}were positively correlated with the H-factor of HCR-20（r=0.119,P=0.049）,the total score（r=0.169,P=0.005）,lifestyle factor（r=0.131,P=0.030）and antisocial factor（r=0.233,P<0.001）of PCL:SV.5.The results of SCZ-PRS correlation analysis showed that in violent patients,SCZ-PRS was positively correlated with FC_{PMSMC.R-PMSMC.L}（r=0.196-0.282,P=0.002-0.029）;In all sample,it was found that SCZ-PRS was negatively correlated with FC_FPC-TRI.Rand FC_{DLPFC.R-TRI.R}（r=-0.129,P=0.037;r=-0.156,P=0.011）.However,no significant correlation was found between SCZ-PRS and violence-related clinical indicators,whether in the violence group or in all samples.6.Among the machine learning models constructed based on clinical characteristics,resting-state prefrontal cortex functional connectivity,and PRS multidimensional data,the random forest model（RF）had achieved the best performance to classify the violent group and the non-violent group.It has an accuracy of 0.942,a sensitivity of 0.985,and a specificity of 0.895.Conclusions:1.Schizophrenia patients with violent criminal behavior have the characteristics of higher proportion of male,low education level,long untreated period,older onset age,and higher psychopathic score.2.The abnormal pivotal function of the right triangle in the prefrontal lobe,which is responsible for language management,and the abnormal activation of the connections between the premotor and auxiliary motor cortex,which are responsible for complex motor control and cognitive control,may be related to violence in schizophrenia.3.The polygenic risk score for schizophrenia may affect the individual’s impulse control ability by affecting the functional connectivity of the prefrontal cortex emotional control and cognitive decision-making related brain regions.4.The classification model constructed based on clinical and imaging genetic multidimensional factors can be used to identify schizophrenia patients with violent criminal behavior.