Share：
Share this content in WeChat
X
[Chinese] [PDF] 7 1
• Clinical Article •
Value of interpretable machine learning models based on sMRI in predicting aggressive and violent behaviors in schizophrenia patients
LIU Shihan  KAN Jianfei  WANG Meijie  ZHENG Changyong  CHEN Fujian 

DOI:10.12015/issn.1674-8034.2026.01.005.


[Abstract] Objective Through voxel-based morphometry (VBM), the structural magnetic resonance imaging (sMRI) features of schizophrenia (SCZ) patients with and without aggressive violence were compared, and a machine learning model was constructed to realize the early identification and prediction of SCZ patients with aggressive violence.Materials and Methods A retrospective analysis of 146 patients diagnosed with SCZ in Shandong Daizhuang Hospital from March 2023 to June 2025, including 77 SCZ patients with aggressive violence and 69 SCZ patients without aggressive violence. The differences of clinical indicators and sMRI features between the aggressive violence group and the non-aggressive violence group were compared, and the prediction model of aggressive violence in SCZ patients was constructed. Among the four machine learning prediction models, the area under the curve (AUC) of logistic regression (LR), decision tree (DT), random forest (RF) and support vector machine (SVM) were 0.824, 0.821, 0.917 and 0.940, respectively. The results of DeLong test showed that the AUC of LR, DT, RF and SVM were 0.824, 0.821, 0.917 and 0.940, respectively. The predictive performance of the SVM model was the highest (P < 0.05). The SHAP (SHapley Additive exPlanations) summary map results showed that the thickness of the left insular cortex was the most important feature for predicting aggressive violence in schizophrenia. Decision curve analysis (DCA) showed that the four models had high guiding significance for clinical practice.Results There was no significant difference in clinical indicators between the aggressive violence group and the non-aggressive violence group (P > 0.05). The thickness of the left insular cortex, the thickness of the left fusiform cortex, the Gaussian curvature of the left operculum, the thickness of the right insular cortex, the mean curvature of the left operculum, and the thickness of the right middle and posterior prefrontal cortex in the aggressive violence group were lower than those in the non-aggressive violence group (P < 0.05).Conclusions Machine learning models based on sMRI can predict aggressive and violent behaviors in SCZ patients, with the SVM model exhibiting the highest predictive performance.
[Keywords] machine learning；structural magnetic resonance imaging；magnetic resonance imaging；schizophrenia；aggression and violence；predictive model

LIU Shihan1   KAN Jianfei2   WANG Meijie3   ZHENG Changyong1   CHEN Fujian1*  

1 Department of Radiology, Shandong Daizhuang Hospital, Jining 272051, China

2 Department of Pharmacy, Shandong Daizhuang Hospital, Jining 272051, China

3 Sixth Department of Psychiatry, Shandong Daizhuang Hospital, Jining 272051, China

Corresponding author: CHEN F J, E-mail: dzyyyxk2014@163.com

Conflicts of interest   None.

Received  2025-08-16
Accepted  2025-12-10
DOI: 10.12015/issn.1674-8034.2026.01.005
DOI:10.12015/issn.1674-8034.2026.01.005.

[1]
ACUÑA V, GUERRA M, COBAISSE M, et al. Women with Schizophrenia: Beyond Psychosis[J]. Community Ment Health J, 2025, 61(1): 22-28. DOI: 10.1007/s10597-024-01321-3.
[2]
LV H, GUO M, GUO C, et al. The Interrelationships between Cytokines and Schizophrenia: A Systematic Review[J/OL]. Int J Mol Sci, 2024, 25(15): 8477 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11313682. DOI: 10.3390/ijms25158477.
[3]
MESSINIS A, PANTELI E, PARASKEVOPOULOU A, et al. Altered lipidomics biosignatures in schizophrenia: A systematic review[J]. Schizophr Res, 2024, 271: 380-390. DOI: 10.1016/j.schres.2024.06.043.
[4]
YU T, PEI W, XU C, et al. Investigation of peripheral inflammatory biomarkers in association with violence in schizophrenia[J/OL]. BMC Psychiatry, 2024, 24(1): 542 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11293062. DOI: 10.1186/s12888-024-05966-y.
[5]
NADER G, QURESHI M, CHAUDHARY Z, et al. Effect of personality traits, psychological resilience and life adversities on lifetime violence trajectories: A novel classification model in schizophrenia[J/OL]. Psychiatry Res, 2024, 339: 116044 [2025-08-16]. https://doi.org/10.1016/j.psychres.2024.116044. DOI: 10.1016/j.psychres.2024.116044.
[6]
YU T, PEI W, ZHANG X, et al. Associations Between Thyroid Hormones Levels and Gray Matter Volume of Frontal Lobe Involved into Violence in Male Schizophrenia Patients[J]. Neuropsychiatr Dis Treat, 2024, 20: 2169-2175. DOI: 10.2147/NDT.S481875.
[7]
HE W M, ZHANG X Y, XIE W G, et al. Expression level of myocardial enzymes in patients with schizophrenia: Predictive value in the occurrence of violence[J]. World J Psychiatry, 2024, 14(9): 1346-1353. DOI: 10.5498/wjp.v14.i9.1346.
[8]
LU J, GOU N, ZHANG S, et al. Dysfunctional striatal-temporal connectivity is associated with a higher risk for violence in patients with schizophrenia[J/OL]. Psychiatry Res, 2025, 348: 116488 [2025-08-16]. https://doi.org/10.1016/j.psychres.2025.116488. DOI: 10.1016/j.psychres.2025.116488.
[9]
HAUKVIK U K, WOLFERS T, TESLI N, et al. Individual-level deviations from normative brain morphology in violence, psychosis, and psychopathy[J/OL]. Transl Psychiatry, 2025, 15(1): 118 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11965457. DOI: 10.1038/s41398-025-03343-1.
[10]
ZOU K, TAN S W, ZHU S X, et al. Relationships between sMRI, fMRI, and dMRI Features and Cerebellar Function in Patients with Sensorineural Hearing Loss[J]. Chin J Clin Med Imag, 2025, 36(6): 385-388, 392. DOI: 10.12117/jccmi.2025.06.002.
[11]
WANG L, XIONG X, LIU J, et al. Gray matter structural and functional brain abnormalities in Parkinson's disease: a meta-analysis of VBM and ALFF data[J/OL]. J Neurol, 2025, 272(4): 276 [2025-08-16]. https://doi.org/10.1007/s00415-025-12934-3. DOI: 10.1007/s00415-025-12934-3.
[12]
VACCA S, SURI J S, SABA L. SBM vs VBM for highlighting similarities and differences between chronotype and Parkinson's MRI scans: a preliminary analysis[J]. Int J Neurosci, 2025, 135(2): 203-212. DOI: 10.1080/00207454.2023.2292958.
[13]
GUAN C, GONG A, ZHAO Y, et al. Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study[J/OL]. Crit Care, 2024, 28(1): 349 [2025-08-16]. https://doi.org/10.1186/s13054-024-05138-0. DOI: 10.1186/s13054-024-05138-0.
[14]
XU J, CHEN T, FANG X, et al. Prediction model of pressure injury occurrence in diabetic patients during ICU hospitalization--XGBoost machine learning model can be interpreted based on SHAP[J/OL]. Intensive Crit Care Nurs, 2024, 83: 103715 [2025-08-16]. https://doi.org/10.1016/j.iccn.2024.103715. DOI: 10.1016/j.iccn.2024.103715.
[15]
PAREEK A, RO D H, KARLSSON J, et al. Machine learning/artificial intelligence in sports medicine: state of the art and future directions[J]. J ISAKOS, 2024, 9(4): 635-644. DOI: 10.1016/j.jisako.2024.01.013.
[16]
JUAN Q, SHIWAN T, YURONG S, et al. Brain structural and functional abnormalities in affective network are associated with anxious depression[J/OL]. BMC Psychiatry, 2024, 24(1): 533 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11270941. DOI: 10.1186/s12888-024-05970-2.
[17]
PANG J, XU J, CHEN L, et al. Family history, inflammation, and cerebellum in major depression: a combined VBM and dynamic functional connectivity study[J/OL]. Transl Psychiatry, 2025, 15(1): 188 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC12137595. DOI: 10.1038/s41398-025-03409-0.
[18]
CHEN M, ZHANG G, WANG W J, et al. Application of Synthetic MRI Combined with VBM Brain Partitioning in the Diagnosis of Early Parkinson's Disease[J]. Chin J Magn Reson Imag, 2023, 14(10): 20-25. DOI: 10.12015/issn.1674-8034.2023.10.004.
[19]
GRIEVE S M, KORGAONKAR M S, KOSLOW S H, et al. Widespread reductions in gray matter volume in depression[J]. Neuroimage Clin, 2013, 3: 332-339. DOI: 10.1016/j.nicl.2013.08.016.
[20]
LI X L, CAI L N, CUI X, et al. Research Progress of Magnetic Resonance Imaging in Amnestic Mild Cognitive Impairment[J]. Chin J Magn Reson Imag, 2021, 12(11): 94-96. DOI: 10.12015/issn.1674-8034.2021.11.023.
[21]
ZHANG R, DENG H, XIAO X. The Insular Cortex: An Interface Between Sensation, Emotion and Cognition[J]. Neurosci Bull, 2024, 40(11): 1763-1773. DOI: 10.1007/s12264-024-01211-4.
[22]
KUROKI N, KASHIWAGI H, OTA M, et al. Brain structure differences among male schizophrenic patients with history of serious violent acts: an MRI voxel-based morphometric study[J/OL]. BMC Psychiatry, 2017, 17(1): 105 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC5361832. DOI: 10.1186/s12888-017-1263-9.
[23]
IEZZI D, CÁCERES-RODRÍGUEZ A, PEREIRA-SILVA J, et al. Gestational CBD Shapes Insular Cortex in Adulthood[J/OL]. Cells, 2024, 13(17): 1486 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11394289. DOI: 10.3390/cells13171486.
[24]
TIKÀSZ A, POTVIN S, DUGRÉ J R, et al. Violent Behavior Is Associated With Emotion Salience Network Dysconnectivity in Schizophrenia[J/OL]. Front Psychiatry, 2020, 11: 143 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC7059347. DOI: 10.3389/fpsyt.2020.00143.
[25]
FORTIER A, DUMAIS A, ATHANASSIOU M, et al. Dysconnectivity between the anterior insula and the dorsal anterior cingulate cortex during an emotion go/nogo paradigm is associated with aggressive behaviors in male schizophrenia patients[J/OL]. Psychiatry Res Neuroimaging, 2023, 328: 111579 [2025-08-16]. https://doi.org/10.1016/j.pscychresns.2022.111579. DOI: 10.1016/j.pscychresns.2022.111579.
[26]
PLAZA P L, RENIER L, ROSEMANN S, et al. Sound-encoded faces activate the left fusiform face area in the early blind[J/OL]. PLoS One, 2023, 18(11): e0286512 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10664868. DOI: 10.1371/journal.pone.0286512.
[27]
WU D, ZHANG P, LIU N, et al. Effects of High-Definition Transcranial Direct Current Stimulation Over the Left Fusiform Face Area on Face View Discrimination Depend on the Individual Baseline Performance[J/OL]. Front Neurosci, 2021, 15: 704880 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC8639859. DOI: 10.3389/fnins.2021.704880.
[28]
WELLER M. Anterior opercular cortex lesions cause dissociated lower cranial nerve palsies and anarthria but no aphasia: Foix-Chavany-Marie syndrome and "automatic voluntary dissociation" revisited[J]. J Neurol, 1993, 240(4): 199-208. DOI: 10.1007/BF00818705.
[29]
DACOSTA-AGUAYO R, PUIG J, LAMONJA-VICENTE N, et al. Reduced Cortical Thickness Correlates of Cognitive Dysfunction in Post-COVID-19 Condition: Insights from a Long-Term Follow-up[J]. AJNR Am J Neuroradiol, 2024, 45(5): 647-654. DOI: 10.3174/ajnr.A8167.
[30]
SINKO L, REGIER P, CURTIN A, et al. Neural correlates of cognitive control in women with a history of sexual violence suggest altered prefrontal cortical activity during cognitive processing[J/OL]. Womens Health (Lond), 2022, 18: 17455057221081326 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC8883288. DOI: 10.1177/17455057221081326.
[31]
SANZ H, VALIM C, VEGAS E, et al. SVM-RFE: selection and visualization of the most relevant features through non-linear kernels[J/OL]. BMC Bioinformatics, 2018, 19(1): 432 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC6245920. DOI: 10.1186/s12859-018-2451-4.
[32]
CHEN D, LIU J, ZHANG L, et al. Integrated Machine Learning and Bioinformatic Analyses Constructed a Novel Stemness-Related Classifier to Predict Prognosis and Immunotherapy Responses for Hepatocellular Carcinoma Patients[J]. Int J Biol Sci, 2022, 18(1): 360-373. DOI: 10.7150/ijbs.66913.
[33]
ELKAHWAGY DMAS, KIRIACOS C J, MANSOUR M. Logistic regression and other statistical tools in diagnostic biomarker studies[J]. Clin Transl Oncol, 2024, 26(9): 2172-2180. DOI: 10.1007/s12094-024-03413-8.
[34]
LORENZ R C, JENNY M, JACOBS A, et al. Fast-and-frugal decision tree for the rapid critical appraisal of systematic reviews[J]. Res Synth Methods, 2024, 15(6): 1049-1059. DOI: 10.1002/jrsm.1754.
[35]
BUDHKAR A, SONG Q, SU J, et al. Demystifying the black box: A survey on explainable artificial intelligence (XAI) in bioinformatics[J]. Comput Struct Biotechnol J, 2025, 27: 346-359. DOI: 10.1016/j.csbj.2024.12.027.
[36]
FAA G, FRASCHINI M, BARBERINI L. Reproducibility and explainability in digital pathology: The need to make black-box artificial intelligence systems more transparent[J/OL]. J Public Health Res, 2024, 13(4): 22799036241284898 [2025-08-16]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11528586. DOI: 10.1177/22799036241284898.
[37]
LIU J, LI X, ZHU P. Effects of Various Heavy Metal Exposures on Insulin Resistance in Non-diabetic Populations: Interpretability Analysis from Machine Learning Modeling Perspective[J]. Biol Trace Elem Res, 2024, 202(12): 5438-5452. DOI: 10.1007/s12011-024-04126-3.
[38]
HASAN M, WU W, ZHAO X. SHAP-Driven Feature Analysis Approach for Epileptic Seizure Prediction[J/OL]. J Med Syst, 2025, 49(1): 77 [2025-08-16]. https://doi.org/10.1007/s10916-025-02211-1. DOI: 10.1007/s10916-025-02211-1.

PREV A study on the brain network connectivity of adolescent depression patients with childhood trauma treated with eye movement desensitization and reprocessing combined with sertraline based on rs-fMRI
NEXT Correlation between arterial recanalization and the susceptibility vessel sign following endovascular therapy in patients with acute ischemic stroke
  


LINK


Tel & Fax: +8610-67113815    E-mail: editor@cjmri.cn