A Case-Control Study On Corpus Callosum Volume and Clinical Correlates in Treatment-Resistant and Responsive Schizophrenia Patients: Addressing MRI Analysis within Single-Subject and a Novel Artificial Intelligence Paradigm.

INTRODUCTION

The corpus callosum (CC) is critical for inter-hemispheric integration, and its structural abnormalities have been implicated in the neurobiology of schizophrenia (SCZ). Ap-proximately 20-35% of SCZ patients exhibit resistance to antipsychotics and are classified as treat-ment-resistant schizophrenia (TRS). To date, no studies have specifically analyzed CC morphology in TRS compared to healthy controls (HCs) and treatment-responsive SCZ patients (non-TRS).

METHODS

The study included 75 HCs and 97 SCZ patients (39 TRS and 58 non-TRS, including cannabis users and non-users). Participants underwent cross-sectional assessment using the Neuro-logical Evaluation Scale (NES), the Brief Assessment of Cognition in Schizophrenia, and the Posi-tive and Negative Syndrome Scale (PANSS). CC volume was measured via MRI and analyzed with a novel artificial intelligence program (MRI-AI) for single-subject assessment, with findings con-firmed through group-level voxel-based morphometry (VBM).

RESULTS

TRS patients exhibited significantly higher total PANSS (p < 0.001) and NES (p < 0.001) scores and performed worse on working memory tasks compared to non-TRS patients. MRI-AI identified CC volume reduction in 55.88% of TRS patients (TRS CC: 3.02 mL vs. non-TRS with cannabis use: 3.69 mL; without cannabis use: 3.28 mL; p = 0.002). VBM confirmed significant CC volume reduction in TRS patients relative to HCs (p = 0.004). Cannabis use in non-TRS patients did not significantly affect CC volume or cognitive performance.

CONCLUSIONS

TRS is associated with pronounced neurological, cognitive, and clinical dysfunctions. Aberrant CC volume may contribute to TRS pathogenesis and executive function deficits. MRI-AI provides a robust, single-subject approach for characterizing CC abnormalities in schizophrenia, of-fering potential for personalized clinical assessment.