6th Edition of Neurology World Conference 2026

Abstract

Psychotic spectrum disorders lack objective diagnostic biomarkers and rely entirely on clinical symptom observation. Although advanced diffusion MRI modeling can quantify brain microstructural abnormalities, conventional single-compartment approaches cannot distinguish overlapping pathological processes, including neuroinflammation, perfusion abnormalities, and alterations in tissue complexity. An integrated diffusion MRI framework that combines Intravoxel Incoherent Motion (IVIM), Free Water Imaging (FWI), and Diffusion Kurtosis Imaging (DKI) to model perfusion, extracellular inflammation, and tissue complexity simultaneously was developed in this study. Diffusion-weighted imaging was acquired from 1,267 participants (1,094 controls; 173 patients with psychotic spectrum disorders, including 83 with schizophrenia, 20 with schizoaffective disorder, and 70 with non-schizophrenia psychosis diagnoses) using multi-shell protocols. Regional diffusion metrics from 68 cortical parcellations served as input features for neural network classification models. This multi-compartment framework was designed specifically as a case-control classification framework to act as an objective clinical decision-support tool for patient stratification. Patient versus control classification achieved exceptional performance (AUC-ROC = 0.968 ± 0.021, Average Precision = 0.791), with mean kurtosis and mean diffusivity in the pars triangularis of both hemispheres emerging as the most discriminatory features. Classification of schizophrenia spectrum versus non-schizophrenia psychosis showed modest discrimination (AUC-ROC = 0.716 ± 0.103, Average Precision = 0.679), with paracentral fractional anisotropy demonstrating the highest feature importance. Integrated diffusion MRI modeling combined with machine learning classification reliably distinguished those with psychotic spectrum disorders from those who do not, with discriminatory features localized to language processing and executive control regions. The substantial neurobiological overlap between diagnostic subtypes supports the idea that psychosis is a continuous spectrum and demonstrates the potential of multi-compartment diffusion modeling as an objective classification framework for psychotic disorders.