Neuroinformatics and Neuroimaging-based schizophrenia modeling and decision support

Abstract

Purpose: Schizophrenia is a common psychiatric disease that affects 1% of the world population. Current standard diagnostic procedures are based on observations on symptoms, which causes low to moderate diagnosis agreement between psychiatrists. We propose a novel approach to build schizophrenia classification models by using objective and quantifiable criteria from neuroinformatics and neuroimaging. Study Subjects: 156 study subjects (92 schizophrenia patients and 64 healthy controls) are recruited by our collaborating hospitals and their neuroinformatic data (demographic data, clinical information, clinical scores, neurocognitive tests) and neuroimaging data (magnetic resonance images, diffusion tensor images) are collected. Results: Various Bayesian Network classification models are constructed to improve the classification accuracy from 70.2% to 89.3%. Major Contributions: Our models reveal the quantitative relationship between schizophrenia and various intermediate phenotypes and brain abnormalities. The decision support system developed using these models can provide additional objective evidence to clinicians and augment the current schizophrenia diagnostic procedures. The methodology we developed has the potential to be applied to other diseases with informatics and imaging data.