USING AI APPROACHES FOR PREDICTING BRAIN AGEING CHANGES IN SCHIZOPHRENIA

Abstract Background Schizophrenia is a severe mental disorder that causes structural and functional abnormalities in the brain. It affects approximately 1% of the global population and has the highest prevalence and most complex symptoms among chronic mental illnesses. Numerous studies on schizophrenia patients have shown a trend of reduced total intracranial volume (TIV) and regional volumes (volumes of different tissue categories), particularly a decrease in gray matter (GM). It is still unclear whether the decrease in brain volume is associated with age-related changes seen in normal individuals or represents processes specific to individuals, such as those occurring at the onset of the disease, genetic factors, diet, and lifestyle. Aims & Objectives Previous studies have held different views on whether schizophrenia is associated with more severe age-related cognitive decline. Until recently, longitudinal studies examining cognitive decline have been lacking. In recent years, brain age prediction has emerged as a neurobiological marker of brain degeneration. Method The brain age gap (BrainAGE) model assesses the difference between the biological age of the brain and the actual age, with higher values indicating accelerated brain aging. This study utilizes neuroimaging data to evaluate BrainAGE and explore the rate of brain aging in individuals across different age groups. A total of 767 structural MRI data were collected from four public MRI datasets (COBRE, MCICShare, UCLA, NUSDAST) consisting of both schizophrenia patients and healthy controls. We constructed the BrainAGE deep learning model using data from the healthy control group. Results This model was trained based on brain images from healthy individuals, and the predicted outcome is represented as an age number, referred to as Brain Age Prediction. It can be applied to predict the brain images of schizophrenia patients. Finally, we used data from schizophrenia patients to provide insights into the brain age aging assessment for individuals with schizophrenia. Discussion & Conclusion In conclusion, this study is using deep learning to establish a predictive health brain age model and also apply it to evaluate the brain age of participants with schizophrenia patients. Moreover, several aspects of this study are worth further research and development. To begin with, the study collected brain imaging data from public databases in schizophrenia research and also used brain imaging data from healthy individuals for training. It is lead to the health control brain imaging data in limited. The problem is able to collect brain imaging data from healthy participants in other research and increase the datasets in our study. Additionally, transfer learning will be used. Transfer learning can use similar data for pre-training, assisting in training models with limited training data. Consequently, the study focuses on exploring the differences in volume and thickness of different brain regions between healthy individuals and schizophrenia patients in T1-weighted images. On the other hand, functional MRI and diffusion tensor imaging (DTI) have also been applied in artificial intelligence in recent years. Therefore, this study will attempt to establish a brain age prediction model using a multimodal in the future.