Computer-aided prognosis: Accurate prediction of patients with neurologic and psychiatric diseases via multi-modal MRI analysis

Abstract

Multi-modal magnetic resonance imaging (MRI) is increasingly used in neuroscience research, as it allowed the non-invasive investigation of structure and function of the human brain in health and pathology. One of the most important applications of multi-modal MRI is the provision of vital diagnostic data for neurologic and psychiatric disorders. As traditional MRI researches using univariate analyses can only reveal disease-related structural and functional alterations at group level which limited the clinical application, and recent attention has turned toward integrating multi-modal neuroimaging and computer-aided prognosis (CAD) technology, especially machine learning, to assist clinical disease diagnose. Research in this area is growing exponentially, and therefore it is meaningful to review the current and future development of this emerging area. Hence, in this paper, based on our own studies and contributions, we review the recent advances in multi-modal MRI and CAD technologies, and their applications to assist the clinical diagnosis of three common neurologic and psychiatric disorders, namely, Alzheimer’s disease, Attention deficit/hyperactivity disorder and Tourette syndrome. We extracted multi-modal features from structural, diffusion and resting-state functional MRI, then different feature selection methods and classifiers were applied. In addition, we applied different feature fusion schemes (e.g. multiple kernel learning) to combining multi-modal features for classification. Our experiments show that using feature fusion techniques to integrate multi-modal features can yield better classification results for diseases prediction, which may outline some future directions for multi-modal neuroimaging where researchers can design more advanced methods and models for neurologic and psychiatric research.