Connecting Patients with Pre-diagnosis: A Multiple Graph Regularized Method for Mental Disorder Diagnosis

Tianqi Zhao, Ming Kong, Kun Kuang, Zhengxing Huang, Qiang Zhu, Fei Wu

Abstract: Computer-aided diagnosis (CAD) plays an important role in medicine. But most of the previous methods only focus on the diagnosis process information like the image data for medical patterns learning, ignoring the pre-diagnosis, which is necessary and important for a doctor’s decision. Besides, traditional CAD methods treat the patients as independent samples in data. To make up this gap, in this paper, we propose to connect patients with pre-diagnosis and propose a novel Multiple Graph REgularized Diagnosis (MuGRED) method for mental disorder diagnosis, which contains two main components: multi-modal representation learning and a multiple graph feature fusion module. We validated our MuGRED method on a practical dataset of children’s attention deficit and hyperactivity disorder and a well-recognized ASD benchmark. Extensive experiments demonstrate that our MuGRED method can achieve a better performance than the state-of-the-art methods for mental disorder diagnosis.