Integrated Equipment for Parkinson’s Disease Early Detection Using Graph Convolution Network

There is an increasing need to diagnose Parkinson’s disease (PD) in early stage. Existing solutions mainly focused on traditional ways such as MRI, thus suffering from the ease-of-use issue. This work presents a new approach using video and skeleton-based techniques solve this problem. In paper, end-to-end diagnosis method based graph convolution networks proposed, which takes patients’ skeletons sequence input returns result. The asymmetric dual-branch network architecture designed process global local information separately capture subtle manifestation of PD. To train network, we present first gait dataset, PD-Walk. dataset consists 95 PD patients 96 healthy people’s walking videos. All data are annotated by experienced doctors. Furthermore, implement our portable equipment, has been operation First Affiliated Hospital, Zhejiang University School Medicine. Experiments show that can achieve 84.1% accuracy real-time performance equipment real environment. Compared with solutions, proposed detect suspicious symptoms quickly conveniently. Integrated be easily placed hospitals or nursing homes provide services for elderly people.