More about HKUST
Applications of Graph Convolutional Networks on Medical Image Analysis: A Survey
PhD Qualifying Examination Title: "Applications of Graph Convolutional Networks on Medical Image Analysis: A Survey" by Mr. Yishuo ZHANG Abstract: Deep learning methods have revolutionised many machine learning tasks such as computer vision and medical data analysis. Benefiting from stacked convolution operations and well-designed architecture, neural networks are capable of learning rich representations and show promising performances on many complicated tasks. The conventional convolution cannot be applied directly on graph data which exist universally. Recently, Graph Convolutional Networks (GCNs) extend convolution operation from grid space to non-grid space such as an irregular graph. These techniques make deep learning methods more flexible and effective when dealing with various kinds of data, especially for graph data. In this survey, we give a comprehensive introduction about the theorem of graph convolution which is the fundamental basis of GCNs. With a focus on GCNs applied in medical image tasks, related papers which are published on top conferences in medical image analysis as well as computer vision areas are reviewed and concluded. We introduce different applications of GCNs to show how to capture and utilize complex relationships from graph data. Finally, we propose some potential research directions in this promising field. Date: Monday, 30 September 2019 Time: 3:30pm - 5:30pm Venue: Room 2408 Lifts 17/18 Committee Members: Prof. Albert Chung (Supervisor) Prof. Nevin Zhang (Chairperson) Dr. Qifeng Chen Prof. Long Quan **** ALL are Welcome ****