We develop cutting-edge methods for healthcare applications, especially for problems in radiology, based on computer vision and deep learning. While fulfilling the clinical requirements of each problem, we pursue making technical advancements that can impact other problems/research/development. In terms of methodologies, we are interested in (but not limited to) anatomically or biologically inspired modeling, data-efficient learning, reinforcement learning, etc. Recent research focuses on analyzing biological network structures in the human body from an image and can be categorized into three major topics as below.