Validation of Knot-Tying Motion by Temporal-Spatial Matching with RGB-D Sensor for Surgical Training

We propose a method for validating surgical knot-tying motions for self-training of surgical trainee. Our method observes their hands by a RGB-D sensor and describes the point cloud as a SHOT feature. We use the features for matching an input point cloud sequence to that of an expert by dynamic programming. On the basis of the matched frames, the method validates relative positions of both hands and each hand shape in each input frame. Then the system detects and shows inappropriate parts in each frame. This paper shows the results of our method on a knot-tying motion dataset of a novice and an expert surgeon.