内視鏡下鼻副鼻腔手術訓練の定量評価にむけた深層学習による鉗子運動解析

抄録

Endoscopic sinus surgery （ESS） is now widely accepted in the field of otolaryngology. However, for doctors, routine and repetitive training is indispensable for safe surgery. Therefore, it is necessary to develop an ESS training system having quantitative and objective evaluation that is easy to use every day. In order to evaluate ESS training, we developed a sensorless forceps motion analysis method. In this study, we adopted the deep learning method for extracting the forceps region from endoscopic videos. As a result of analyzing the behavioral features, significant differences were observed between experts and novices in the several proposed indexes, and the effectiveness of the proposed method was shown.