鉗子シャフトの画像認識に基づく手術助手ロボットの操作インタフェース

抄録

With the collaboration of locally operated small surgical robot in a sterile area, a surgeon can perform robotically assisted laparoscopic solo surgery. A new operation interface based on image recognition of forceps performed by the surgeon to control a surgical assistant manipulator has been developed. The forceps shaft captured using RGB and depth images is extracted by Mask R-CNN, and the orientation of the shaft is estimated for the control command. The prediction on test images of actual operation for the Mask R-CNN trained model on simulated operation images was evaluated.