Kyunghwa Jung1, Hyunseok Choi1, Hanpyo Hong2, Arnold Adikrishna2, In-Ho Jeon3, Jaesung Hong4. 1. Department of Robotics Engineering, Daegu Gyeongbuk Institute of Science and Technology (DGIST), 333, Techno-jungang-daero, Hyeonpung-myeon, Dalseong-gun, Daegu, 711-873, Korea. 2. Department of Orthopedic Surgery, Asan Medical Center, School of Medicine, University of Ulsan, 86 Asanbyeongwon-gil, Songpa-gu, Seoul, 138-736, Korea. 3. Department of Orthopedic Surgery, Asan Medical Center, School of Medicine, University of Ulsan, 86 Asanbyeongwon-gil, Songpa-gu, Seoul, 138-736, Korea. jeonchoi@gmail.com. 4. Department of Robotics Engineering, Daegu Gyeongbuk Institute of Science and Technology (DGIST), 333, Techno-jungang-daero, Hyeonpung-myeon, Dalseong-gun, Daegu, 711-873, Korea. jhong@dgist.ac.kr.
Abstract
BACKGROUND: A hands-free region-of-interest (ROI) selection interface is proposed for solo surgery using a wide-angle endoscope. A wide-angle endoscope provides images with a larger field of view than a conventional endoscope. With an appropriate selection interface for a ROI, surgeons can also obtain a detailed local view as if they moved a conventional endoscope in a specific position and direction. METHODS: To manipulate the endoscope without releasing the surgical instrument in hand, a mini-camera is attached to the instrument, and the images taken by the attached camera are analyzed. When a surgeon moves the instrument, the instrument orientation is calculated by an image processing. Surgeons can select the ROI with this instrument movement after switching from 'task mode' to 'selection mode.' The accelerated KAZE algorithm is used to track the features of the camera images once the instrument is moved. Both the wide-angle and detailed local views are displayed simultaneously, and a surgeon can move the local view area by moving the mini-camera attached to the surgical instrument. RESULTS: Local view selection for a solo surgery was performed without releasing the instrument. The accuracy of camera pose estimation was not significantly different between camera resolutions, but it was significantly different between background camera images with different numbers of features (P < 0.01). The success rate of ROI selection diminished as the number of separated regions increased. However, separated regions up to 12 with a region size of 160 × 160 pixels were selected with no failure. Surgical tasks on a phantom model and a cadaver were attempted to verify the feasibility in a clinical environment. CONCLUSIONS: Hands-free endoscope manipulation without releasing the instruments in hand was achieved. The proposed method requires only a small, low-cost camera and an image processing. The technique enables surgeons to perform solo surgeries without a camera assistant.
BACKGROUND: A hands-free region-of-interest (ROI) selection interface is proposed for solo surgery using a wide-angle endoscope. A wide-angle endoscope provides images with a larger field of view than a conventional endoscope. With an appropriate selection interface for a ROI, surgeons can also obtain a detailed local view as if they moved a conventional endoscope in a specific position and direction. METHODS: To manipulate the endoscope without releasing the surgical instrument in hand, a mini-camera is attached to the instrument, and the images taken by the attached camera are analyzed. When a surgeon moves the instrument, the instrument orientation is calculated by an image processing. Surgeons can select the ROI with this instrument movement after switching from 'task mode' to 'selection mode.' The accelerated KAZE algorithm is used to track the features of the camera images once the instrument is moved. Both the wide-angle and detailed local views are displayed simultaneously, and a surgeon can move the local view area by moving the mini-camera attached to the surgical instrument. RESULTS: Local view selection for a solo surgery was performed without releasing the instrument. The accuracy of camera pose estimation was not significantly different between camera resolutions, but it was significantly different between background camera images with different numbers of features (P < 0.01). The success rate of ROI selection diminished as the number of separated regions increased. However, separated regions up to 12 with a region size of 160 × 160 pixels were selected with no failure. Surgical tasks on a phantom model and a cadaver were attempted to verify the feasibility in a clinical environment. CONCLUSIONS: Hands-free endoscope manipulation without releasing the instruments in hand was achieved. The proposed method requires only a small, low-cost camera and an image processing. The technique enables surgeons to perform solo surgeries without a camera assistant.
Keywords:
Image processing; Region of interest (ROI); Solo surgery; Wide-angle endoscope