Daichi Kitaguchi1,2,3, Nobuyoshi Takeshita4,5, Hiroki Matsuzaki1, Hiro Hasegawa1,2, Ryoya Honda2, Koichi Teramura2, Tatsuya Oda3, Masaaki Ito6,7. 1. Surgical Device Innovation Office, National Cancer Center Hospital East, 6-5-1 Kashiwanoha, Kashiwa, Chiba, 277-8577, Japan. 2. Department of Colorectal Surgery, National Cancer Center Hospital East, 6-5-1 Kashiwanoha, Kashiwa, Chiba, 277-8577, Japan. 3. Department of Gastrointestinal and Hepato-Biliary-Pancreatic Surgery, Faculty of Medicine, University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki, 305-8575, Japan. 4. Surgical Device Innovation Office, National Cancer Center Hospital East, 6-5-1 Kashiwanoha, Kashiwa, Chiba, 277-8577, Japan. ntakeshi@east.ncc.go.jp. 5. Department of Colorectal Surgery, National Cancer Center Hospital East, 6-5-1 Kashiwanoha, Kashiwa, Chiba, 277-8577, Japan. ntakeshi@east.ncc.go.jp. 6. Surgical Device Innovation Office, National Cancer Center Hospital East, 6-5-1 Kashiwanoha, Kashiwa, Chiba, 277-8577, Japan. maito@east.ncc.go.jp. 7. Department of Colorectal Surgery, National Cancer Center Hospital East, 6-5-1 Kashiwanoha, Kashiwa, Chiba, 277-8577, Japan. maito@east.ncc.go.jp.
Abstract
BACKGROUND: Urethral injuries (UIs) are significant complications pertaining to transanal total mesorectal excision (TaTME). It is important for surgeons to identify the prostate during TaTME to prevent UI occurrence; intraoperative image navigation could be considered useful in this regard. This study aims at developing a deep learning model for real-time automatic prostate segmentation based on intraoperative video during TaTME. The proposed model's performance has been evaluated. METHODS: This was a single-institution retrospective feasibility study. Semantic segmentation of the prostate area was performed using a convolutional neural network (CNN)-based approach. DeepLab v3 plus was utilized as the CNN model for the semantic segmentation task. The Dice coefficient (DC), which is calculated based on the overlapping area between the ground truth and predicted area, was utilized as an evaluation metric for the proposed model. RESULTS: Five hundred prostate images were randomly extracted from 17 TaTME videos, and the prostate area was manually annotated on each image. Fivefold cross-validation tests were performed, and as observed, the average DC value equaled 0.71 ± 0.04, the maximum value being 0.77. Additionally, the model operated at 11 fps, which provides acceptable real-time performance. CONCLUSIONS: To the best of the authors' knowledge, this is the first effort toward realization of computer-assisted TaTME, and results obtained in this study suggest that the proposed deep learning model can be utilized for real-time automatic prostate segmentation. In future endeavors, the accuracy and performance of the proposed model will be improved to enable its use in practical applications, and its capability to reduce UI risks during TaTME will be verified.
BACKGROUND: Urethral injuries (UIs) are significant complications pertaining to transanal total mesorectal excision (TaTME). It is important for surgeons to identify the prostate during TaTME to prevent UI occurrence; intraoperative image navigation could be considered useful in this regard. This study aims at developing a deep learning model for real-time automatic prostate segmentation based on intraoperative video during TaTME. The proposed model's performance has been evaluated. METHODS: This was a single-institution retrospective feasibility study. Semantic segmentation of the prostate area was performed using a convolutional neural network (CNN)-based approach. DeepLab v3 plus was utilized as the CNN model for the semantic segmentation task. The Dice coefficient (DC), which is calculated based on the overlapping area between the ground truth and predicted area, was utilized as an evaluation metric for the proposed model. RESULTS: Five hundred prostate images were randomly extracted from 17 TaTME videos, and the prostate area was manually annotated on each image. Fivefold cross-validation tests were performed, and as observed, the average DC value equaled 0.71 ± 0.04, the maximum value being 0.77. Additionally, the model operated at 11 fps, which provides acceptable real-time performance. CONCLUSIONS: To the best of the authors' knowledge, this is the first effort toward realization of computer-assisted TaTME, and results obtained in this study suggest that the proposed deep learning model can be utilized for real-time automatic prostate segmentation. In future endeavors, the accuracy and performance of the proposed model will be improved to enable its use in practical applications, and its capability to reduce UI risks during TaTME will be verified.
Authors: Jurriaan B Tuynman; Stefan E van Oostendorp; Miranda Kusters; Colin Sietses; Roel Hompes; H Jaap Bonjer Journal: Br J Surg Date: 2020-09-17 Impact factor: 6.939
Authors: T W A Koedam; M Veltcamp Helbach; P M van de Ven; Ph M Kruyt; N T van Heek; H J Bonjer; J B Tuynman; C Sietses Journal: Tech Coloproctol Date: 2018-03-22 Impact factor: 3.781
Authors: Lawrence Lee; Justin Kelly; George J Nassif; Teresa C deBeche-Adams; Matthew R Albert; John R T Monson Journal: Surg Endosc Date: 2018-07-11 Impact factor: 4.584
Authors: M Veltcamp Helbach; S E van Oostendorp; T W A Koedam; J J Knol; H B A C Stockmann; S J Oosterling; R C L M Vuylsteke; E J R de Graaf; P G Doornebosch; R Hompes; H J Bonjer; C Sietses; J B Tuynman Journal: Surg Endosc Date: 2019-03-19 Impact factor: 4.584
Authors: R B den Boer; C de Jongh; W T E Huijbers; T J M Jaspers; J P W Pluim; R van Hillegersberg; M Van Eijnatten; J P Ruurda Journal: Surg Endosc Date: 2022-08-04 Impact factor: 3.453