IDDF2020-ABS-0165 Accuracy and applicability of the artificial intelligence integrated software in Z-line segmentation. (18th November 2020)

• Record Type: Journal Article
• Authors: Dao, Hang; Le, Hung; Nguyen, Binh; Nguyen, Hung; Manh, Huy; Ha, Kien; Vu, Hai
• Is Part Of: Gut. Volume 69(2020)Supplement 2
• Page Start: A4
• Page End: A4
• Publication Date: 2020-11-18
• Subjects: Gastroenterology -- Periodicals; 616.33
• DOI: 10.1136/gutjnl-2020-IDDF.7
• Languages: English
• ISSNs: 0017-5749

Abstract

Abstract : Background: Artificial intelligence (AI) and its applications in developing software for assisting endoscopic education are novel research directions around the world in general and in Vietnam particularly. This study aims to assess the software accuracy in assisting Z-line segmentation by comparing with medical doctors' detection results and record doctors' satisfaction in scale, time-of-implementation in interactive mode, and integrated mode. Methods: This research was conducted from April 2019 to July 2020. For the development of the Z-line detection algorithm, a dataset of 533 high-definition endoscopic WLI (white-light) images in diverse forms of Z-line were collected. Preliminary assessment of the AI algorithm yielded a good IoU (Interception over Union) of 72%. The software was subsequently developed in 4 modes, including manual mode, interactive mode (using Superpixels-BPT), automatic mode (using AI algorithm), and integrated mode (the combination of BPT and U-Net). 30 endoscopic images were assigned to 2 groups of doctors (naive and experienced group) for the Z-line detection using the software in 4 modes. Assessment indicators including time-of-implementation, number of mouse clicks, satisfaction in scales, and IoU metric with expert's ground-truth were taken into account. Results: The IoU metric of interactive and integrated modes in the experimental dataset deviated from 87.3% to 88.5% with no statistical difference to the IoU value of manual mode, and mean IoU metrics from the results of 2 doctor groups were over 85%. The mean values of time-of-implementation in interactive mode and integrated mode were not statistically different from manual mode. The median number of mouse clicks each use in the interactive mode and the integrated mode were 24.5 and 15.5 times, respectively. The software received good feedbacks from the doctors, with the mean values of satisfaction scores of automatic mode, interactive mode and integrated mode were 7.19, 7.26, and 7.18, respectively. Conclusions: The development of the software for detecting endoscopic anatomy landmarks is a novel and feasible research direction in Vietnam. Further studies could focus on detecting some specific lesions.