IDDF2022-ABS-0065 The investigations of a novel artificial intelligence approach to bowel preparation in different age populations based on a multicenter randomized controlled study

• Published in: Gut Vol. 71; no. Suppl 2; pp. A8 - A9
• Main Authors: Xie, YouXian, Lu, YangBor, Chiang, TungYing, Weng, YuChieh, Hu, YanXing, Huang, YungNing
• Format: Journal Article
• Language: English
• Published: London BMJ Publishing Group Ltd and British Society of Gastroenterology 02.09.2022; BMJ Publishing Group LTD
• Subjects: Age; Artificial intelligence; Colon; Colonoscopy; Colorectal cancer; Gut – IDDF Young Investigator Award; Intestine; Laxatives; Neural networks; Smartphones
• ISSN: 0017-5749; 1468-3288
• DOI: 10.1136/gutjnl-2022-IDDF.9

BackgroundAdequate bowel preparation is key to a successful colonoscopy. Some studies have proposed different methods to improve bowel preparation, such as text messages, videos, and even artificial intelligence recently. However, there is lacking study to investigate the age effect on these modern enforced educations for bowel preparation. In the study, we try to figure out the effects of AI applications via smartphones in different age populations.MethodsThe AI platform used a convolutional neural network and was able to assign either a ‘pass’ or ‘not pass’ designation to an uploaded image. We designed a multicentre, colonoscopist-blinded, randomized study. After taking the laxatives, all subjects were asked to use their smartphones to scan a quick response code for randomization. The system displayed instructions for using the application, taking photos of their feces, and uploading images to the system. We took the age ≤50 as the cut-off point to further analyze the differences between the age-specific groups of smartphones with and without the AI-CNN model.ResultsA total of 1454 eligible subjects completed a colonoscopy, but 20 of them were excluded due to missing data (IDDF2022-ABS-0065 Figure 1). There was no significant difference between the age-specific groups in BBPS, PDRs, ADRs, and cleanliness of each intestinal segment. But we found out all subjects with a ‘pass’ designation in the younger group had a higher BBPS in the AI-CNN group than in the control group (7.32 ± 1.41 vs. 7.15 ± 1.44, p =0.044). Also, the younger group with the proportion of BBPS ≥ 7 was higher in the AI-CNN group (72.5% vs. 66.9%, p =0.046) (IDDF2022-ABS-0065 Table 1). BBPS ≥ 7 was considered the ‘Good’ bowel preparation. We, therefore, think that younger people might benefit from the model by achieving higher BBPS.Abstract IDDF2022-ABS-0065 Table 1Quality of bowel preparation and outcomes of subjects with PASS results (age ≤50) Parameters PASSN = 1067 ControlN = 522 AI-CNNN = 545 P-value+ Bowel preparation quality, n (%) Good/Excellent (BBPS >6) 744 (69.7) 349 (66.9) 395 (72.5) 0.046* Inadequate/poor/fair (BBPS ≤ 6) 323 (30.3) 173 (33.1) 150 (27.5) BBPS score, mean ± SD Total 7.24 ± 1.43 7.15 ± 1.44 7.32 ± 1.41 0.044* Right colon 2.31 ± 0.61 2.27 ± 0.60 2.34 ± 0.61 0.032* Transverse colon 2.59 ± 0.54 2.56 ± 0.55 2.61 ± 0.53 0.183 Left colon 2.35 ± 0.57 2.32 ± 0.58 2.37 ± 0.56 0.212 *Statistical significance +Comparison between AI-CNN and Control groups. Abbreviations: AI-CNN, artificial intelligence-convolutional neural network; BBPS, Boston Bowel Preparation ScaleAbstract IDDF2022-ABS-0065 Figure 1ConclusionsSmartphones equipped with the AI-CNN model have better BBPS performance in younger groups. This may be related to the unskilled operation of smartphones by the elderly. In the future, for the bowel preparation of the elderly, smartphone applications with AI assistance should be designed from more aspects, which will be worthy of further research.