Xu Xirong, Liu Jiahao, Qiu Jianwei, Fan Benfang, He Tao, Feng Shichun, Sun Jinjie, Ge Zhenming
Digestive Endoscopy Center, Nantong First People's Hospital, Affiliated Hospital 2 of Nantong University, Nantong, Jiangsu, China.
School of Public Health, Nantong University, Nantong, Jiangsu, China.
Br J Hosp Med (Lond). 2025 Jan 24;86(1):1-11. doi: 10.12968/hmed.2024.0577.
Artificial intelligence (AI), with advantages such as automatic feature extraction and high data processing capacity and being unaffected by fatigue, can accurately analyze images obtained from colonoscopy, assess the quality of bowel preparation, and reduce the subjectivity of the operating physician, which may help to achieve standardization and normalization of colonoscopy. In this study, we aimed to explore the value of using an AI-driven intestinal image recognition model to evaluate intestinal preparation before colonoscopy. In this retrospective analysis, we analyzed the clinical data of 98 patients who underwent colonoscopy in Nantong First People's Hospital from May 2023 to October 2023. Among them, 47 cases were evaluated based on the intestinal preparation map and the last fecal characteristics (Regular group), and 51 cases were evaluated using an AI-driven intestinal image recognition model (AI group). The duration of colonoscopy examination, intestinal cleanliness, incidence of adverse reactions, and satisfaction with intestinal preparation of the two groups were analyzed. The time for colonoscopy in the AI group was shorter than that in the Regular group, and the intestinal cleanliness score in the AI group was higher than that in the Regular group ( < 0.05). The incidence of adverse reactions in the AI group (3.92%) was lower than that in the Regular group (10.64%), but the difference was not statistically significant ( > 0.05). The satisfaction rate of intestinal preparation in the AI group (96.08%) was comparable to that of the Regular group (82.98%) ( > 0.05). Compared with the assessment based solely on the intestinal preparation map and the last fecal characteristics, the application of AI intestinal image recognition model in intestinal preparation before colonoscopy can shorten the time of colonoscopy and improve intestinal cleanliness, but with comparable patient satisfaction and safety.
人工智能（AI）具有自动特征提取、高数据处理能力且不受疲劳影响等优势，能够准确分析结肠镜检查所获得的图像，评估肠道准备质量，并降低操作医师的主观性，这可能有助于实现结肠镜检查的标准化和规范化。在本研究中，我们旨在探讨使用人工智能驱动的肠道图像识别模型评估结肠镜检查前肠道准备情况的价值。在这项回顾性分析中，我们分析了2023年5月至2023年10月在南通市第一人民医院接受结肠镜检查的98例患者的临床资料。其中，47例根据肠道准备图谱和末次粪便特征进行评估（常规组），51例使用人工智能驱动的肠道图像识别模型进行评估（人工智能组）。分析了两组的结肠镜检查时间、肠道清洁度、不良反应发生率以及对肠道准备的满意度。人工智能组的结肠镜检查时间短于常规组，人工智能组的肠道清洁度评分高于常规组（<0.05）。人工智能组的不良反应发生率（3.92%）低于常规组（10.64%），但差异无统计学意义（>0.05）。人工智能组的肠道准备满意率（96.08%）与常规组（82.98%）相当（>0.05）。与仅基于肠道准备图谱和末次粪便特征的评估相比，人工智能肠道图像识别模型在结肠镜检查前肠道准备中的应用可缩短结肠镜检查时间并提高肠道清洁度，但患者满意度和安全性相当。
