以人工智能肠道图像识别模型评估结肠镜检查前肠道准备

目的 观察人工智能肠道图像识别模型用于评估结肠镜检查前肠道准备的价值。方法 回顾性分析190例接受肠道准备评估及结肠镜检查患者，根据评估肠道准备方法将其分为观察组(以人工智能肠道图像识别模型进行判断，n=100)和对照组(仅由患者将末次粪便性状与肠道清洁准备图进行对比而判断，n=90);比较2组肠道清洁度、肠镜检查时间及腺瘤检出率。结果 观察组波士顿肠道准备量表(BBPS)评分及腺瘤检出率高于对照组，而肠镜检查时间短于对照组(P均<0.05)。结论 检查前采用人工智能肠道图像识别模型评估肠道准备情况可提高BBPS评分及腺瘤检出率并缩短肠镜检查时间。

Artificial intelligence colonic image recognition model for evaluating bowel preparation before colonoscopy

Objective To observe the value of artificial intelligence colonic image recognition model for evaluating bowel preparation before colonoscopy. Methods Data of 190 patients who underwent bowel preparation assessment and colonoscopy examination were retrospectively analyzed. The patients were divided into observation group (judging bowel preparation with artificial intelligence colonic image recognition model, n=100) or control group (judging bowel preparation by patient according to comparison of the last fecal characteristics with the bowel cleaning preparation map, n=90). The bowel cleanliness, operation time of colonoscopy and detection rate of adenomas were compared between groups. Results Boston bowel preparation scale (BBPS) score and detection rate of adenoma in observation group were both higher than those in control group, while colonoscopy time in observation group was shorter than that in control group (all P<0.05). Conclusion Artificial intelligence colonic image recognition model for evaluating bowel preparation could improve BBPS score and detection rate of adenoma, also shorten colonoscopy time of colonoscopy.