人工智能技术对社区大肠癌筛查高危人群息肉检出率的影响

Effect of artificial intelligence on polyp detection rate of community colorectal carcinoma screening in high-risk population

摘要:

目的 探讨人工智能技术对社区大肠癌（CRC）筛查高危人群息肉检出率（PDR）的影响。方法 选取2020年12月－2021年3月社区CRC筛查结果阳性并接受结肠镜检查的患者202例。分析受检者的临床资料、肠道准备情况和内镜精灵辅助诊断等指标，比较PDR。结果 PDR为37.62%。单因素分析结果提示，检出组年龄大于未检出组［（61.26±6.45）和（58.37±7.01）岁，P = 0.004］，体重重于未检出组［（65.28±9.51）和（60.81±9.20）kg，P = 0.001］，体重指数（BMI）高于未检出组［（24.80±2.69）和（23.51±2.96）kg/m²，P = 0.002］，男性占比高于未检出组（38.16%和24.60%，P = 0.041），有慢性腹泻或便秘症状者占比为42.11%，未检出组为55.56%，差异无统计学意义（P = 0.064），使用内镜精灵的占比高于未检出组（67.11%和43.65%，P = 0.001）。多因素回归分析结果表明，年龄（OR＾ = 1.066，95％CI：1.017～1.119，P = 0.008）和使用内镜精灵（OR＾ = 3.101，95％CI：1.625～5.916，P = 0.001）为PDR的独立影响因素。对使用内镜精灵组进一步分析发现，年龄、体重、BMI、退镜时间、肠道准备评分和性别对PDR有影响。多因素回归分析结果显示，退镜时间（OR＾ = 1.007，95％CI：1.001～1.013，P = 0.027）和肠道准备评分（OR＾ = 1.535，95％CI：1.018～2.316，P = 0.041）为PDR的独立预测因素。结论 对于社区CRC筛查高危患者，男性、高龄和肥胖患者息肉发生风险可能更高。提高肠道准备评分，使用内镜精灵辅助诊断时，延长退镜时间，可能有助于病灶检出。

Abstract:

Objective To investigate the influencing factors of polyp detection rate (PDR) and the role of artificial intelligence in high-risk population in community colorectal carcinoma (CRC) screening.Methods 202 patients with CRC screening positive from December 2020 to March 2021 were selected and subjected to colonoscopy. PDR was compared according to the clinical characteristics, quality of bowel preparation, and ENDOANGEL-assisted diagnoses.Results The PDR was 37.62% in this study. Univariate analysis showed that the age of the detection group was elder than that of the non-detection group [(61.26 ± 6.45) and (58.37 ± 7.01) years old, P = 0.004), the weight of the detection group was higher than that of the non-detection group [(65.28 ± 9.51) and (60.81 ± 9.20) kg, P = 0.001), the body mass index (BMI) of the detection group was higher than that of the non-detection group [(24.80 ± 2.69) and (23.51 ± 2.96) kg/m², P = 0.002], the proportion of males was higher than that of the non-detection group (38.16% and 24.60%, P = 0.041), the proportion of chronic diarrhea or constipation symptoms was 42.11%, while the non-detection group was 55.56%, with no statistically significant difference (P = 0.064), and the proportion of using of ENDOANGEL-assisted diagnosis was higher than that of the non-detection group (67.11% and 43.65%, P = 0.001). Multivariate analysis showed that age (OR＾ = 1.066, 95%CI: 1.017 ~ 1.119, P = 0.008) and the use of ENDOANGEL (OR＾ = 3.101, 95%CI: 1.625 ~ 5.916, P = 0.001) were independent influencing factors to PDR. For the further analysis of the ENDOANGEL-assisted group, it was found that age, body weight, BMI, withdrawal time, intestinal preparation score and gender had an impact on PDR. Withdrawal time (OR＾ = 1.007, 95%CI: 1.001 ~ 1.013, P = 0.027) and intestinal preparation score (OR＾ = 1.535, 95%CI: 1.018 ~ 2.316, P = 0.041) were independent predictors.Conclusion For high-risk patients who are screened for CRC, male, elderly, and obese patients have a greater risk of polyps. Improving the intestinal preparation score and extending the withdrawal time when using ENDOANGEL assistance for diagnosis may be more helpful in detecting lesions.