基于深度学习的人工智能技术在结直肠息肉性质鉴别中的应用

Application of deep learning based artificial intelligence technology in identification of colorectal polyps

结直肠癌是明确可以从结肠镜筛查中获益且最宜开展筛查的癌症,2017年中国结直肠癌新发病例约431951例,发病人数在28年间增长了203.5%.在结肠镜检查中对腺瘤性息肉等癌前病变早发现、早切除,可预防结直肠癌的发生,然而各种因素导致结肠镜检查过程中息肉的漏诊,使结直肠癌发生的风险增加.近年来随着人工智能技术(artificial intelligence,AI)在医学领域的应用,AI辅助结肠镜检查能增加息肉的检出率(adenoma detection rate,ADR),提高结肠镜检查的质量.本文主要就以卷积神经网络(convolutional neural network,CNN)为架构的AI辅助系统在结肠镜检查中的的质量控制(automatic quality control system,AQCS)、肠道准备、结直肠息肉的诊断与分型以及AI在结肠镜领域未来面对的机遇与挑战等方面进行综述,希望能给临床工作提供一些参考.

Colorectal cancer is a cancer type that is most suitable for screening since subjects at risk of this malignancy can clearly benefit from colonoscopy screening.In 2017,there were about 431951 new cases of colorectal cancer in China,with an increase of 203.5%in 28 years.Early detection and early removal of adenomatous polyps and other precancerous lesions during colonoscopy can prevent the occurrence of colorectal cancer.However,various factors lead to missed diagnosis of polyps during colonoscopy,which increases the risk of colorectal cancer.In recent years,with the rapid development of artificial intelligence technology in the medical field,colonoscopy assisted by artificial intelligence can increase the detection rate of polyps and improve the quality of colonoscopy.This paper mainly reviews the quality control,bowel preparation,diagnosis and classification of colorectal polyps,and the future opportunities and challenges faced by convolutional neural network based artificial intelligence technology in the field of colonoscopy,hoping to provide some reference for clinical work.