AID-GI – Artificial Intelligence-supported Diagnostics of Gastrointestinal diseases with video capsule endoscopy

“该项目旨在通过应用机器学习来辅助内部图像分析，提高下消化道疾病，特别是炎症的诊断准确性。目前，胃肠道的分析是通过粪便中的标记物，或通过可食用相机或传统的管式相机拍摄的图像进行的。前者仅限于检测癌症（FIT）或一般性炎症（钙保护蛋白），因此相当不具体，后者由人类操作员进行，因此容易因疲劳、分心、经验变化或其他认知限制而导致人为错误。图像的数量和需要筛查的患者数量，在工作和质量方面给操作者带来了难以管理的负担。这个项目的重点是视频胶囊内窥镜——一种可摄取的“胶囊”相机，它越来越有可能成为胃肠成像的主要方法，胶囊捕捉患者日常生活的图像，而不是去医院或诊所。通过该项目，我们将测试自动化图像分析的最新方法，量化对患者、临床医生和NHS的经济和临床效益，并就如何实施解决方案提出建议。如果成功，这种方法可以消除限制最佳IBD治疗、肠癌预防、早期检测和其他疾病调查的‘诊断瓶颈’。”

"This project is aiming to improve the diagnostic accuracy for lower gastrointestinal diseases, especially inflammations, by applying _machine learning_ to aid the analysis of internal images.Currently, the gastrointestinal tract is analysed using markers in the stool or with images taken via ingestible cameras or traditional tube-mounted cameras. While the former is limited to detecting cancers (FIT) or generic inflammation (calprotectin) and therefore rather unspecific, the latter is undertaken by human operators and therefore prone to human errors resulting from fatigue, distraction, variable experience or other cognitive limitations. The volume of images and the volume of patients requiring screening, places unmanageable loads on the operators in terms of effort and quality.This project focuses on Video Capsule Endoscopy -- ingestible 'capsule' cameras which are increasingly likely to become the dominant approach to gastrointestinal imagery, where the capsules capture images as patients go about their daily lives rather than attending a hospital or clinic.Through the project we will test the latest approaches to automated image analysis, quantify benefits to the patient, clinician and NHS -- financially and clinically -- and make recommendations on how to implement the solution.When successful, this approach can remove the 'diagnostic bottleneck' that limits optimal IBD treatments, bowel cancer prevention, early detection and other disease investigations."