BowelSys: an AI tool to enhance FIT diagnostic yield and improve bowel cancer detection in symptomatic patients

粪便免疫化学检测(FIT)测量粪便血红蛋白浓度(f-Hb)。FIT可用于管理在紧急转诊之前(或同时)在初级保健中出现不明原因的下胃肠道(GI)症状的患者，或用于二级护理中的分类以指导转诊患者的管理。有症状的患者的阳性匹配结果的阈值为10微克Hb/g粪便。任何超过这个预设FIT阈值的参与者都会被推荐接受内窥镜检查(例如乙状结肠镜检查或结肠镜检查)，以确定任何医学上可操作的出血原因，即结直肠癌(CRC)或癌前病变(息肉)。与传统的基于愈创木酚的f-Hb检测相比，FIT的灵敏度提高，因此转诊至初级保健的内窥镜手术的数量正在上升。2021年的一项研究确定，在有症状的低风险人群中，FIT的阳性预测值为7.0%。这表明假阳性率为93%，这意味着13名f-Hb\&gt；10微克Hb/g粪便患者接受乙状结肠镜或结肠镜检查以确定一例CRC。有一项迫在眉睫的、尚未得到满足的需求**是提高FIT对检测肠癌的诊断效率。这将减少内窥镜检查的转诊，保护患者免受可能导致肠穿孔(一种可能致命的严重不良事件)的不必要的侵入性筛查，每个手术的费用为GB 322**-**GB 548。这个项目的临床合作者已经收集的数据揭示了根据特定的个人数据变量调整有症状患者的个性化匹配阈值的范围，这可能有助于改进后续转诊。**解决方案：**高级专家系统有限公司(AES)已经与来自大学医院考文垂和华威郡中部地区NHS信托基金、约克和斯卡伯勒教学医院NHS基金会信托基金(约克)和莱斯特大学医院NHS信托基金(莱斯特)的临床专家一起开发了一个基于人工智能的软件平台‘BowelSys’，以在与肠癌发展相关的其他临床和非临床风险因素的背景下解释Fit结果。作为一个多变量分层模型，BowelSys能够对FIT结果进行个性化解释，并根据风险优化临床资源。高危患者被快速追踪以进行进一步调查，而低风险或无风险患者的进一步调查被推迟。这将使临床资源得到更有效的利用，高危患者可以快速进行内窥镜检查，从而能够更早、更好地检测肿瘤。**及时性：**由于新冠肺炎，现在有额外的动力减少不必要的内窥镜转诊，以遵守感染控制措施，解决漫长的等待名单，提高国民健康保险制度的效率，并降低成本。

Faecal immunochemical testing (FIT) measures faecal haemoglobin concentration (f-Hb).FIT can be used to manage patients presenting in primary care with unexplained lower gastrointestinal (GI) symptoms that may be indicative of bowel cancer in primary care prior to (or alongside) an urgent referral, or for triaging in secondary care to guide the management of referred patients.The threshold for a positive FIT result in symptomatic patients is 10µg Hb/g faeces. Any participant exceeding this pre-set FIT threshold value is referred for endoscopic services (e.g., sigmoidoscopy or colonoscopy) to identify any medically actionable cause of bleeding i.e., colorectal cancer (CRC) or a precancerous adenoma (polyp).Coupled to increased sensitivity of FIT vs. traditional guaiac-based f-Hb detection, the number of referrals from primary care for endoscopic procedures is rising.A 2021 study determined the positive predictive value of FIT in a low-risk symptomatic population was 7.0%. This indicates a false-positive rate of 93%, meaning 13 patients with f-Hb \>10 µg Hb/g faeces underwent sigmoidoscopy or colonoscopy to identify one CRC.There is an immediate and pressing **unmet need** to improve FIT's diagnostic yield for detecting bowel cancer. This will reduce endoscopy referrals, protecting patients from unnecessary invasive screening that can cause bowel perforation (a serious adverse event that can be lethal), and costs £322**-**£548 per procedure. Data already collected by the clinical collaborators to this project has revealed scope to modulate personalised FIT thresholds for symptomatic patients based on specific individual data variables that could help to refine onward referrals.**SOLUTION:** Advanced Expert Systems Ltd (AES) have joined clinical specialists from University Hospitals Coventry & Warwickshire Midlands NHS Trust, York and Scarborough Teaching Hospital NHS Foundation Trust (York), and University Hospitals of Leicester NHS Trust (Leicester) to develop an AI-based software platform 'BowelSys' to interpret FIT result in the context of other clinical and non-clinical risk factors associated with the development of bowel cancer.As a multivariate stratification model, BowelSys enables a personalised interpretation of FIT result and optimises clinical resources by risk. High-risk patients are fast-tracked for further investigation, whilst patients with low-to-no risk have further investigations delayed. This will enable clinical resources to be used more efficiently, with high-risk patients fast-tracked for endoscopy, enabling earlier and better detection of neoplasias.**TIMELINESS:** Due to Covid-19, there is now additional impetus to reduce unnecessary endoscopy referrals to comply with infection control measures, tackle long waiting lists, improve NHS efficiency, and reduce costs.