{
  "abstract": "Introduction Adequate bowel preparation is essential for colonoscopy quality, yet endoscopist-based scores are limited by subjectivity. We aimed to externally validate the Artificial Intelligence (AI)-based Open-Source Automated Bowel Preparation Scale (OSABPS) — a Danish-developed computer vision algorithm quantifying the ‘faecal ratio’ from mucosal and faecal pixel counts — and assess its association with clinical outcomes.Methods In a post-hoc analysis of the UCL-led CADDIE multicentre randomised controlled trial of AI polyp detection, OSABPS scores (scale 0–1; 0=optimal, 1=poor) were calculated for procedures with video data and compared against original endoscopist ratings recorded during the trial. Mann-Whitney U tests evaluated the ability of OSABPS to discriminate between adequate and inadequate bowel preparation. Adequate bowel preparation was defined as a Boston Bowel Preparation Scale (BBPS) score ≥2 per segment or a total Aronchick score ≤3. Multivariable logistic and linear regression models used established confounders selected a priori to account for differences in polyp prevalence (patient age, sex, history of polyps) and detection (endoscopist). Models estimated association between OSABPS and detection of adenomas, non-diminutive polyps, and withdrawal time (adjusted for number of polyps resected).Results In 543 procedures, OSABPS significantly discriminated between endoscopist-rated adequate and inadequate preparation (median 0.11 vs 0.19; p=0.002). In multivariable analysis (BBPS cohort; n=329), endoscopist-rated scores were not associated with polyp detection or withdrawal time. In contrast, higher OSABPS scores (worse preparation) were significantly associated with lower detection of adenomas (adjusted Odds Ratio [aOR] 0.77 per 0.1-unit increase; 95% CI 0.60–0.99; p=0.047) and non-diminutive polyps (aOR 0.75; 95% CI 0.57–0.99; p=0.041). Poorer bowel preparation was associated with shorter withdrawal times (β = -0.77 min per 0.1-unit; p=0.017), independent of number of polypectomies performed.Conclusions This study externally validates OSABPS as an objective measure of bowel preparation quality, demonstrating stronger associations with adenoma and non-diminutive polyp detection compared to endoscopist-rated BBPS. The association between worse bowel preparation and shorter withdrawal time may suggest that mucosal inspection is limited in poorly prepared colons. Further studies are required to identify if AI-based bowel preparation scoring has a role as a standardised metric for quality assurance.",
  "authors": [
    {
      "affiliations": [
        "Division of Surgery and Interventional Science, University College London, London, United Kingdom",
        "Department of Gastroenterology, University College London Hospital NHS Foundation Trust, London, United Kingdom"
      ],
      "name": "Julian Gertner"
    },
    {
      "affiliations": [
        "Copenhagen Academy for Medical Education and Simulation (CAMES), Center for HR & Education, Copenhagen, Denmark",
        "Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark"
      ],
      "name": "Kristoffer Cold"
    },
    {
      "affiliations": [
        "Copenhagen Academy for Medical Education and Simulation (CAMES), Center for HR & Education, Copenhagen, Denmark"
      ],
      "name": "Amihai Heen"
    },
    {
      "affiliations": [
        "Copenhagen Academy for Medical Education and Simulation (CAMES), Center for HR & Education, Copenhagen, Denmark",
        "DTU HEALTH TECH, Department of Health Technology, Technical University of Denmark, Denmark"
      ],
      "name": "Morten Svendsen"
    },
    {
      "affiliations": [
        "Copenhagen Academy for Medical Education and Simulation (CAMES), Center for HR & Education, Copenhagen, Denmark",
        "Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark"
      ],
      "name": "Lars Konge"
    },
    {
      "affiliations": [
        "Division of Surgery and Interventional Science, University College London, London, United Kingdom",
        "Department of Gastroenterology, University College London Hospital NHS Foundation Trust, London, United Kingdom"
      ],
      "name": "Omer Ahmad"
    },
    {
      "affiliations": [
        "Division of Surgery and Interventional Science, University College London, London, United Kingdom",
        "Department of Gastroenterology, University College London Hospital NHS Foundation Trust, London, United Kingdom"
      ],
      "name": "Laurence Lovat"
    }
  ],
  "title": "P27 External validation of the open-source automated bowel preparation scale (OSABPS) in a CADDIE trial post-hoc analysis",
  "uid": "128c0fc4-57b1-5d26-959e-4f9fe9e5641b"
}