Spada C, Piccirelli S, Hassan C, Ferrari C, Toth E, González-Suárez B, Keuchel M, McAlindon M, Finta Á, Rosztóczy A, Dray X, Salvi D, Riccioni ME, Benamouzig R, Chattree A, Humphries A, Saurin JC, Despott EJ, Murino A, Johansson GW, Giordano A, Baltes P, Sidhu R, Szalai M, Helle K, Nemeth A, Nowak T, Lin R, and Costamagna G
Background: Capsule endoscopy reading is time consuming, and readers are required to maintain attention so as not to miss significant findings. Deep convolutional neural networks can recognise relevant findings, possibly exceeding human performances and reducing the reading time of capsule endoscopy. Our primary aim was to assess the non-inferiority of artificial intelligence (AI)-assisted reading versus standard reading for potentially small bowel bleeding lesions (high P2, moderate P1; Saurin classification) at per-patient analysis. The mean reading time in both reading modalities was evaluated among the secondary endpoints., Methods: Patients aged 18 years or older with suspected small bowel bleeding (with anaemia with or without melena or haematochezia, and negative bidirectional endoscopy) were prospectively enrolled at 14 European centres. Patients underwent small bowel capsule endoscopy with the Navicam SB system (Ankon, China), which is provided with a deep neural network-based AI system (ProScan) for automatic detection of lesions. Initial reading was performed in standard reading mode. Second blinded reading was performed with AI assistance (the AI operated a first-automated reading, and only AI-selected images were assessed by human readers). The primary endpoint was to assess the non-inferiority of AI-assisted reading versus standard reading in the detection (diagnostic yield) of potentially small bowel bleeding P1 and P2 lesions in a per-patient analysis. This study is registered with ClinicalTrials.gov, NCT04821349., Findings: From Feb 17, 2021 to Dec 29, 2021, 137 patients were prospectively enrolled. 133 patients were included in the final analysis (73 [55%] female, mean age 66·5 years [SD 14·4]; 112 [84%] completed capsule endoscopy). At per-patient analysis, the diagnostic yield of P1 and P2 lesions in AI-assisted reading (98 [73·7%] of 133 lesions) was non-inferior (p<0·0001) and superior (p=0·0213) to standard reading (82 [62·4%] of 133; 95% CI 3·6-19·0). Mean small bowel reading time was 33·7 min (SD 22·9) in standard reading and 3·8 min (3·3) in AI-assisted reading (p<0·0001)., Interpretation: AI-assisted reading might provide more accurate and faster detection of clinically relevant small bowel bleeding lesions than standard reading., Funding: ANKON Technologies, China and AnX Robotica, USA provided the NaviCam SB system., Competing Interests: Declaration of interests CS, SP, ET, MK, XD, and EJD received support for attending meetings or travel from AnX Robotica. CS, MK, and EJD received honoraria from AnX Robotica for the adjudication committee. J-CS received teaching fees from AnX Robotica. RS participated in an Ankon-X-advisory meeting. TN received consulting fees from AnX Robotica. AnX Robotica provided all centres with the Navicam SB system at no cost, in instances where it was not readily available. Complimentary capsules were supplied in centres where regional reimbursement did not encompass the procedure., (Copyright © 2024 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC 4.0 license. Published by Elsevier Ltd.. All rights reserved.)