Gaius Longcroft-Wheaton

Gastrointestinal (GI) endoscopy comprises both diagnostic and therapeutic procedures involving the luminal GI tract as well as the biliary tree, liver, and pancreas. GI endoscopy is challenging to learn, requiring both cognitive (nontechnical) and technical skills, and requires extensive practice to attain proficiency. Simulation-based training has been shown to assist trainees and young endoscopists in acquiring new skills and accelerating the learning curve. Moreover, simulation-based training creates an ideal environment for trainees to initially learn and practice skills while making mistakes with no risk to patients. This review, divided in two parts, offers a comprehensive summary of the different classes of simulators available for GI endoscopic training. In Part I, only mechanical simulators are reported and described. In Part II, animal simulators (ex vivo/in vivo) and virtual reality models are detailed, together with prototypes that are currently not commercially available. ^‡ Joint last authors Article published online: 04 April 2025 © 2025. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany

Barrett’s oesophagus is the only known precursor to oesophageal adenocarcinoma, a cancer with very poor prognosis. The main risk factors for Barrett’s oesophagus are a history of gastro-oesophageal acid reflux symptoms and obesity. Men, smokers and those with a family history are also at increased risk. Progression from Barrett’s oesophagus to cancer occurs via an intermediate stage, known as dysplasia. However, dysplasia and early cancer usually develop without any clinical signs, often in individuals whose symptoms are well controlled by acid suppressant medications; therefore, endoscopic surveillance is recommended to allow for early diagnosis and timely clinical intervention. Individuals with Barrett’s oesophagus need to be fully informed about the implications of this diagnosis and the benefits and risks of monitoring strategies. Pharmacological treatments are recommended for control of symptoms, but not for chemoprevention. Dysplasia and stage 1 oesophageal adenocarcinoma have excellent prognoses, since they can be cured with endoscopic or surgical therapies. Endoscopic resection is the most accurate staging technique for early Barrett’s-related oesophageal adenocarcinoma. Endoscopic ablation is effective and indicated to eradicate Barrett’s oesophagus in patients with dysplasia. Future research should focus on improved accuracy for dysplasia detection via new technologies and providing more robust evidence to support pathways for follow-up and treatment.

Background Device-assisted enteroscopy (DAE) has become a well-established diagnostic and therapeutic tool for the management of small-bowel pathology. We aimed to evaluate the performance measures for DAE across the UK against the quality benchmarks proposed by the European Society of Gastrointestinal Endoscopy (ESGE). Methods We retrospectively collected data on patient demographics and DAE performance measures from electronic endoscopy records of consecutive patients who underwent DAE for diagnostic and therapeutic purposes across 12 enteroscopy centers in the UK between January 2017 and December 2022. Results A total of 2005 DAE procedures were performed in 1663 patients (median age 60 years; 53% men). Almost all procedures (98.1%) were performed for appropriate indications. Double-balloon enteroscopy was used for most procedures (82.0%), followed by single-balloon enteroscopy (17.2%) and spiral enteroscopy (0.7%). The estimated depth of insertion was documented in 73.4% of procedures. The overall diagnostic yield was 70.0%. Therapeutic interventions were performed in 42.6% of procedures, with a success rate of 96.6%. Overall, 78.0% of detected lesions were marked with a tattoo. Patient comfort was significantly better with the use of deep sedation compared with conscious sedation (99.7% vs. 68.5%; P<0.001). Major adverse events occurred in only 0.6% of procedures. Conclusions Performance measures for DAE in the UK meet the ESGE quality benchmarks, with high diagnostic and therapeutic yields, and a low incidence of major adverse events. However, there is room for improvement in optimizing sedation practices, standardizing the depth of insertion documentation, and adopting marking techniques to aid in the follow-up of detected lesions.

Aims “Resect and discard” strategy for colorectal polyps is based on the negligible risk of cancer in diminutive (1000 lifetime colonoscopies). Results 292 polyps (106 non-diminutive and 186 diminutive) were included in Phase 1 study. Sensitivity, specificity, and accuracy of AI system classifying into non-diminutive polyps were 93.40%, 79.03% and 84.25% respectively. In Phase 2, 135 polyps (75 non-diminutive (median size of 10mm) and 60 diminutive (median size of 4mm)) from 38 colonoscopies were included. The sensitivity, specificity, and accuracy of AI system sizing into non-diminutive polyps were 82.67%, 88.33% and 85.19% respectively. Conclusions Our data from both phases of the study demonstrates that AI-based sizing is effective in differentiating non-diminutive from diminutive polyps. This can facilitate the introduction of “resect and discard” strategy ([Table 1]). Article published online: 14 April 2023 © 2023. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany

Aims Barrett’s intramucosal neoplasia can be endoscopically cured with minimal risk of lymph node metastasis while submucosal neoplasia poses higher risk and usually requires surgery. Endoscopic optical diagnosis is often challenging to differentiate between the two. We aim to develop an artificial intelligence (AI) system using a deep neural network to stage Barrett’s neoplasia into intramucosal (LGD, HGD, pT1a) and submucosal (pT1b) lesions. Methods A VGG16-based deep neural network was used for the staging AI system. The model was trained with 69 images which were carefully selected from 79,159 frame images with histological ground truth. As a pre-trained model for training, the model from initial Barrett’s detection and delineation system was used, which was trained for 1,090,171 images from 161 Barrett's patients¹. We tested staging AI on pre-recorded images and its performance was compared to two experts who regularly perform Barrett's endoscopic resections. Results 68 neoplastic images (white light, enhanced imaging, near focused/magnification) were used. 43 were intramucosal (2 LGD, 31 HGD, 10 pT1a) and 25 were submucosal (6 pT1bSM1 and 19 pT1bSM2/3). The sensitivity, specificity and accuracy of AI were 80%, 84% and 82% while that of experts were 74%, 78% and 76% with no statistical difference (p=0.507) [1]. Conclusions This suggests that the staging AI system is good at differentiating submucosal neoplasia and it can perform as well as experts. It has the potential to help guide clinicians to decide if endoscopic resection of Barrett's neoplasia can be curative. However, more work is still needed to further refine and validate this staging AI model. Article published online: 14 April 2023 © 2023. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany

Aims Despite improvements in endoscopic imaging modalities, there are still significant miss rates in detecting Barrett’s neoplasia. We have previously reported the development of novel deep neural network in detecting Barrett’s neoplasia¹. The aim of this study was to understand the added value of AI system across a range of expert and non-expert endoscopists. Methods We used WiseVision (NEC Corporation, Japan) as an AI platform. Pre-recorded videos were edited and histology was used as a ground truth. 10 experts and 10 non-experts were asked to review the videos without AI and then with AI and asked to categorise into neoplastic, non-neoplastic or uncertain. Videos were randomly distributed to reduce recall bias ([Fig. 1]). Results 114 videos (57 neoplastic and 57 non-neoplastic) were collected. There were 15 polypoidal and 45 non-polypoidal neoplasia. Sensitivity of AI is 84.5%. There is a significant improvement in performance of non-experts with AI (sensitivity 63% to 81%). Improvement in experts with AI is statistically significant but not clinically significant (69% to 73%). With AI, non-experts have a greater improvement in achieving fewer uncertain answers compared to experts (12.8% to 4.8% in non-experts vs 11.6% to 8.9% in experts) [1]. Conclusions Sensitivity of AI looks promising and non-experts seem to agree more with AI, resulting in a greater gain in detecting additional neoplasia. Experts, however, seem to disregard AI results, resulting in very little additional gain and this could potentially deprive patients from additional AI-related benefits. Our data gives an interesting insight into human-machine interaction which needs to be further explored before the widespread use of AI. Article published online: 14 April 2023 © 2023. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany

Objectives: The EndoRings device is a distal attachment consisting of two layers of circular flexible rings that evert mucosal folds. The aim of this study was to investigate whether EndoRing assisted colonoscopy (ER) improves polyp and adenoma detection compared to standard colonoscopy (SC).Methods: Multicenter, parallel group, randomized controlled trial.Results: Total of 556 patients randomized to ER (n = 275) or SC (n = 281). Colonoscopy completed in 532/556 (96%) cases. EndoRings removed in 74/275 (27%) patients. Total number of polyps in ER limb 582 vs. 515 in SC limb, P = 0.04. Total number of adenomas in ER limb 361 vs. 343 for SC limb, P = 0.49. A statistically significant difference in the mean number of polyps per patient in both the intention to treat (1.84 SC vs. 2.10 ER, P = 0.027) and per protocol (PP) (1.84 SC vs. 2.25 ER, P = 0.004).Conclusions: Our study shows promise for the EndoRings device to improve polyp detection.

Introduction Endoscopic differentiation between intramucosal and submucosal Barrett’s neoplasia has several important implications but remains challenging even for expert endoscopists. Recent studies demonstrated promising results on AI-assisted detection of Barrett’s neoplasia, but data on AI-assisted staging is limited. We aimed to develop and validate an AI system for classification of Barrett’s neoplasia into intramucosal or submucosal, and compare its performance to expert endoscopists. Methods The model, based on VGG-16 architecture, was trained on 117 images of prospectively collected and annotated Barrett’s neoplastic lesions. Rotation and random flip were used for data augmentation. The ground truth was the histological staging of endoscopically resected specimens performed by two pathologists with expertise in Barrett’s neoplasia. Images comprised of WLI, enhanced imaging, and magnification views. The model was designed to classify images as either intramucosal (pT1a) or submucosal (pT1b). Performance of the AI system was compared to a group of three experts. Results The AI model was tested on an independent dataset of 90 images. The accuracy, sensitivity and specificity and AUC of the AI model in differentiating between intramucosal and submucosal neoplasia was 70.9%, 72.5%, 65.7%, and 0.781 respectively. Mean accuracy, sensitivity and specificity of experts were 73.3%, 63.3% and 83.3% respectively. Processing speed of the AI system was 5 ms/image. Performance of AI system compared to three experts on differentiating between mucosal and submucosal Barrett’s neoplasia (n=90 images) Class activation maps (CAM) images showing interpretable correct prediction of Barrett’s neoplasia staging by a Deep Learning model. The image at the top (A) was correctly classified as an intramucosal, while the image at the bottom (B) was correctly classified as submucosal neoplasia Conclusion This study demonstrates the feasibility of AI-assisted staging of Barrett’s neoplasia on endoscopic images. The AI model’s performance was comparable to that of experts. More work is needed to further develop this early model and validate its use on real-time video sequences.