A deep learning system for detection of early Barrett's neoplasia: a model development and validation study.

BACKGROUND: Computer-aided detection (CADe) systems could assist endoscopists in detecting early neoplasia in Barrett's oesophagus, which could be difficult to detect in endoscopic images. The aim of this study was to develop, test, and benchmark a CADe system for early neoplasia in Barrett's oesophagus. METHODS: The CADe system was first pretrained with ImageNet followed by domain-specific pretraining with GastroNet. We trained the CADe system on a dataset of 14 046 images (2506 patients) of confirmed Barrett's oesophagus neoplasia and non-dysplastic Barrett's oesophagus from 15 centres. Neoplasia was delineated by 14 Barrett's oesophagus experts for all datasets. We tested the performance of the CADe system on two independent test sets. The all-comers test set comprised 327 (73 patients) non-dysplastic Barrett's oesophagus images, 82 (46 patients) neoplastic images, 180 (66 of the same patients) non-dysplastic Barrett's oesophagus videos, and 71 (45 of the same patients) neoplastic videos. The benchmarking test set comprised 100 (50 patients) neoplastic images, 300 (125 patients) non-dysplastic images, 47 (47 of the same patients) neoplastic videos, and 141 (82 of the same patients) non-dysplastic videos, and was enriched with subtle neoplasia cases. The benchmarking test set was evaluated by 112 endoscopists from six countries (first without CADe and, after 6 weeks, with CADe) and by 28 external international Barrett's oesophagus experts. The primary outcome was the sensitivity of Barrett's neoplasia detection by general endoscopists without CADe assistance versus with CADe assistance on the benchmarking test set. We compared sensitivity using a mixed-effects logistic regression model with conditional odds ratios (ORs; likelihood profile 95% CIs). FINDINGS: Sensitivity for neoplasia detection among endoscopists increased from 74% to 88% with CADe assistance (OR 2·04; 95% CI 1·73-2·42; p<0·0001 for images and from 67% to 79% [2·35; 1·90-2·94; p<0·0001] for video) without compromising specificity (from 89% to 90% [1·07; 0·96-1·19; p=0·20] for images and from 96% to 94% [0·94; 0·79-1·11; ] for video; p=0·46). In the all-comers test set, CADe detected neoplastic lesions in 95% (88-98) of images and 97% (90-99) of videos. In the benchmarking test set, the CADe system was superior to endoscopists in detecting neoplasia (90% vs 74% [OR 3·75; 95% CI 1·93-8·05; p=0·0002] for images and 91% vs 67% [11·68; 3·85-47·53; p<0·0001] for video) and non-inferior to Barrett's oesophagus experts (90% vs 87% [OR 1·74; 95% CI 0·83-3·65] for images and 91% vs 86% [2·94; 0·99-11·40] for video). INTERPRETATION: CADe outperformed endoscopists in detecting Barrett's oesophagus neoplasia and, when used as an assistive tool, it improved their detection rate. CADe detected virtually all neoplasia in a test set of consecutive cases. FUNDING: Olympus.

Evaluation of a novel infra-red endoscopy system in the assessment of early neoplasia in Barretts esophagus: pilot study from a single center.

Infrared endoscopy (IRE) has been shown to be useful in detecting submucosal (SM) invasion in early gastric cancer. Its role in the endoscopic assessment of Barrett's neoplasia has not been reported to date. We aimed in this study to evaluate the role of IRE in the detection and characterization of early neoplastic lesions within Barrett's esophagus (BE). The secondary aim was to explore its usefulness for the assessment of the presence of submucosal invasion in these early neoplastic Barrett's lesions. We included in the study patients with dysplastic BE who were referred to our institution for endoscopic therapy of a previously diagnosed early Barrett's neoplasia. An examination with white light high resolution endoscopy (HRE) and near IRE after intravenous injection of indocyanine green was performed for all patients using an infrared endoscope prototype. Staining on IRE and correlation with final histological staging by endoscopic mucosal resection/surgery or histological diagnosis on mapping biopsies was analyzed. A total of 23 patients were enrolled in our study: 17 of them with 19 visible lesions and 6 patients with flat BE and no lesions. Staining on IRE was noted in 18 cases: 17 (94%) had at least high grade dysplasia (HGD). No stain was noted in 7 cases: final histology was

Incidence of metachronous visible lesions in patients referred for radiofrequency ablation (RFA) therapy for early Barrett's neoplasia: a single-centre experience.

OBJECTIVE: Evaluate the incidence of metachronous visible lesions (VLs) in patients referred for radiofrequency ablation (RFA) for early Barrett's neoplasia. DESIGN: This study was conducted as part of the service evaluation audit. SETTING: Tertiary referral centre. PATIENTS: All patients with dysplastic Barrett's oesophagus referred for RFA were included for analysis. White light high-resolution endoscopy (HRE), autofluorescence imaging and narrow band imaging were sequentially performed. Endoscopic mucosal resection (EMR) was performed for all VL. Three to six months after EMR, all patients underwent initial RFA and then repeat RFA procedures at three monthly intervals. INTERVENTIONS: All endoscopy reports and final staging by EMR/surgery were evaluated and included for analysis. RESULTS: Fifty patients were analysed; median age 73 years, 84% men. 38/50 patients (76%) had a previous EMR due to the presence of VL before referred for ablation; twelve patients had no previous treatment. In total, 151 ablation procedures were performed, median per patient 2.68. Twenty metachronous VL were identified in 14 patients before the first ablation or during the RFA protocol; incidence was 28%. All metachronous lesions were successfully resected by EMR. Upstaging after rescue EMR compared with the initial histology was observed in four patients (28%). CONCLUSIONS: In total, 28% of patients enrolled in the RFA programme were diagnosed to have metachronous lesions. This high-incidence rate highlights the importance of a meticulous examination to identify and resect any VL before every ablation session. RFA treatment for early Barrett's neoplasia should be performed in tertiary referral centres with HRE and EMR facilities and expertise.

Animal models for endoscopic training: do we really need them?

Gastrointestinal endoscopy currently includes many therapeutic methods that are technically challenging and frequently associated with a significant risk of complications. Several issues such as the limited number of clinical cases and practice in emergency situations, and technical difficulty may limit the opportunity for training, and increased exposure in more relaxed situations would be desirable. Moreover, providing the patient with the best possible standard of care is a must. Animal models are the most easily available simulators. Training in these models has been recommended for several complex techniques, among which hemostasis, endoscopic ultrasound, endoscopic retrograde cholangiopancreatography, and endoscopic submucosal dissection are reviewed here. Ex vivo models are much easier to set up and, from an ethical standpoint, they should be used for the initial step in training whenever possible before moving on to in vivo models. Although simulation with animal models has been the subject of a good number of studies, very few of them have evaluated the impact on clinical outcomes, and clearly more studies are needed. Nevertheless, available evidence does suggest that practicing on animal models has an influence on the learning curve and facilitates the acquisition of skills in the complex endoscopic techniques reviewed.