A new artificial intelligence system successfully detects and localises early neoplasia in Barrett's esophagus by using convolutional neural networks.

BACKGROUND AND AIMS: Seattle protocol biopsies for Barrett's Esophagus (BE) surveillance are labour intensive with low compliance. Dysplasia detection rates vary, leading to missed lesions. This can potentially be offset with computer aided detection. We have developed convolutional neural networks (CNNs) to identify areas of dysplasia and where to target biopsy. METHODS: 119 Videos were collected in high-definition white light and optical chromoendoscopy with i-scan (Pentax Hoya, Japan) imaging in patients with dysplastic and non-dysplastic BE (NDBE). We trained an indirectly supervised CNN to classify images as dysplastic/non-dysplastic using whole video annotations to minimise selection bias and maximise accuracy. The CNN was trained using 148,936 video frames (31 dysplastic patients, 31 NDBE, two normal esophagus), validated on 25,161 images from 11 patient videos and tested on 264 iscan-1 images from 28 dysplastic and 16 NDBE patients which included expert delineations. To localise targeted biopsies/delineations, a second directly supervised CNN was generated based on expert delineations of 94 dysplastic images from 30 patients. This was tested on 86 i-scan one images from 28 dysplastic patients. FINDINGS: The indirectly supervised CNN achieved a per image sensitivity in the test set of 91%, specificity 79%, area under receiver operator curve of 93% to detect dysplasia. Per-lesion sensitivity was 100%. Mean assessment speed was 48 frames per second (fps). 97% of targeted biopsy predictions matched expert and histological assessment at 56 fps. The artificial intelligence system performed better than six endoscopists. INTERPRETATION: Our CNNs classify and localise dysplastic Barrett's Esophagus potentially supporting endoscopists during surveillance.

Endoscopic eradication therapy for Barrett's esophagus-related neoplasia: a final 10-year report from the UK National HALO Radiofrequency Ablation Registry.

BACKGROUND AND AIMS: Long-term durability data for effectiveness of radiofrequency ablation (RFA) to prevent esophageal adenocarcinoma in patients with dysplastic Barrett's esophagus (BE) are lacking. METHODS: We prospectively collected data from 2535 patients with BE (mean length, 5.2 cm; range, 1-20) and neoplasia (20% low-grade dysplasia, 54% high-grade dysplasia, 26% intramucosal carcinoma) who underwent RFA therapy across 28 UK hospitals. We assessed rates of invasive cancer and performed detailed analyses of 1175 patients to assess clearance rates of dysplasia (CR-D) and intestinal metaplasia (CR-IM) within 2 years of starting RFA therapy. We assessed relapses and rates of return to CR-D (CR-D2) and CR-IM (CR-IM2) after further therapy. CR-D and CR-IM were confirmed by an absence of dysplasia and intestinal metaplasia on biopsy samples taken at 2 consecutive endoscopies. RESULTS: Ten years after starting treatment, the Kaplan-Meier (KM) cancer rate was 4.1% with a crude incidence rate of .52 per 100 patient-years. CR-D and CR-IM after 2 years of therapy were 88% and 62.6%, respectively. KM relapse rates were 5.9% from CR-D and 18.7% from CR-IM at 8 years, with most occurring in the first 2 years. Both were successfully retreated with rates of CR-D2 of 63.4% and CR-IM2 of 70.0% 2 years after retreatment. EMR before RFA increased the likelihood of rescue EMR from 17.2% to 41.7% but did not affect the rate of CR-D, whereas rescue EMR after RFA commenced reduced CR-D from 91.4% to 79.7% (χ2P < .001). CONCLUSIONS: RFA treatment is effective and durable to prevent esophageal adenocarcinoma. Most treatment relapses occur early and can be successfully retreated.

Learning curves and the influence of procedural volume for the treatment of dysplastic Barrett's esophagus.

BACKGROUND AND AIMS: Endoscopic resections and radiofrequency ablation (RFA) are the established treatments for Barrett's-associated dysplasia and early esophageal neoplasia. The UK RFA Registry collects patient outcomes from 24 centers treating patients in the United Kingdom and Ireland. Learning curves for treatment of Barrett's dysplasia and the impact of center caseload on patient outcomes is still unknown. METHODS: We examined outcomes of 678 patients treated with RFA in the UK Registry using risk-adjusted cumulative sum control chart (RA-CUSUM) analysis to identify change points in complete resolution of intestinal metaplasia (CR-IM) and complete resolution of dysplasia (CR-D) outcomes. We compared outcomes between those treated at high-volume (>100 enrolled patients), medium-volume (51-100), and low-volume (<50) centers. RESULTS: There was no association between center volume and CR-IM and CR-D rates, but recurrence rates were lower in high-volume versus low-volume centers (log rank P = .001). There was a significant change point for outcomes at 12 cases for CR-D (reduction from 24.5% to 10.4%; P < .001) and at 18 cases for CR-IM (30.7% to 18.6%; P < .001) from RA-CUSUM curve analysis. CONCLUSION: Our data suggest that 18 supervised cases of endoscopic ablation may be required before competency in endoscopic treatment of Barrett's dysplasia can be achieved. The difference in outcomes between a high-volume and low-volume center does not support further centralization of services to only high-volume centers.

Radiofrequency ablation for patients with refractory symptomatic anaemia secondary to gastric antral vascular ectasia.

Background: Gastric antral vascular ectasia (GAVE) is a rare cause of gastrointestinal bleeding, often causing iron deficiency anaemia. Previous studies have looked at the management of this with argon plasma coagulation, laser therapy and endoscopic band ligation. Methods: This was a single-centre prospective study to evaluate the efficacy and safety of radiofrequency ablation (RFA) in patients with GAVE with persistent anaemia refractory to at least one session of first-line endoscopic therapy. Patients were treated with a through-the-scope (TTS) radiofrequency catheter at two endoscopic sessions six weeks apart. The primary outcome was change in haemoglobin at six months posttreatment. The secondary outcomes were reduction in blood or iron requirements, endoscopic surface area regression and complications. Results: Twenty patients were treated. The mean change in haemoglobin at six months was +12.6 g/l (95% confidence interval 11.7-24.3 g/l), paired t test p < 0.001. At six months, three of 14 individuals who had required blood transfusions had ongoing blood transfusions and five of 17 who had required iron had ongoing iron needs. Surface area regression was scored as 74% ± 25% but no correlation was seen between this and other outcomes. Three of 20 patients experienced pain which was managed with oral analgesia. Of the 14 patients who had reached 12-month follow-up, three required retreatment (21%). Discussion: This small study suggests that RFA is a safe and effective treatment for GAVE. Our study uses the TTS catheter compared to other studies, and demonstrates prolonged improvement in haemoglobin and reduction in blood and iron requirements with a novel assessment of surface area regression.

Machine Learning Creates a Simple Endoscopic Classification System that Improves Dysplasia Detection in Barrett's Oesophagus amongst Non-expert Endoscopists.

INTRODUCTION: Barrett's oesophagus (BE) is a precursor to oesophageal adenocarcinoma (OAC). Endoscopic surveillance is performed to detect dysplasia arising in BE as it is likely to be amenable to curative treatment. At present, there are no guidelines on who should perform surveillance endoscopy in BE. Machine learning (ML) is a branch of artificial intelligence (AI) that generates simple rules, known as decision trees (DTs). We hypothesised that a DT generated from recognised expert endoscopists could be used to improve dysplasia detection in non-expert endoscopists. To our knowledge, ML has never been applied in this manner. METHODS: Video recordings were collected from patients with non-dysplastic (ND-BE) and dysplastic Barrett's oesophagus (D-BE) undergoing high-definition endoscopy with i-Scan enhancement (PENTAX®). A strict protocol was used to record areas of interest after which a corresponding biopsy was taken to confirm the histological diagnosis. In a blinded manner, videos were shown to 3 experts who were asked to interpret them based on their mucosal and microvasculature patterns and presence of nodularity and ulceration as well as overall suspected diagnosis. Data generated were entered into the WEKA package to construct a DT for dysplasia prediction. Non-expert endoscopists (gastroenterology specialist registrars in training with variable experience and undergraduate medical students with no experience) were asked to score these same videos both before and after web-based training using the DT constructed from the expert opinion. Accuracy, sensitivity, and specificity values were calculated before and after training where p < 0.05 was statistically significant. RESULTS: Videos from 40 patients were collected including 12 both before and after acetic acid (ACA) application. Experts' average accuracy for dysplasia prediction was 88%. When experts' answers were entered into a DT, the resultant decision model had a 92% accuracy with a mean sensitivity and specificity of 97% and 88%, respectively. Addition of ACA did not improve dysplasia detection. Untrained medical students tended to have a high sensitivity but poor specificity as they "overcalled" normal areas. Gastroenterology trainees did the opposite with overall low sensitivity but high specificity. Detection improved significantly and accuracy rose in both groups after formal web-based training although it did it reach the accuracy generated by experts. For trainees, sensitivity rose significantly from 71% to 83% with minimal loss of specificity. Specificity rose sharply in students from 31% to 49% with no loss of sensitivity. CONCLUSION: ML is able to define rules learnt from expert opinion. These generate a simple algorithm to accurately predict dysplasia. Once taught to non-experts, the algorithm significantly improves their rate of dysplasia detection. This opens the door to standardised training and assessment of competence for those who perform endoscopy in BE. It may shorten the learning curve and might also be used to compare competence of trainees with recognised experts as part of their accreditation process.

Systematic assessment with I-SCAN magnification endoscopy and acetic acid improves dysplasia detection in patients with Barrett's esophagus.

Background and study aims Enhanced endoscopic imaging with chromoendoscopy may improve dysplasia recognition in patients undergoing assessment of Barrett's esophagus (BE). This may reduce the need for random biopsies to detect more dysplasia. The aim of this study was to assess the effect of magnification endoscopy with I-SCAN (Pentax, Tokyo, Japan) and acetic acid (ACA) on dysplasia detection in BE using a novel mucosal and vascular classification system. Methods BE segments and suspicious lesions were recorded with high definition white-light and magnification endoscopy enhanced using all I-SCAN modes in combination. We created a novel mucosal and vascular classification system based on similar previously validated classifications for narrow-band imaging (NBI). A total of 27 videos were rated before and after ACA application. Following validation, a further 20 patients had their full endoscopies recorded and analyzed to model use of the system to detect dysplasia in a routine clinical scenario. Results The accuracy of the I-SCAN classification system for BE dysplasia improved with I-SCAN magnification from 69 % to 79 % post-ACA (P = 0.01). In the routine clinical scenario model in 20 new patients, accuracy of dysplasia detection increased from 76 % using a "pull-through" alone to 83 % when ACA and magnification endoscopy were combined (P = 0.047). Overall interobserver agreement between experts for dysplasia detection was substantial (0.69). Conclusions A new I-SCAN classification system for BE was validated against similar systems for NBI with similar outcomes. When used in combination with magnification and ACA, the classification detected BE dysplasia in clinical practice with good accuracy.Trials registered at ISRCTN (58235785).

Cost-effectiveness analysis of endoscopic eradication therapy for treatment of high-grade dysplasia in Barrett's esophagus.

AIM: The aim was to evaluate the cost-effectiveness of endoscopic eradication therapy (EET) with combined endoscopic mucosal resection and radiofrequency ablation for the treatment of high-grade dysplasia (HGD) arising in patients with Barrett's esophagus compared with endoscopic surveillance alone in the UK. MATERIALS & METHODS: The cost-effectiveness model consisted of a decision tree and modified Markov model. A lifetime time horizon was adopted with the perspective of the UK healthcare system. RESULTS: The base case analysis estimates that EET for the treatment of HGD is cost-effective at a GB£20,000 cost-effectiveness threshold compared with providing surveillance alone for HGD patients (incremental cost-effectiveness ratio: GB£1272). CONCLUSION: EET is likely to be a cost-effective treatment strategy compared with surveillance alone in patients with HGD arising in Barrett's esophagus in the UK.

Monitoring the premalignant potential of Barrett's oesophagus'.

The landscape for patients with Barrett's oesophagus (BE) has changed significantly in the last decade. Research and new guidelines have helped gastroenterologists to better identify those patients with BE who are particularly at risk of developing oesophageal adenocarcinoma. In parallel, developments in endoscopic image enhancement technology and optical biopsy techniques have improved our ability to detect high-risk lesions. Once these lesions have been identified, the improvements in minimally invasive endoscopic therapies has meant that these patients can potentially be cured of early cancer and high-risk dysplastic lesions without the need for surgery, which still has a significant morbidity and mortality. The importance of reaching an accurate diagnosis of BE remains of paramount importance. More work is needed, however. The vast majority of those undergoing surveillance for their BE do not progress towards cancer and thus undergo a regular invasive procedure, which may impact on their psychological and physical well-being while incurring significant cost to the health service. New work that explores cheaper endoscopic or non-invasive ways to identify the at-risk individual provides exciting avenues for research. In future, the diagnosis and monitoring of patients with BE could move away from hospitals and into primary care.

Comparing outcome of radiofrequency ablation in Barrett's with high grade dysplasia and intramucosal carcinoma: a prospective multicenter UK registry.

BACKGROUND AND STUDY AIM: Mucosal neoplasia arising in Barrett's esophagus can be successfully treated with endoscopic mucosal resection (EMR) followed by radiofrequency ablation (RFA). The aim of the study was to compare clinical outcomes of patients with high grade dysplasia (HGD) or intramucosal cancer (IMC) at baseline from the United Kingdom RFA registry. PATIENTS AND METHODS: Prior to RFA, visible lesions and nodularity were removed entirely by EMR. Thereafter, patients underwent RFA every 3 months until all visible Barrett's mucosa was ablated or cancer developed (end points). Biopsies were taken at 12 months or when end points were reached. RESULTS: A total of 515 patients, 384 with HGD and 131 with IMC, completed treatment. Prior to RFA, EMR was performed for visible lesions more frequently in the IMC cohort than in HGD patients (77 % vs. 47 %; P < 0.0001). The 12-month complete response for dysplasia and intestinal metaplasia were almost identical in the two cohorts (HGD 88 % and 76 %, respectively; IMC 87 % and 75 %, respectively; P = 0.7). Progression to invasive cancer was not significantly different at 12 months (HGD 1.8 %, IMC 3.8 %; P = 0.19). A trend towards slightly worse medium-term durability may be emerging in IMC patients (P = 0.08). In IMC, EMR followed by RFA was definitely associated with superior durability compared with RFA alone (P = 0.01). CONCLUSION: The Registry reports on endoscopic therapy for Barrett's neoplasia, representing real-life outcomes. Patients with IMC were more likely to have visible lesions requiring initial EMR than those with HGD, and may carry a higher risk of cancer progression in the medium term. The data consolidate the approach to ensuring that these patients undergo thorough endoscopic work-up, including EMR prior to RFA when necessary.