Optical Diagnosis (OD) based strategy (PIVI criteria) in the management of colorectal polyps has been a revelation in modern endoscopic therapy, however, studies have shown that OD in non- expert hands have not meet PIVI criteria, for both ' diagnose and leave' and 'resect and discard' strategies. We aim to create a simplified optical strategy which could accurately identify covert cancer as well as reduce the number of pathological examinations based on prevalence of cancer in 6–10 mm polyps.
Methods
We analysed outcomes of all patients who underwent screening colonoscopy between January 2007 to December 2018 and were found to have polyps. Data was prospectively collected on an online endoscopy reporting system and pathology reporting system. Statistical analysis was performed using multinomial logistic regression.
Results
A total of 15906 polyps were removed at colonoscopy over the specified period. Mean size was 7.3 mm (range: 1 to 120 mm). 86.6% of all polyps were non pedunculated and 56.3% polyps were located in the left colon. The size, site, morphology and histology of these polyps is shown in table 1. A histopathological diagnosis of polyp cancer was made in 104/15906 polyps (0.65%). 94/104 polyp cancers (90.25%) were associated with non pedunculated morphology [OR 1.45, 95%CI 0.75–2.78 p=0.005]. Risk of developing in cancer in polyps ≥20 mm was significantly higher than in smaller polyps [ OR 6.57 95% CI 5.7- 13.1 p< 0.001 ]. 89 cancers were found in the left colon and rectum compared with 15 cancers in the right colon ( 85.5% vs 14.5%) [OR 1.98, 95%CI 0.9–3.1 p=0.007].
Conclusion
This is the largest report of the prevalence of cancer in colorectal lesions 6–10 mm in size. We have demonstrated that the prevalence of covert cancer in colorectal lesions <5 mm is negligible and that of polyps 6–10 mm is very low (0.17%). All these cancers were in non-pedunculated adenomas in left colon. Based on the data, we have demonstrated in the 6–10 mm polyp subgroup, we suggest a modified ' resect and discard' concept ( based on OD AND location based strategy) extending to 6–10 mm polyps in the right colon. Given the fact, that most non experts fail to reach PIVI criteria based on OD alone, this modified strategy would reduce the need for optical assessment and increase the scope of 'resect and discard' to a larger number of polyps.
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.
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.
Training in gastrointestinal endoscopy in the UK occurs predominantly in a real world one-to-one trainer to trainee interaction. Previous surveys have shown surgical and gastroenterology trainees have had mixed experiences of supervision and training, and no surveys have explored specifically the role of trainee to trainer feedback. This study aimed to explore the experience of training and of providing trainer feedback for all disciplines of endoscopy trainees.An online survey designed in collaboration with Joint Advisory Committee training committee and trainee representatives was distributed from January 2020 but was interrupted by the COVID-19 pandemic and hence terminated early.There were 129 responses, including trainees from all disciplines and regions, of which 86/129 (66.7%) rated the culture in their endoscopy units favourably-either good or excellent. 65/129 (50.4%) trainees reported having one or more training lists allocated per week, with 41/129 (31.8%) reporting only ad hoc lists. 100/129 (77.5%) respondents were given feedback and 97/129 (75.2%) were provided with learning points from the list. 65/129 (50.4%) respondents reported their trainer completed a direct observation of procedure or direct observation of polypectomies. 73/129 (56.6%) respondents reported that they felt able to give feedback to their trainer, with 88/129 (68.2%) feeling they could do this accurately. Barriers to trainer feedback cited included time constraints, lack of anonymity and concerns about affecting the trainer-trainee relationship.Overall, the training environment has improved since previous surveys. There are still issues around interdisciplinary differences with some surgical trainees finding the training environment less welcoming, and trainee perceptions of hierarchical barriers and trainer responsiveness to feedback limiting the accuracy of their feedback.
Abstract Background Big data has the potential to revolutionize echocardiography by enabling novel research and rigorous, scalable quality improvement. Text reports are a critical part of such analyses, and ontology is a key strategy for promoting interoperability of heterogeneous data through consistent tagging. Currently, echocardiogram reports include both structured and free text and vary across institutions, hampering attempts to mine text for useful insights. Natural language processing (NLP) can help and includes both non-deep learning and deep-learning (e.g., large language model, or LLM) based techniques. Challenges to date in using echo text with LLMs include small corpus size, domain-specific language, and high need for accuracy and clinical meaning in model results. Methods We tested whether we could map echocardiography text to a structured, three-level hierarchical ontology using NLP. We used two methods: statistical machine learning (EchoMap) and one-shot inference using the Generative Pre-trained Transformer (GPT) large language model. We tested against eight datasets from 24 different institutions and compared both methods against clinician-scored ground truth. Results Despite all adhering to clinical guidelines, there were notable differences by institution in what information was included in data dictionaries for structured reporting. EchoMap performed best in mapping test set sentences to the ontology, with validation accuracy of 98% for the first level of the ontology, 93% for the first and second level, and 79% for the first, second, and third levels. EchoMap retained good performance across external test datasets and displayed the ability to extrapolate to examples not initially included in training. EchoMap’s accuracy was comparable to one-shot GPT at the first level of the ontology and outperformed GPT at second and third levels. Conclusions We show that statistical machine learning can achieve good performance on text mapping tasks and may be especially useful for small, specialized text datasets. Furthermore, this work highlights the utility of a high-resolution, standardized cardiac ontology to harmonize reports across institutions.
Benign colonic polyps are traditionally resected via endoscopic mucosal resection (EMR). This technique is safe but carries risk of recurrence of 10 – 15% in polyps larger than 20 mm. Post-EMR recurrences are often scarred making endoscopic management of these polyps challenging with a high risk of complications. EndoRotor is a novel non-diathermic EMR device designed to reduce diathermy related complications (eg. perforation and delayed bleeding). We present a video demonstrating the use of this device in the management of scarred polyps.
Real-time in-vivo characterisation of colorectal polyps remains limited outside expert centres. Data on AI polyp detection and characterisation is promising but accurate sizing remains the missing jigsaw piece. We aimed to study the impact of a novel AI system on non-expert endoscopists detection, characterisation and sizing of colorectal polyps compared to experts.
Methods
Prospectively collected endoscopy videos from twelve centres in Europe and Japan were uploaded on a bespoke online platform. All polyps were histologically proven and sized by three experts. The AI model detects polyps and classifies them as neoplastic/non-neoplastic and diminutive/non-diminutive. We asked Six experts to detect, characterise and size polyps without AI support, and Six non-experts to detect polyps assisted by AI, and to characterise and size polyps without and then with AI.
Results
199 videos (100-polyps) were included. On polyp detection, average sensitivity and specificity of non-experts +AI compared to experts was 96.0% and 84.6% compared to 95.7% and 89.9% respectively (p>0.5). Non-experts+AI showed superior sensitivity (95.5% vs 83.3%) and NPV (90.8% vs 70.4%) of characterisation on enhanced imaging compared to non-experts alone (p<0.5). On sizing, non-experts+AI achieved accuracy and sensitivity of 84.0% and 93.6%, respectively. Experts' characterisation and sizing metrics were not significantly different from non-experts+AI.
Conclusion
This interim analysis suggests our AI system may support non-experts to perform at experts' level and achieve PIVI-2 threshold (diagnose and leave). Further analysis is underway to understand the impact of the AI system on surveillance interval (PIVI-1). To our knowledge, this is the first report incorporating AI-assisted sizing with detection and characterisation.
Peptic ulcers are the commonest cause of upper Gastrointestinal bleeding (UGIB). Hemospray (Cook Medical, North Carolina, USA) is a novel haemostatic powder aimed to treat UGIB. The aim of this study is to look at outcomes in patients with peptic ulcer GI bleeds treated with hemospray in 13 centres.
Methods
Data was prospectively collected on hemospray use in UGIBs in the UK, France and Germany (Jan'16-Sept'18). Hemospray was used for peptic ulcer UGIBs as a monotherapy, dual-therapy with standard haemostatic techniques or rescue therapy. Haemostasis was defined as cessation of bleeding within 5 minutes of hemospray application.
Results
196 patients with UGIBs secondary to peptic ulcers were recruited (133 M, 63 F, 123/196(63%) duodenal, 44/196(22%) gastric, 29/196(15%) oesophageal). Immediate haemostasis was achieved in 171/196(87%) patients. The median rockall score was 7 (IQR, 6–8). Rebleeding rates were significantly lower in forrest 2a relative to current predicted rebleeding rates based on forrest classifications, 2/21(10%, P<0.005). In the 25/196(13%) patients who did not achieve haemostasis 18/25 (72%) were Forrest 1b ulcers. In the total cohort, 33/169(20%) had a rebleed, median rockall score was 7(IQR,7–8). Outcomes with different Forrest classifications (table 1)
Conclusions
Hemospray is effective in achieving immediate haemostasis in peptic ulcer UGIBs. The baseline Blatchford/rockall scores in our cohort are high with patients recruited from tertiary centres with high-risk cases. The rebleeding and mortality rates are in keeping/below the predicted rate based on the scores. The best outcome with hemospray was with forrest 2a ulcers.
Bleeding is a well recognised complication of endoscopic resection (ER), particularly in endoscopic submucosal dissection (ESD). Electrocautery can be used to control bleeding but does increase the risk of perforation. A novel extracellular scaffold matrix (Purastat) has recently been approved for gastrointestinal haemostasis. This self-assembling peptide forms a transparent gel that can be applied via a catheter through the scope over the bleeding area. We conducted a feasibility study in a high bleeding risk cohort to assess its applicability, safety and efficacy. We also aimed to ascertain the mean volume of Purastat required to cover the resection base prophylactically.
Method
This was a prospective observational cohort study of patients undergoing complex ER in a tertiary referral centre from December 2015–2016. Purastat was used for prophylaxis over the resection base in high bleeding risk procedures or for primary haemostasis in active bleeding. Data was collected on patient and lesion characteristics including surface area, technical feasibility of gel application, haemostasis and delayed bleeding rate.
Results
Purastat was used in 74 patients (average age 69 years, male to female ratio of 2:1). All lesions were >2 cm and 33.8% (25/74) had cardiac co-morbidities with anticoagulant or antiplatelet usage reflecting a high bleeding risk. 60 (81.1%) had ESD and 14 (18.9%) had endoscopic mucosal resection. Table 1 shows the distribution of lesions according to location and size. Abstract PTH-019 Table 1 Abstract PTH-019 Table 2
Purastat on its own was effective in stopping bleeding in 35/48 (72.9%) cases (see Table 2). It was successfully applied in all patients with no interference in visibility or catheter blockage
The mean surface area of the resection base was 16.2cm2 requiring a mean Purastat® volume of 2.7mls, or 0.2mls/cm2. On follow up in 1 month, delayed bleeding was noted in 3/74 (4%) patients. All were managed with endoscopic intervention and no transfusion was required.
Conclusion
Purastat was effective in controlling bleeding in almost ¾ of the cases where it was used for primary haemostasis. It is safe, easy to use and does not hamper ongoing ER. Only a small amount is needed to effectively cover the resection base for prophylaxis. Our data has demonstrated its potential as a novel haemostatic agent that can minimise bleeding during ER.
Disclosure of Interest
S. Subramaniam: None Declared, K Kandiah: None Declared, S Thayalasekaran: None Declared, G Longcroft-Wheaton: None Declared, P Bhandari Conflict with: Receives educational grants from Fujifilm, Olympus and Pentax
Aims ESD is a minimally invasive therapeutic option for early oesophageal neoplasia, however is not without risk. In Europe, the complication profile is most established for Barrett’s neoplasia, being the predominant pathology, and stricture risk has been shown to be related to lesion circumference. Our aim was to compare the safety of ESD between Barrett’s and squamous neoplasia in a Western population.
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
'/%3e%3cpath id='l' d='M-26.25 0h52.5v-12h-40.5v-16h33v-12h-33v-11H25v-12h-51.25z'/%3e%3cpath id='k' d='M-31.5 0h12v-48l14 48h11l14-48V0h12v-70H14L0-22l-14-48h-17.5z'/%3e%3cpath id='b' fill-rule='evenodd' d='M0 0a31.5 35 0 0 0 0-70A31.5 35 0 0 0 0 0m0-13a18.5 22 0 0 0 0-44 18.5 22 0 0 0 0 44'/%3e%3cpath id='d' fill-rule='evenodd' d='M-31.5 0h13v-26h28a22 22 0 0 0 0-44h-40zm13-39h27a9 9 0 0 0 0-18h-27z'/%3e%3cpath id='n' d='M-15.75-22C-15.75-15-9-11.5 1-11.5s14.74-3.25 14.75-7.75c0-14.25-46.75-5.25-46.5-30.25C-30.5-71-6-70 3-70s26 4 25.75 21.25H13.5c0-7.5-7-10.25-15-10.25-7.75 0-13.25 1.25-13.25 8.5-.25 11.75 46.25 4 46.25 28.75C31.5-3.5 13.5 0 0 0c-11.5 0-31.55-4.5-31.5-22z'/%3e%3cuse xlink:href='%23a' id='o' transform='scale(31.5)'/%3e%3cuse xlink:href='%23a' id='p' transform='scale(26.25)'/%3e%3cuse xlink:href='%23a' id='r' transform='scale(21)'/%3e%3cuse xlink:href='%23a' id='q' transform='scale(15)'/%3e%3cuse xlink:href='%23a' id='s' transform='scale(10.5)'/%3e%3cg id='m'%3e%3cclipPath id='c'%3e%3cpath d='M-31.5 0v-70h63V0zM0-47v12h31.5v-12z'/%3e%3c/clipPath%3e%3cuse xlink:href='%23b' clip-path='url(%23c)'/%3e%3cpath d='M5-35h26.5v10H5z'/%3e%3cpath d='M21.5-35h10V0h-10z'/%3e%3c/g%3e%3cg id='h'%3e%3cuse xlink:href='%23d'/%3e%3cpath d='M28 0c0-10 0-32-15-32H-6c22 0 22 22 22 32'/%3e%3c/g%3e%3cg id='a' fill='%23fff'%3e%3cg id='f'%3e%3cpath id='e' d='M0-1v1h.5' transform='rotate(18 0 -1)'/%3e%3cuse xlink:href='%23e' transform='scale(-1 1)'/%3e%3c/g%3e%3cuse xlink:href='%23f' transform='rotate(72)'/%3e%3cuse xlink:href='%23f' transform='rotate(-72)'/%3e%3cuse xlink:href='%23f' transform='rotate(144)'/%3e%3cuse xlink:href='%23f' transform='rotate(216)'/%3e%3c/g%3e%3c/defs%3e%3crect width='100%25' height='100%25' x='-50%25' y='-50%25' fill='%23009b3a'/%3e%3cpath fill='%23fedf00' d='M-1743 0 0 1113 1743 0 0-1113z'/%3e%3ccircle r='735' fill='%23002776'/%3e%3cclipPath id='g'%3e%3ccircle r='735'/%3e%3c/clipPath%3e%3cpath fill='%23fff' d='M-2205 1470a1785 1785 0 0 1 3570 0h-105a1680 1680 0 1 0-3360 0z' clip-path='url(%23g)'/%3e%3cg fill='%23009b3a' transform='translate(-420 1470)'%3e%3cuse xlink:href='%23b' y='-1697.5' transform='rotate(-7)'/%3e%3cuse xlink:href='%23h' y='-1697.5' transform='rotate(-4)'/%3e%3cuse xlink:href='%23i' y='-1697.5' transform='rotate(-1)'/%3e%3cuse xlink:href='%23j' y='-1697.5' transform='rotate(2)'/%3e%3cuse xlink:href='%23k' y='-1697.5' transform='rotate(5)'/%3e%3cuse xlink:href='%23l' y='-1697.5' transform='rotate(9.75)'/%3e%3cuse xlink:href='%23d' y='-1697.5' transform='rotate(14.5)'/%3e%3cuse xlink:href='%23h' y='-1697.5' transform='rotate(17.5)'/%3e%3cuse xlink:href='%23b' y='-1697.5' transform='rotate(20.5)'/%3e%3cuse xlink:href='%23m' y='-1697.5' transform='rotate(23.5)'/%3e%3cuse xlink:href='%23h' y='-1697.5' transform='rotate(26.5)'/%3e%3cuse xlink:href='%23j' y='-1697.5' transform='rotate(29.5)'/%3e%3cuse xlink:href='%23n' y='-1697.5' transform='rotate(32.5)'/%3e%3cuse xlink:href='%23n' y='-1697.5' transform='rotate(35.5)'/%3e%3cuse xlink:href='%23b' y='-1697.5' transform='rotate(38.5)'/%3e%3c/g%3e%3cuse xlink:href='%23o' x='-600' y='-132'/%3e%3cuse xlink:href='%23o' x='-535' y='177'/%3e%3cuse xlink:href='%23p' x='-625' y='243'/%3e%3cuse xlink:href='%23q' x='-463' y='132'/%3e%3cuse xlink:href='%23p' x='-382' y='250'/%3e%3cuse xlink:href='%23r' x='-404' y='323'/%3e%3cuse xlink:href='%23o' x='228' y='-228'/%3e%3cuse xlink:href='%23o' x='515' y='258'/%3e%3cuse xlink:href='%23r' x='617' y='265'/%3e%3cuse xlink:href='%23p' x='545' y='323'/%3e%3cuse xlink:href='%23p' x='368' y='477'/%3e%3cuse xlink:href='%23r' x='367' y='551'/%3e%3cuse xlink:href='%23r' x='441' y='419'/%3e%3cuse xlink:href='%23p' x='500' y='382'/%3e%3cuse xlink:href='%23r' x='365' y='405'/%3e%3cuse xlink:href='%23p' x='-280' y='30'/%3e%3cuse xlink:href='%23r' x='200' y='-37'/%3e%3cuse xlink:href='%23o' y='330'/%3e%3cuse xlink:href='%23p' x='85' y='184'/%3e%3cuse xlink:href='%23p' y='118'/%3e%3cuse xlink:href='%23r' x='-74' y='184'/%3e%3cuse xlink:href='%23q' x='-37' y='235'/%3e%3cuse xlink:href='%23p' x='220' y='495'/%3e%3cuse xlink:href='%23r' x='283' y='430'/%3e%3cuse xlink:href='%23r' x='162' y='412'/%3e%3cuse xlink:href='%23o' x='-295' y='390'/%3e%3cuse xlink:href='%23s' y='575'/%3e%3c/svg%3e)