VALIDATION OF A NEW OPTICAL DIAGNOSIS TRAINING PLATFORM TO IMPROVE DYSPLASIA CHARACTERISATION IN INFLAMMATORY BOWEL DISEASE (OPTIC-IBD): A MULTICENTRE RANDOMISED CONTROLLED STUDY
Bibliographic record
Abstract
Aims To develop and validate online training in optical diagnosis of dysplastic lesions in IBD (NCT04924543). Methods We designed an interactive, self-directed, multi-modality learning module (includes optical diagnosis methods, classification systems, examples and self-assessments). We invited participants from Canada, Italy and the UK. Assessments comprised short endoscopic videos, divided into 8 non-dysplastic (hyperplastic, inflammatory, SSL) and 16 dysplastic IBD colonic lesions (SSL-D, low/high-grade dysplasia, cancer). Participants classified lesions, predicted histology and rated their confidence, then completed online training. The videos were repeated in a random order after≥7-14d. Participants were then randomized 1:1 to feedback and extra training, with a final assessment after 60d. We report diagnostic performance for dysplasia, interrater reliability and rater confidence. Results We present a planned interim analysis of 77 participants ([ Table 1 ]). Diagnostic accuracy for dyplasia improved (primary endpoint: 44.5 to 54.0%, P<0.0001), particularly for novice and intermediate endoscopists (data not shown). In multilevel logistic regression, training was associated with correct diagnoses for high confidence (OR 1.40, 1.13-1.77) but not low confidence ratings (OR 1.09, 0.96-1.25). Consistency between participants improved from slight to fair (Fleiss’ κ 0.16 to 0.24, P=0.015), proportionate with experience. Training increased participants’ confidence to correctly identify dysplasia (high confidence 25% to 46%, P<0.0001). Table 1 Diagnostic Performance for Dysplasia Overall 77 44.5 vs. 54.0, P <0.0001 (43.2-46.2, 52.4-55.6) 50.3 vs. 59.1 (46.8-56.2, 54.5-63.2) 33.0 vs. 43.8 (28.9-37.7, 39.1-48.5) Participant confidence in prediction: Low and High Pre: 18 (13-23) Post: 13 (7-20) 40.4 vs. 46.1, P =0.0455 (38.8-42.3, 44.3-48.1) 46.5 vs. 45.1 (42.2-53.9, 39.4-52.1) 28.5 vs. 48.3 (23.8-34.4, 42.5-54.8) Pre: 6 (1-11) Post: 11 (4-17) 55.0 vs. 64.6, P =0.0574 (52.2-57.8, 61.3-66.5) 59.7 vs. 61.2 (51.3-67.3, 51.4-66.2) 44.9 vs. 70.3 (36.0-53.3, 61.8-75.6) Participant endoscopic experience Novice (<100 lifetime colonoscopies) : 30 Novice : 35.0 vs. 45.8, P =0.0003 (31.7-42.7, 43.7-48.4) Novice : 39.6 vs. 51.7 (34.6-51.0, 45.3-60.2) Novice : 25.8 vs. 34.2 (20.2-34.8, 28.0-42.0) Intermediate : 25 Intermediate : 43.3 vs. 55.7, P =0.0004 (39.1-49.9, 52.7-58.3) Intermediate : 49.8 vs. 59.3 (43.2-59.3, 51.1-66.1) Intermediate : 30.5 vs. 48.5 (24.0-39.8, 39.8-55.9) Experienced (>=1000): 22 Experienced : 58.9 vs. 63.3, P =0.0703 (53.0-63.2, 59.9-65.8) Experienced : 65.6 vs. 69.0 (57.1-72.4, 58.6-75.2) Experienced : 45.5 vs. 51.7 (36.9-53.9, 42.0-59.1) Conclusions The OPTIC-IBD training module improved participants’ accuracy, precision and confidence in optical diagnosis of dysplasia. Next, we will study the training approaches and classification systems that can best be adopted by non-experts and trainees. Our refined training platform will be made available to improve quality of care for people with IBD. Publication History Article published online: 14 April 2022 © 2022. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.009 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.003 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.009 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".