Abstract
Background and study aims
Endoscopic and histologic remission (ER, HR) are therapeutic targets in ulcerative colitis (UC) and virtual chromoendoscopy (VCE) improves the endoscopic assessment and the prediction of histology. However, interobserver variability is a limitation for widespread standardised endoscopic assessment using all scoring systems. We aimed to develop an artificial intelligence tool to distinguish ER/activity, and predict histology and risk of flare from white-light-endoscopy (WLE) and VCE videos.
Patients and methods
1090 endoscopic videos (638287 frames), from 283 patients, were used to develop a convolutional neural network (CNN). UC endoscopic activity was graded by experts with Ulcerative Colitis Endoscopic Index of Severity (UCEIS) and PICaSSO. The CNN was trained to distinguish ER/activity on endoscopy videos, and retrained to predict HR/activity, defined according to multiple indices, and predict outcome; CNN and humans agreement was measured.
Results
The AI system detected ER (UCEIS ≤1) in WLE videos with 72% sensitivity (Se), 87% specificity (Sp), and area under the ROC curve (AUROC) of 0.85; For detection of ER in VCE videos (PICaSSO ≤3) Se was 79%, Sp 95%, and the AUROC 0.94. Prediction of HR was similar between WLE and VCE videos (accuracies ranging 80%-85%). The model’s stratification of risk of flare was similar to that of physician-assessed endoscopy scores.
Conclusions
Our system accurately distinguished ER/activity and predicted HR and clinical outcomes from colonoscopy videos. This is the first computer model developed to detect inflammation/healing using VCE through the PICaSSO score and the first computer tool providing endoscopic, histologic, and clinical assessment
Endoscopic and histologic remission (ER, HR) are therapeutic targets in ulcerative colitis (UC) and virtual chromoendoscopy (VCE) improves the endoscopic assessment and the prediction of histology. However, interobserver variability is a limitation for widespread standardised endoscopic assessment using all scoring systems. We aimed to develop an artificial intelligence tool to distinguish ER/activity, and predict histology and risk of flare from white-light-endoscopy (WLE) and VCE videos.
Patients and methods
1090 endoscopic videos (638287 frames), from 283 patients, were used to develop a convolutional neural network (CNN). UC endoscopic activity was graded by experts with Ulcerative Colitis Endoscopic Index of Severity (UCEIS) and PICaSSO. The CNN was trained to distinguish ER/activity on endoscopy videos, and retrained to predict HR/activity, defined according to multiple indices, and predict outcome; CNN and humans agreement was measured.
Results
The AI system detected ER (UCEIS ≤1) in WLE videos with 72% sensitivity (Se), 87% specificity (Sp), and area under the ROC curve (AUROC) of 0.85; For detection of ER in VCE videos (PICaSSO ≤3) Se was 79%, Sp 95%, and the AUROC 0.94. Prediction of HR was similar between WLE and VCE videos (accuracies ranging 80%-85%). The model’s stratification of risk of flare was similar to that of physician-assessed endoscopy scores.
Conclusions
Our system accurately distinguished ER/activity and predicted HR and clinical outcomes from colonoscopy videos. This is the first computer model developed to detect inflammation/healing using VCE through the PICaSSO score and the first computer tool providing endoscopic, histologic, and clinical assessment
| Original language | English |
|---|---|
| Pages (from-to) | 332-341 |
| Number of pages | 10 |
| Journal | Endoscopy |
| Volume | 55 |
| Issue number | 4 |
| DOIs | |
| Publication status | Published - 13 Oct 2022 |