Introduction

Capsule endoscopy (CE) is a valuable tool for assessing inflammation in patients with Crohn's disease (CD). The current standard for evaluating inflammation are validated scores like Lewis score (LS), Capsule Endoscopy Crohn’s Disease Activity Index (CECDAI) and ELIAKIM score and laboratory values. Recent advances in artificial intelligence (AI) have made it possible to automatically select the most relevant frames in capsule endoscopy.

Aims & Methods

In this study, our objective was to develop an automated scoring system using CE images to objectively grade inflammation. Pan-enteric CE videos (PillCam Crohn's) performed in CD patients between 09/2020 and 01/2023 were retrospectively reviewed and LS, CECDAI and ELIAKIM scores calculated. We developed a convolutional neural network based automated score consisting in the percentage of positive frames selected by the algorithm (for small bowel and colon separately). We correlated clinical data and the validated scores with the artificial intelligence generated score (AIS).

Results

A total of 61 patients were included. The median SL was 225 [0-6,06], CECDAI was 6 [0-33], ELIAKIM was 4 [0-38] and SB_AIS was 0.5659 [0-29.45]. We found a strong correlation between SB_AIS and LS, CECDAI and ELIAKIM scores (Pearson’s r= 0.751, r=0.707, r= 0.655, p = 0.001). We found a strong correlation between SL and ELIAKIM (r= 0,768, p = 0.001) and very strong correlation between CECDAI and SL scores (r= 0,854, p = 0.001) and CECDAI and ELIAKIM (r= 0,827, p = 0.001).

Conclusion

Our study showed that the AI-generated score had a strong correlation with validated scores, indicating that it could serve as an objective and efficient method for evaluating inflammation in CD patients. As a pre-proof study, our findings provide a promising basis for future reffining a CE score that can accurately correlate with prognostic factors and aid in the management and treatment of CD patients.

Introduction

Capsule endoscopy (CE) is a minimally invasive exam suitable of panendoscopic evaluation of the gastrointestinal tract. Nevertheless, CE is time-consuming, with suboptimal diagnostic yield in the upper GI tract. Convolutional neural networks (CNN) are human brain architecture-based models suitable for image analysis. However, there is no study about their role in capsule panendoscopy.

Aims & Methods

Our aim was to develop an artificial intelligence (AI) model for panendoscopic automatic detection of pleomorphic lesions (namely vascular lesions, protuberant lesions, hematic residues, ulcers, and erosions). 355110 images (6977 esophageal, 12918 gastric, 258443 small bowel, 76772 colonic) from eight different CE and colon CE (CCE) devices were divided into a training and validation dataset in a patient split design. The model classification was compared to three CE experts’ classifications. The model’s performance was evaluated by its sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and area under the precision-recall curve.

Results

The binary esophagus CNN had a diagnostic accuracy for pleomorphic lesions of 83.6%. The binary gastric CNN identified pleomorphic lesions with a 96.6% accuracy. The undenary small bowel CNN distinguished pleomorphic lesions with different hemorrhagic potential with 97.6% accuracy. The trinary colonic CNN (detection and differentiation of normal mucosa, pleomorphic lesions, and hematic residues) had 94.9% global accuracy.

Conclusion

Our group developed the first AI model for panendoscopic automatic detection of pleomorphic lesions in both CE and colon CE from multiple brands, solving a critical interoperability technological challenge. Deep learning-based tools may change the landscape of minimally invasive capsule panendoscopy.

References

1. Pennazio M, Rondonotti E, Despott EJ, Dray X, Keuchel M, Moreels T, et al. Small-bowel capsule endoscopy and device-assisted enteroscopy for diagnosis and treatment of small-bowel disorders: European Society of Gastrointestinal Endoscopy (ESGE) Guideline - Update 2022. Endoscopy. 2023 Jan;55(1):58-95. PubMed PMID: 36423618. Epub 20221124.
2. Le Berre C, Trang-Poisson C, Bourreille A. Small bowel capsule endoscopy and treat-to-target in Crohn's disease: A systematic review. World J Gastroenterol. 2019 Aug 21;25(31):4534-54. PubMed PMID: 31496630. PMCID: PMC6710184.
3. Iddan G, Meron G, Glukhovsky A, Swain P. Wireless capsule endoscopy. Nature. 2000 May 25;405(6785):417. PubMed PMID: 10839527.
4. Rondonotti E, Pennazio M. Colon capsule for panendoscopy: a narrow window of opportunity. Endosc Int Open. 2021 Dec;9(12):E1860-E2. PubMed PMID: 34917450. PMCID: PMC8670997. Epub 20211214.
5. Vuik FER, Moen S, Spaander MCW. Colon capsule endoscopy as panendoscopy: Using current knowledge to enhance possibilities. Endosc Int Open. 2022 May;10(5):E584. PubMed PMID: 35571484. PMCID: PMC9106432. Epub 20220513.

Introduction

Crohn's disease (CD) has a segmental distribution and possible involvement of the entire gastrointestinal (GI) tract. Pan-enteric capsule endoscopy (CE) is shown to be a highly sensitive tool for detecting pathology in patients with suspected Crohn's disease. A disadvantage of CE is the time-consuming analysis of the images. Artificial intelligence (AI) algorithms might offer a possibility to assist in the review and reduce the analysis time of CE.

Aims & Methods

This study examines the agreement between pan-enteric CE analyzed by specialists utilizing an AI algorithm to regular CE readings in patients with suspected CD.
Patients and methods: Pan-enteric CE’s from a prospective blinded, multicenter study, including patients suspected of CD (ClinicalTrials.Gov NCT03134586)[1], were anonymized, and videos from the first duodenal- to the last rectal picture were generated using Pillcam Rapid reader v.9 (Medtronic, Dublin, Ireland). Videos were processed by the deep learning solution AXARO® (Augmented Endoscopy, Paris, France), and two GI specialists reviewed images of interest. Based on the image output, the observers classified the patient’s CE as suggestive of CD, ulcerative colitis (UC), cancer, or as having a normal CE. Observers were blinded to patient histories, the results of the complete CE analysis, and the ileocolonoscopy. Previous GI-specialist evaluation of the CE, following standard procedure in the Rapid reader software v9, served as the reference standard. The time used for the AI review was recorded. Per-patient sensitivities, specificities, and positive and negative predictive values for detecting CD and Inflammatory bowel disease (IBD) (CD+UC) were calculated for the two observers, and Cohen’s kappa assessed interobserver agreement. Following the blinded assessment, all disagreement was reviewed.

Results

Pan-enteric CEs from 131 patients were included in this analysis (95 women, 73%) with a median age of 28 years (range 16-72). CE was complete in 109 (83.0%) patients, and the mean CE recording time was 374 minutes.
The AXARO® framework reduced the image output to a median of 2.1%, corresponding to 470 images per patient, and the two observers had a pooled mean review time of 208 seconds per patient. The following results were obtained for Observer 1 and Observer 2, respectively, a sensitivity for detecting CD of 96% and 92% and a specificity of 93% and 90%. IBD was detected with a 97% and 97% sensitivity and specificity of 91% and 90%. The negative predictive value in this cohort was 95% for CD and 97% for IBD. The interobserver agreement was very high, with a kappa of 0.91 (CI 0.82-0.97) and 0.92 (CI 0.85-0.98) for CD and IBD, respectively. A disagreement review showed challenges with very mild disease and correctly classifying inflammation. The number of false positives was mainly due to non-specific inflammation, which did not meet the diagnostic criteria for CD in the reference standard evaluation. Two cancers were correctly detected, but a single adenomatous polyp was not detected.

Conclusion

Using the AXARO® framework to select pictures of interest reduced the initial review time to a median of less the four minutes while maintaining high diagnostic accuracy. The interobserver agreement was very high, and disagreement occurred in very mild cases or classifications of inflammation. The high sensitivity and negative predictive value show that this AI algorithm could be used as a rapid tool to ”rule out” IBD in pan-enteric CEs of newly referred patients suspected of Crohn’s disease.

References

1. Brodersen JB, Knudsen T, Kjeldsen J et al. Diagnostic accuracy of pan-enteric capsule endoscopy and magnetic resonance enterocolonography in suspected Crohn's disease. United European gastroenterology journal 2022. DOI: 10.1002/ueg2.12307

Disclosure

JBB, MDJ, JK, and TK have nothing to disclose
RL: Co-founder and shareholder of Augmented Endoscopy
AH: Co-founder and shareholder of Augmented Endoscopy
XD: Co-founder and shareholder of Augmented Endoscopy; Consultancy for Norgine and Provepharma; Lecture fees for Alfasigma, Bouchara Recordati, Medtronic, Norgine, Sandoz, Fujifilm

Introduction

Intestinal ultrasound (IUS) is effective for monitoring treatment efficacy in Crohn's disease (CD). Bowel wall thickness (BWT) is used as a surrogate of inflammatory activity. However, the added value of other IUS parameters ((bowel wall flow (BWF), stratification (BWS), lymph node enlargement, mesenteric fat hypertrophy, and complications)) for monitoring and predicting treatment failure during anti-TNFs’ maintenance therapy is still poorly defined and was the goal of this study.

Aims & Methods

This was a retrospective cross-sectional study. The study population included patients treated on infliximab or adalimumab maintenance treatment who had undergone an index intestinal ultrasound examination. We examined the added value of various ultrasonographic activity signs, in addition to bowel wall thickness, for their association with subsequent treatment failure. Treatment failure was defined as the need for CD-related surgery, hospitalization, corticosteroid use, dose escalation, and drug discontinuation.

Results

The study included 103 CD patients treated with anti-TNF therapy (67% adalimumab and 33% infliximab) who underwent an IUS at a median of 10 months (IQR 6-18) after starting therapy. Of 103 patients, 59.2% experienced treatment failure at a median of 5 months (2-12.5) after the index IUS examination. Among the failure group, bowel thickness at preceding index IUS was significantly higher compared with patients without failure (3.9 vs 3mm, respectively, P= 0.007), BWF was higher (49.2% vs. 23.8%, P= 0.009), and mesenteric hypertrophy was more prevalent (20.4% vs 7.1%, P=0.006). Logistic regression identified increased BWT and mesenteric hypertrophy as having a significant association with therapy failure. A sensitivity analysis with the outcome of therapy discontinuation revealed that mesenteric fat hypertrophy was a significant predictive factor along with increased BWT (Table 1).

Table 1: Sonographic Characteristics Associated with Treatment Failure and Drug Discontinuation.

Conclusion

Our study highlights the significance of BWT, BWF, and mesenteric fat hypertrophy as key sonographic markers associated with anti-TNF treatment failure in Crohn's disease. Integrating these indicators into scoring systems may enhance predictive accuracy, providing a quantitative tool for evaluating disease severity and treatment response.