Introduction
The achievement of histological remission (HR) is a key treatment goal in ulcerative colitis (UC), as it is suggestive of better disease management and predicts response to therapy.
The absence of neutrophils in epithelium and lamina is crucial to define HR. However, the role of neutrophil number and localisation remains unclear. We aimed to develop a novel artificial intelligence (AI)-driven system to automate neutrophil detection, localisation, and quantification, supporting assessment of HR and enable the prediction of response to therapy.
Aims & Methods
We developed an AI-driven system by integrating two deep learning models to (1) segment whole slide images (WSIs) into epithelium, crypts and lamina propria and to (2) detect and quantify neutrophils. Outputs from these models were combined to compute neutrophil densities in the lamina propria, epithelium, and overall tissue. The optimal neutrophil density cut-offs to assess disease activity (Geboes score >2B.0) in UC patients were identified in the AMAC phase 2 Mirikizumab trial, and subsequently validated in the multicentre prospective PICaSSO cohort study. Cut-offs to determine treatment response at weeks 12 and 52 (histological improvement [HI; Geboes score ≤3.1] and histological remission [HR; Geboes score ≤2B.0]), were also evaluated.
Results
Table 1 shows the system diagnostic performance in assessing disease activity and predicting response to therapy at W12 and W52 in the Mirikizumab cohort.
303 WSIs from active UC patients from the multicentre, randomised, double-blind, parallel-arm, placebo-controlled phase 2 clinical trial of Mirikizumab were analysed. The model yielded a DICE SØrensen of 0.57 for segmentation, precision of 82.3% and recall of 77.0% for neutrophil detection. The system identified an overall optimal neutrophil density cut-off (cells/mm²) predictive of disease activity of >21.7.0 (>10.1 and >30.4 for epithelium and lamina propria, respectively), with an accuracy of 86% (85% and 85%, respectively). Further validation in the multicenter PICaSSO cohort yielded to similar results.
Moreover, the overall optimal cut-off to predict histological remission at W52 was <20.8 (<10.4 and <27.9 for epithelium and lamina propria, respectively), with an accuracy of 82.0% (81% and 80%, respectively).
Conclusion
Our novel AI-based system demonstrates strong potential as an automated tool to detect, localise, and quantify neutrophils to assess histological activity and predict response to therapy in a clinical trial UC cohort and real-world settings. This framework could be applied in clinical trials and routine clinical practice, offering objective guidance for personalised management in UC patients.
| Neutrophil localisation | Cut-off (cells/mm2) | Sensitivity (%) | Specificity (%) | Accuracy (%) |
| Assessment of disease activity |
| Epithelium | 10.1 | 90 (83-96) | 75 (56-94) | 85 (78-91) |
| Lamina propria | 30.4 | 93 (87-98) | 72 (55-90) | 85 (79-91) |
| Total | 21.7 | 94 (88-99) | 74 (54-89) | 86 (79-92) |
| Prediction of response to therapy - Histological Remission |
| | W12 | W52 | W12 | W52 | W12 | W52 | W12 | W52 |
| Epithelium | 10.9 | 10.4 | 73 (41-94) | 75 (52-95) | 82 (67-95) | 90 (77-100) | 78 (64-90) | 81 (67-94) |
| Lamina propria | 33 | 27.9 | 72 (41-95) | 74 (50-95) | 84 (67-96) | 89 (71-100) | 78 (65-89) | 80 (67-91) |
| Total | 26.8 | 20.8 | 74 (47-95) | 74 (48-94) | 84 (70-96) | 93 (80-100) | 79 (66-91) | 82 (68-94) |
References
[1] Santacroce G, Meseguer P, Zammarchi I, et al. P406 A novel active learning-based digital pathology protocol annotation for histologic assessment in Ulcerative Colitis using PICaSSO Histologic Remission Index (PHRI). Journal of Crohn's and Colitis, Volume 18, Issue Supplement_1, January 2024, Pages i843–i844
[2] Graham S, Vu QD, Jahanifar M, et al. CoNIC Challenge: Pushing the frontiers of nuclear detection, segmentation, classification and counting. Med Image Anal. 2024;92:103047.
Disclosure
MI received research grants and consultant fees from Olympus, Pentax, Eli Lilly, and Janssen. All other authors declare no competing interests.