Citations

Abstract Background Eosinophils have been implicated in the pathogenesis of ulcerative colitis and have been associated with disease course and therapeutic response. However, associations between eosinophil density, histologic activity, and clinical features have not been rigorously studied. Methods A deep learning algorithm was trained to identify eosinophils in colonic biopsies and validated against pathologists’ interpretations. The algorithm was applied to sigmoid colon biopsies from a cross-sectional cohort of 88 ulcerative colitis patients with histologically active disease as measured by the Geboes score and Robarts histopathology index (RHI). Associations between eosinophil density, histologic activity, and clinical features were determined. Results The eosinophil deep learning algorithm demonstrated almost perfect agreement with manual eosinophil counts determined by 4 pathologists (interclass correlation coefficients: 0.805–0.917). Eosinophil density varied widely across patients (median 113.5 cells per mm2, interquartile range 108.9). There was no association between eosinophil density and RHI (P = 0.5). Significant differences in eosinophil density were seen between patients with Montreal E3 vs E2 disease (146.2 cells per mm2 vs 88.2 cells per mm2, P = 0.005). Patients on corticosteroids had significantly lower eosinophil density (62.9 cells per mm2 vs 124.1 cells per mm2, P = 0.006). No association between eosinophil density and biologic use was observed (P = 0.5). Conclusions We developed a deep learning algorithm to quantify eosinophils in colonic biopsies. Eosinophil density did not correlate with histologic activity but did correlate with disease extent and corticosteroid use. Future studies applying this algorithm in larger cohorts with longitudinal follow-up are needed to further elucidate the role of eosinophils in ulcerative colitis.

Objective Effective medical therapy and validated trial outcomes are lacking for small bowel Crohn’s disease (CD) strictures. Histopathology of surgically resected specimens is the gold standard for correlation with imaging techniques. However, no validated histopathological scoring systems are currently available for small bowel stricturing disease. We convened an expert panel to evaluate the appropriateness of histopathology scoring systems and items generated based on panel opinion. Design Modified RAND/University of California Los Angeles methodology was used to determine the appropriateness of 313 candidate items related to assessment of CD small bowel strictures. Results In this exercise, diagnosis of naïve and anastomotic strictures required increased bowel wall thickness, decreased luminal diameter or internal circumference, and fibrosis of the submucosa. Specific definitions for stricture features and technical sampling parameters were also identified. Histopathologically, a stricture was defined as increased thickness of all layers of the bowel wall, fibrosis of the submucosa and bowel wall, and muscularisation of the submucosa. Active mucosal inflammatory disease was defined as neutrophilic inflammation in the lamina propria and any crypt or intact surface epithelium, erosion, ulcer and fistula. Chronic mucosal inflammatory disease was defined as crypt architectural distortion and loss, pyloric gland metaplasia, Paneth cell hyperplasia, basal lymphoplasmacytosis, plasmacytosis and fibrosis, or prominent lymphoid aggregates at the mucosa/submucosa interface. None of the scoring systems used to assess CD strictures were considered appropriate for clinical trials. Conclusion Standardised assessment of gross pathology and histopathology of CD small bowel strictures will improve clinical trial efficiency and aid drug development.

Abstract Background and Aims To correlate histologic activity in surveillance colonoscopies with the development of colorectal neoplasia in ulcerative colitis [UC]. Methods Colorectal biopsies during surveillance [N = 764] from 52 UC patients with colorectal neoplasia were compared to 122 patients without neoplasia enrolled in a prospective natural history registry. All biopsies were scored using validated histologic scoring systems (Geboes score, Nancy histopathologic index [NHI], and Robarts histopathologic index [RHI]). Clinical, endoscopic, and histologic data were correlated with the development of colorectal neoplasia. Results In multivariable analysis, mean RHI (hazard ratio [HR] 1.07 for each 1-unit increase in RHI, 95% confidence interval [CI] 1.03–1.12, p = 0.002) and mean NHI [HR 1.89 for each 1-unit increase in NHI, 95% CI 1.34–2.67, p = 0.002] for the entire surveillance period were significantly associated with colorectal neoplasia development. Shorter surveillance interval and increasing age were associated with increased risk of neoplasia development whereas mean Mayo endoscopic score was not significant. To generate a clinically useful measure of neoplasia risk, mean histologic activity in the preceding 5 years before the study endpoint was correlated with neoplasia development. In the preceding 5 years of surveillance, a mean RHI ≥ 8 had a 7.53-fold increased risk [95% CI 2.56–12.16, p < 0.001] and mean NHI ≥ 1.9 had a 5.89-fold increased risk [95% CI 2.18–15.92, p < 0.001] of developing colorectal neoplasia. Conclusions Persistent histologic activity during multiple surveillance episodes is an independent predictor of colorectal neoplasia. Mean RHI and mean NHI during a 5-year colonoscopic surveillance period can be used to assess risk for colorectal neoplasia and optimize UC surveillance.

AimsTo develop and validate a deep learning algorithm to quantify a broad spectrum of histological features in colorectal carcinoma.Methods and resultsA deep learning algorithm was trained on haematoxylin and eosin‐stained slides from tissue microarrays of colorectal carcinomas (N = 230) to segment colorectal carcinoma digitised images into 13 regions and one object. The segmentation algorithm demonstrated moderate to almost perfect agreement with interpretations by gastrointestinal pathologists, and was applied to an independent test cohort of digitised whole slides of colorectal carcinoma (N = 136). The algorithm correctly classified mucinous and high‐grade tumours, and identified significant differences between mismatch repair‐proficient and mismatch repair‐deficient (MMRD) tumours with regard to mucin, inflammatory stroma, and tumour‐infiltrating lymphocytes (TILs). A cutoff of >44.4 TILs per mm2 carcinoma gave a sensitivity of 88% and a specificity of 73% in classifying MMRD carcinomas. Algorithm measures of tumour budding (TB) and poorly differentiated clusters (PDCs) outperformed TB grade derived from routine sign‐out, and compared favourably with manual counts of TB/PDCs with regard to lymphatic, venous and perineural invasion. Comparable associations were seen between algorithm measures of TB/PDCs and manual counts of TB/PDCs for lymph node metastasis (all P < 0.001); however, stronger correlations were seen between the proportion of positive lymph nodes and algorithm measures of TB/PDCs. Stronger associations were also seen between distant metastasis and algorithm measures of TB/PDCs (P = 0.004) than between distant metastasis and TB (P = 0.04) and TB/PDC counts (P = 0.06).ConclusionsOur results highlight the potential of deep learning to identify and quantify a broad spectrum of histological features in colorectal carcinoma.

Objective We evaluated the reliability and responsiveness of available but incompletely validated UC histological disease activity indices using standardised rules for centralised assessment. Design Disease activity was assessed in biopsies collected in a phase II placebo-controlled ozanimod trial by four blinded pathologists using the Geboes (GS) and modified Riley (MRS) scores, the Robarts Histopathology (RHI) and Nancy Histological (NHI) indices and a Visual Analogue Scale. Reliability was assessed with intraclass correlation coefficients (ICCs). Index responsiveness was evaluated by assessing longitudinal validity (Pearson correlations of changes in index scores and other disease measures), and effect size estimates (standardised effect size (SES)) using two criteria for change (treatment assignment and >2 point decrease in total Mayo Clinic score). Area under the receiver operating characteristic (AUROC) curve estimates evaluated the probability of the indices to discriminate between treatment and placebo. Results Inter-rater reliability of the histological indices was substantial to almost perfect (ICC>0.61), and responsiveness was moderate to large (SES estimates>0.5); 0.81 (0.52, 1.10), 0.87 (0.58, 1.17), 0.57 (0.30, 0.84) and 0.81 (0.52, 1.09) when treatment assignment was the criterion for change and 1.05 (0.80, 1.31), 1.13 (0.87, 1.39), 0.88 (0.64, 1.12) and 1.06 (0.80, 1.31) for the change in Mayo score criterion for the GS, MRS, RHI and NHI, respectively. The indices had similar drisciminative ability based on AUROC estimates (range 0.608–0.649). Conclusion All four existing histological indices were similarly reliable and responsive based on this dataset.