Radiomics to Detect Inflammation and Fibrosis on Magnetic Resonance Enterography in Stricturing Crohn's Disease

Abstract

Background and aims: Non-invasive cross-sectional imaging via magnetic resonance enterography [MRE] offers excellent accuracy for the diagnosis of stricturing complications in Crohn's disease [CD] but is limited in determining the degrees of fibrosis and inflammation within a stricture. We developed and validated a radiomics-based machine-learning model for separately characterizing the degree of histopathological inflammation and fibrosis in CD strictures and compared it to centrally read visual radiologist scoring of MRE.

Methods: This single-centre, cross-sectional study included 51 CD patients [n = 34 for discovery; n = 17 for validation] with terminal ileal strictures confirmed on diagnostic MRE within 15 weeks of resection. Histopathological specimens were scored for inflammation and fibrosis and spatially linked with corresponding pre-surgical MRE sequences. Annotated stricture regions on MRE were scored visually by radiologists as well as underwent 3D radiomics-based machine learning analysis; both were evaluated against histopathology.

Results: Two distinct sets of radiomic features capturing textural heterogeneity within strictures were linked with each of severe inflammation or severe fibrosis across both the discovery (area under the curve [AUC = 0.69, 0.83] and validation [AUC = 0.67, 0.78] cohorts. Radiologist visual scoring had an AUC = 0.67 for identifying severe inflammation and AUC = 0.35 for severe fibrosis. Use of combined radiomics and radiologist scoring robustly augmented identification of severe inflammation [AUC = 0.79] and modestly improved assessment of severe fibrosis [AUC = 0.79 for severe fibrosis] over individual approaches.

Conclusions: Radiomic features of CD strictures on MRE can accurately identify severe histopathological inflammation and severe histopathological fibrosis, as well as augment performance of the radiologist visual scoring in stricture characterization.

Keywords: Stenosis; extracellular matrix; fibrostenosis; therapy.

Figure 1.

Experimental workflow for developing the stricturing CD radiomics model. [A] An inflammatory bowel disease [IBD] radiologist identified the region of stricturing disease in conjunction with an IBD pathologist with access to the resection. [B] Quantitative radiomic features are extracted on a per-pixel basis within the green region of interest [ROI]. [C] A subset of radiomic features that strongly discriminate between pathologically severe and non-severe fibrosis [or inflammation] is selected via machine learning. [D] Standardized scoring of the corresponding pathology section used as ground truth for severity of inflammation or fibrosis. [E] Radiomic and histopathology features are correlated and used to train a machine learning [ML] model. [F] The ML model predicts the severity of histopathological findings within the stricturing region based on the radiomic profile.

Figure 2.

Accuracy of the radiomics model, the radiologist assessment, and their combination to discriminate pathologically non-severe and severe inflammation within stricturing Crohn’s disease via MRE. [A] Depiction of two representative patients with non-severe and severe inflammation, respectively, based on histopathology scoring of the resection tissue [left], together with annotated region of interest [ROI] [middle] and the top ranked radiomic feature heatmaps [right]. Higher heterogeneity in radiomic expression is observed via increasing proportion of blue/red within the heatmap. [B] Average radiomic classifier probability for severe inflammation plotted against ground truth histopathology pVAS. The central solid line shows the linear regression between these two variables with the thinner curves showing error bounds. [C] Violin plots of non-severe [blue] and severe [red] inflammation patients using the radiologist visual assessment score [rVAS] [left] and the top ranked radiomic feature [right]. [D] Receiver–operating characteristic curves showing ML model performance for radiomic features [in both the discovery and validation cohorts], rVAS score, and the augmented rVAS/radiomics model, together with bar plots for sensitivity, specificity, and F1 scores.

Figure 3.

Accuracy of the radiomics model, the radiologist assessment, and their combination to discriminate pathologically non-severe and severe fibrosis within stricturing Crohn’s disease via MRE. [A] Depiction of two representative patients with non-severe and severe fibrosis, respectively, based on histopathology scoring of the resection tissue [left], together with annotated region of interest [ROI] [middle] and the top ranked radiomic feature heatmaps [right]. Higher heterogeneity in radiomic expression is observed via increasing proportion of red within the heatmap. [B] Average radiomic classifier probability for severe fibrosis plotted against ground truth histopathology pVAS. The central solid line shows the linear regression between these two variables with the thinner curves showing error bounds. [C] Violin plots of non-severe [blue] and severe [red] fibrosis patients using the radiologist visual assessment score [rVAS] [left] and the top ranked radiomic feature [right]. [D] Receiver–operating characteristic curves showing ML model performance for radiomic features [in both the discovery and validation cohorts], rVAS, and the augmented rVAS/radiomics model, together with bar plots for sensitivity, specificity, and F1 scores.

Figure 4.

Cell map visualization depicting the association of top ranked radiomic features with individual histopathological components comprising inflammation and fibrosis. The colour of the cell represents the p-value between the feature and the histopathological score with a darker blue indicating a lower p-value and yellow a higher p-value. Stars indicate statistically significant associations, obtained via Kruskal–Wallis testing.