Abstract

Crohn’s disease (CD) is a chronic inflammatory bowel disease whose activity and complications can be assessed using magnetic resonance enterography (MRE). Existing MRE-based approaches for automated CD assessment either rely on manually extracted features or are limited to coarse disease detection rather than classification of specific radiological findings. Furthermore, most approaches utilize only a single MRE contrast, despite the complementary information available from multi-contrast imaging. Therefore, we introduce a model that extracts eight CD radiological findings using T1- and T2-weighted MRE without handcrafted features. Our model consists of contrast-specific feature extractors, masked Generalized Mean (GeM) pooling, and a modality-weighting module preceding the classification head. We use a pretrained supervised segmentation model to extract sub-organ intestinal segmentations and use them to guide region-based feature pooling from the feature maps. We evaluate our model through two experiments: First, we compare the multi-contrast model, which uses both T1 and T2 contrasts, with single-contrast models to assess the contributions of each contrast. Second, we compare different modality weightings, including linear, learned, and attention-based. In the first experiment, the multi-contrast model achieves a weighted AUPRC of 0.683±0.011, outperforming the single-contrast T1 or T2 models, which achieved 0.657±0.010 and 0.677±0.010, respectively. In the second experiment, the attention-based weighting produced the best performance with a weighted AUPRC of 0.691±0.011. These results demonstrate the value of segmentation-guided multi-contrast MRE for automated classification of CD imaging findings.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CLIP_003.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to Open Review

Open Review Page: https://openreview.net/forum?id=7HGk6Fsz8d

BibTex

@InProceedings{TeiSar_SegClassCD_MICCAISAT2026,
        author = { Teitz, Sarah AND Gavrielov, Naama AND Gitelman, Leah AND Focht, Gili AND Cytter-Kuint, Ruth AND Hagopian, Talar AND Vainberg, Elena AND Cohen, Israel AND Turner, Dan AND Freiman, Moti},
        title = { { SegClass-CD: Segmentation-Guided Region-Based Classification of Crohn’s Disease Imaging Findings from Multi-Contrast MRE } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 17273},
        month = {pending},
        page = {pending}
}


back to top