A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer
Abstract
OBJECTIVES: Accurate MRI-based identification of extramural vascular invasion (EVI) and mesorectal fascia invasion (MFI) is crucial for risk-stratified rectal cancer treatment. However, subjective visual assessment and inter-institutional variability limit diagnostic consistency. This study developed and evaluated a multi-center, foundation model-driven framework that automatically classifies EVI and MFI on axial and sagittal MRI. MATERIALS AND METHODS: A total of 331 pre-treatment rectal cancer T2-weighted MRI scans from three European hospitals were retrospectively recruited. A self-supervised frequency domain harmonization strategy was applied to reduce scanner variability. Three classifiers, SeResNet, the universal biomedical pretrained model (UMedPT) with a multilayer perceptron head, and a logistic-regression variant using frozen UMedPT features (UMedPT_LR), were trained (n = 265) and tested (n = 66). Gradient-weighted class activation mapping (Grad-CAM) visualized model predictions. RESULTS: UMedPT_LR achieved the best EVI performance with multiplanar fusion (AUC = 0.82, test set). For MFI, UMedPT trained on axial harmonized images yielded the highest performance (AUC = 0.77). Both tasks outperformed the CHAIMELEON 2024 benchmark (EVI: 0.82 vs 0.74; MFI: 0.77 vs 0.75). Harmonization enhanced MFI classification, and multiplanar fusion further boosted EVI performance. Grad-CAM confirmed biologically plausible attention on peritumoral regions (EVI) and mesorectal fascia margins (MFI). CONCLUSION: The proposed foundation model-driven framework, leveraging frequency domain harmonization and multiplanar fusion, achieves state-of-the-art performance for automated EVI and MFI classification on MRI, demonstrating strong generalizability across multiple centers. CRITICAL RELEVANCE STATEMENT: Addressing inter-center inconsistencies in rectal cancer MRI, a multiplanar foundation model with cross-scanner harmonization significantly improves the detection of EVI and MFI, potentially standardizing staging and guiding therapy. KEY POINTS: Among the first studies to investigate automated classification of both EVI and MFI using axial and sagittal T2-weighted MRI. Foundation model-derived features outperform conventional convolutional neural networks (CNNs) for EVI and MFI classification. Frequency domain harmonization and multiplanar fusion selectively enhance diagnostic performance. Automated prediction of EVI and MFI may support more consistent staging and clinical decision-making across institutions.
Related
- MR-based radiomics of mesorectal fat for improved prediction of perirectal lymph node metastasis and extramural venous invasion in locally advanced rectal cancer
- CERVEX: A Foundation-Model Framework With Built-In Explainable AI for Automated Cervical Cytology Classification From Pap Smear Images
- Circumferential Meniscal Extrusion Mapping Using Radial Multiplanar Reconstructions
- Short-Term Outcomes and Stoma-Related Morbidity of Diverting Ileostomy After Laparoscopic Total Mesorectal Excision Following Neoadjuvant Chemoradiotherapy for Mid- and Distal Rectal Cancer: A Retrospective Single-Center Study
- Short‐Term Outcomes of Transanal Total Mesorectal Excision for Rectal Cancer: A Multicenter Propensity Score‐Matched Analysis Comparing Patients With and Without Preoperative Neoadjuvant Therapy
BibTeX
@article{zhang2026pretrained,
title = {{A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer}},
author = {Yumeng Zhang and Shruti Atul Mali and Danial Khan and Sina Amirrajab and Eduardo Ibor-Crespo and Ana Jimenez-Pastor and Gloria Ribas and Silvia Flor-Arnal and Marta Zerunian and Christophe Aubé and Luis Martí-Bonmatí and Zohaib Salahuddin and Philippe Lambin},
journal = {Insights into Imaging},
year = {2026},
month = may,
doi = {10.1186/s13244-026-02296-3},
url = {https://hal.science/hal-05841807v1},
note = {CC BY},
}
RIS
TY - JOUR
TI - A pre-trained foundation model framework for multiplanar MRI classification of extramural vascular invasion and mesorectal fascia invasion in rectal cancer
AU - Yumeng Zhang
AU - Shruti Atul Mali
AU - Danial Khan
AU - Sina Amirrajab
AU - Eduardo Ibor-Crespo
AU - Ana Jimenez-Pastor
AU - Gloria Ribas
AU - Silvia Flor-Arnal
AU - Marta Zerunian
AU - Christophe Aubé
AU - Luis Martí-Bonmatí
AU - Zohaib Salahuddin
AU - Philippe Lambin
PY - 2026
DA - 2026/05/22
JO - Insights into Imaging
DO - 10.1186/s13244-026-02296-3
UR - https://hal.science/hal-05841807v1
AB - OBJECTIVES: Accurate MRI-based identification of extramural vascular invasion (EVI) and mesorectal fascia invasion (MFI) is crucial for risk-stratified rectal cancer treatment. However, subjective visual assessment and inter-institutional variability limit diagnostic consistency. This study developed and evaluated a multi-center, foundation model-driven framework that automatically classifies EVI and MFI on axial and sagittal MRI. MATERIALS AND METHODS: A total of 331 pre-treatment rectal cancer T2-weighted MRI scans from three European hospitals were retrospectively recruited. A self-supervised frequency domain harmonization strategy was applied to reduce scanner variability. Three classifiers, SeResNet, the universal biomedical pretrained model (UMedPT) with a multilayer perceptron head, and a logistic-regression variant using frozen UMedPT features (UMedPT_LR), were trained (n = 265) and tested (n = 66). Gradient-weighted class activation mapping (Grad-CAM) visualized model predictions. RESULTS: UMedPT_LR achieved the best EVI performance with multiplanar fusion (AUC = 0.82, test set). For MFI, UMedPT trained on axial harmonized images yielded the highest performance (AUC = 0.77). Both tasks outperformed the CHAIMELEON 2024 benchmark (EVI: 0.82 vs 0.74; MFI: 0.77 vs 0.75). Harmonization enhanced MFI classification, and multiplanar fusion further boosted EVI performance. Grad-CAM confirmed biologically plausible attention on peritumoral regions (EVI) and mesorectal fascia margins (MFI). CONCLUSION: The proposed foundation model-driven framework, leveraging frequency domain harmonization and multiplanar fusion, achieves state-of-the-art performance for automated EVI and MFI classification on MRI, demonstrating strong generalizability across multiple centers. CRITICAL RELEVANCE STATEMENT: Addressing inter-center inconsistencies in rectal cancer MRI, a multiplanar foundation model with cross-scanner harmonization significantly improves the detection of EVI and MFI, potentially standardizing staging and guiding therapy. KEY POINTS: Among the first studies to investigate automated classification of both EVI and MFI using axial and sagittal T2-weighted MRI. Foundation model-derived features outperform conventional convolutional neural networks (CNNs) for EVI and MFI classification. Frequency domain harmonization and multiplanar fusion selectively enhance diagnostic performance. Automated prediction of EVI and MFI may support more consistent staging and clinical decision-making across institutions.
SN - 1869-4101
ER -