Related Experiment Videos
Explainable Multiparametric MRI Radiomics Reveals Prognostically Relevant Intratumoral Heterogeneity in Rectal Cancer
Zhe Dong1, Shubo Ding2, Chen Huang3
1Department of Radiation Oncology, Jinhua Central Hospital, Teaching Hospital of Mathematical Medicine College, Zhejiang Normal University, Zhejiang, China (Z.D.).
Academic Radiology
|August 10, 2026
Summary
An explainable MRI radiomics signature effectively quantifies intratumoral heterogeneity (ITH) in rectal cancer. High SHAP heterogeneity predicts worse disease-free survival (DFS) and aids in risk stratification.
Area of Science:
- Radiology
- Oncology
- Medical Imaging Analysis
Background:
- Intratumoral heterogeneity (ITH) in rectal cancer is challenging to assess noninvasively via standard MRI.
- Developing quantitative methods for ITH assessment is crucial for accurate risk stratification.
Purpose of the Study:
- To develop and validate an explainable multiparametric MRI radiomics signature for quantifying ITH in rectal cancer.
- To assess the association of this signature with disease-free survival (DFS) and treatment outcomes.
Main Methods:
- A retrospective multicenter study involving 794 rectal cancer patients.
- Integration of radiomics features from T2-weighted imaging, diffusion-weighted imaging, and ADC maps using similarity network fusion.
- Development of an 18-feature texture-based ITH score using an extreme gradient boosting model with SHAP for risk stratification.
Main Results:
- High SHAP heterogeneity was significantly associated with worse DFS across training and validation cohorts.
- High SHAP heterogeneity independently predicted inferior DFS and was linked to lymphovascular invasion in patients receiving neoadjuvant therapy.
- A nomogram incorporating SHAP heterogeneity, CEA, and MRI-T stage improved 3-year DFS prediction.
Conclusions:
- The developed explainable multiparametric MRI radiomics signature effectively identifies an aggressive MRI phenotype in rectal cancer.
- High SHAP heterogeneity is linked to poor DFS and residual invasive pathology, supporting its use for pretreatment risk stratification.
- This approach offers a promising noninvasive tool for personalized treatment strategies in rectal cancer.