Related Experiment Video
Updated: May 8, 2026
07:41
Performing Data Mining And Integrative Analysis Of Biomarker in Breast Cancer Using Multiple Publicly Accessible Databases
Published on: May 17, 2019
Deep interpretable radiogenomic workflow deciphers tumor microenvironment from breast MRI and identifies
Huijun Li1,2, Qiuxia Yang3, Rui Zhang2
1Faculty of Health Sciences, University of Macau, Macao SAR, China.
NPJ Precision Oncology
|May 6, 2026
Summary
This study introduces an interpretable workflow correlating MRI with the tumor microenvironment (TME). It identifies imaging biomarkers and subtypes, improving cancer diagnosis and treatment prediction.
Area of Science:
- Oncology
- Radiology
- Bioinformatics
Background:
- The tumor microenvironment (TME) significantly impacts cancer prognosis and immunotherapy response.
- Current TME assessment methods rely on invasive pathology, facing challenges from tumor heterogeneity.
- There's a need for non-invasive biomarkers to accurately assess TME composition and guide treatment.
Purpose of the Study:
- To develop an interpretable workflow correlating Magnetic Resonance Imaging (MRI) with TME characteristics.
- To enable unsupervised lesion annotation and identify cancer imaging biomarkers and subtypes.
- To discover novel, clinically relevant radiomic features for TME assessment and patient stratification.
Main Methods:
- A novel deconvolution workflow was developed to infer TME profiles from bulk data by integrating gene and image information.
- Unsupervised lesion annotation was performed incorporating TME data.
- Radiogenomics models were built using interpretable radiomic features to predict TME components and identify cancer subtypes.
Main Results:
- The workflow successfully deconvoluted TME data, outperforming existing methods across multiple datasets.
- An inverse relationship was observed between cancer-associated fibroblasts (CAFs) and T-cell infiltration in triple-negative breast cancer (TNBC).
- A radiogenomics model achieved 0.87 accuracy in predicting CAF proportions and identified novel CAF-associated imaging biomarkers. Radiomic features for subtyping showed high accuracy (>0.8) in multicenter validations.
Conclusions:
- The developed interpretable workflow effectively correlates MRI with TME, offering a non-invasive assessment method.
- This approach facilitates biomarker discovery and clinical application, enhancing understanding of TME heterogeneity.
- The identified imaging biomarkers and subtypes hold promise for improved cancer diagnosis, prognosis, and personalized treatment strategies.
