Related Experiment Video
Updated: Dec 1, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.2K
Comparison of Radiomic Feature Aggregation Methods for Patients with Multiple Tumors
Enoch Chang1, Marina Joel2, Hannah Y Chang3
1Department of Therapeutic Radiology, Yale School of Medicine.
Medrxiv : the Preprint Server for Health Sciences
|November 11, 2020
Summary
Combining radiomic features from the largest three brain metastases using a weighted average improves patient prognostication for multifocal metastatic cancer. This approach enhances personalized treatment decisions for patients with multiple brain metastases.
Area of Science:
- Oncology
- Medical Imaging
- Radiomics
Background:
- Radiomic feature analysis shows promise for predicting cancer outcomes.
- Optimal methods for combining radiomic features in multifocal disease remain unclear.
- Rising incidence of multifocal metastatic cancer necessitates improved patient-level prognostication.
Approach:
- Compared six mathematical methods for combining radiomic features from 3596 tumors across 831 patients with multiple brain metastases.
- Evaluated aggregation method performance using Cox proportional hazards, LASSO-regularized Cox, and random survival forest models.
Key Points:
- The weighted average of the largest three metastases achieved the highest concordance index across all evaluated survival models.
- Concordance indices ranged from 0.627 to 0.652, indicating effective prognostication.
- This method offers a robust approach to patient-level outcome prediction in multifocal brain metastases.
Conclusions:
- Radiomic features can be effectively combined to predict patient outcomes in multifocal brain metastases.
- The volume-weighted average of the largest three tumors is a promising method for feature aggregation.
- Further research should validate this approach across diverse imaging modalities and cancer types.
