CT-based radiomics for predicting breast cancer radiotherapy side effects

Area of Science:

• Oncology
• Medical Imaging
• Radiotherapy

Background:

• Breast radiotherapy (RT) commonly causes acute skin side effects like inflammation, moist epitheliolysis, and edema.
• Predicting these side effects is crucial for improving patient care and treatment planning.

Purpose of the Study:

• To compare the predictive value of radiomics features derived from tissue with total breast volume (TBV) for predicting moist epitheliolysis and edema.
• To evaluate the performance of machine learning models using radiomics and clinical features for predicting RT-induced skin toxicity.

Main Methods:

• Radiomics features were extracted from computed tomography (CT) scans of 252 breast cancer patients.
• Features were analyzed from total breast volume (TBV) and glandular tissue (GT) regions of interest.
• Machine learning classifiers, including LASSO, were trained and evaluated for predicting skin side effects.

Main Results:

• The best radiomics model using TBV features achieved an AUROC of 0.74 for predicting moist epitheliolysis, comparable to TBV alone (AUROC 0.75).
• Mammary tissue radiomics demonstrated higher predictive value than glandular tissue.
• Excluding volume-correlated features slightly decreased predictive performance (AUROC 0.71).

Conclusions:

• CT-based radiomics models show potential for predicting breast radiotherapy-induced skin side effects.
• Radiomics features derived from mammary tissue are more predictive than those from glandular tissue.
• The predictive performance of radiomics features is influenced by their correlation with total breast volume.