MRI-based Radiomic Features for Risk Stratification of Ductal Carcinoma in Situ in a Multicenter Setting (ECOG-ACRIN...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Ductal carcinoma in situ (DCIS) is a preinvasive breast cancer often subject to overtreatment.
• Accurate characterization of DCIS extent is crucial for appropriate management.
• Breast MRI is a key imaging modality for assessing DCIS extent.

Purpose of the Study:

• To develop and evaluate logistic regression models for predicting disease upstaging and DCIS score in DCIS patients.
• To integrate clinical and MRI-based radiomic features for improved risk stratification.
• To assess the performance of radiomic models in predicting surgical outcomes and DCIS grade.

Main Methods:

• Secondary analysis of DCE-MRI data from the ECOG-ACRIN E4112 trial.
• Computation of 65 radiomic features using CaPTk software.
• Development of logistic regression models incorporating clinical, qualitative MRI, and radiomic features (principal components).
• Model performance evaluated using AUC on a held-out test set.

Main Results:

• Two radiomic phenotypes were associated with disease upstaging (P = .007).
• Combined radiomic and clinical models achieved an AUC of 0.77 for predicting upstaging, outperforming clinical information alone.
• The combined model identified 25% more true-negative findings (53% specificity vs. 28% specificity).
• Radiomic models did not predict DCIS score (P > .05).

Conclusions:

• Integrating radiomic metrics with clinical data significantly improves the prediction of disease upstaging in DCIS.
• Radiomic models show potential for reducing overtreatment by better identifying patients who do not require more aggressive surgical intervention.
• Radiomic features did not demonstrate predictive value for DCIS score in this study.