A Radiogenomic Model using MRI and Gene Signature to Predict Complete Response in Breast Cancer.

Area of Science:

• Oncology
• Radiology
• Genomics

Background:

• Pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) is a key predictor of improved survival in breast cancer patients.
• Radiogenomic approaches integrating MRI radiomics and gene signatures offer potential for enhancing pretreatment prediction of therapeutic response.

Purpose of the Study:

• To develop and validate a radiogenomic model for predicting pCR in breast cancer.
• To evaluate the performance of MRI radiomics, an immune-related 23-gene signature (IRSN-23), and their combination in predicting pCR.

Main Methods:

• Retrospective analysis of 112 breast cancer patients who underwent pre-NAC MRI and had IRSN-23 scores.
• Extraction of 100 radiomic features from pre-NAC MRI.
• Training and validation of prediction models (radiomics alone, IRSN-23 alone, combined radiogenomic model) using a support vector classifier.

Main Results:

• The combined radiogenomic model achieved higher AUCs in both training (0.93) and validation (0.89) cohorts compared to MRI radiomics or IRSN-23 alone.
• The combined model significantly outperformed individual modalities in the training cohort (p < 0.05).
• While not statistically significant, patients predicted to achieve pCR showed a trend toward longer relapse-free and overall survival.

Conclusions:

• Integration of MRI radiomics with IRSN-23 significantly improves the prediction of pCR in breast cancer.
• The developed radiogenomic model shows potential for supporting treatment planning.
• Further research is needed to confirm the association between model-predicted pCR and long-term survival outcomes.