Deep Learning-Based Synthetic Contrast-Enhanced Breast MRI for Monitoring Response to Neoadjuvant Therapy

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Contrast-enhanced breast MRI (CE-MRI) is crucial for assessing breast cancer extent and treatment response.
• Gadolinium-based contrast agents (GBCAs) in CE-MRI pose risks and increase costs.
• Deep learning (DL) offers a potential solution to reduce GBCA use.

Purpose of the Study:

• To explore the feasibility of using a DL model to synthesize CE-MRI from non-contrast MRI for breast cancer treatment monitoring.
• To evaluate the accuracy of synthetic CE-MRI in measuring tumor volume changes and predicting treatment outcomes.

Main Methods:

• A retrospective pilot study involving women undergoing neoadjuvant therapy (NAT) for breast cancer.
• A pre-trained DL model synthesized CE-MRI from T1-, T2-, and diffusion-weighted MRI.
• Tumor volumes and prediction of residual cancer burden (RCB) class 0/1 were compared between synthetic and acquired CE-MRI.

Main Results:

• Synthetic CE-MRI showed strong correlation with acquired CE-MRI for tumor volumes pre-treatment and early in treatment.
• Agreement in tumor volume measurement decreased at mid-treatment.
• Synthetic CE-MRI demonstrated comparable performance to acquired CE-MRI in predicting favorable RCB class (0/1 vs. 2/3) outcomes.

Conclusions:

• Synthetic CE-MRI shows preliminary feasibility as a GBCA-free alternative for predicting favorable breast cancer treatment outcomes.
• Further model refinement and validation are needed due to inconsistencies in tumor volume measurements compared to acquired CE-MRI.