Predicting the response of I-SPY 2 breast cancer patients to treatment using a biology-based mathematical model calibrated with quantitative MRI

Motivation: Optimizing treatment to improve outcomes necessitates a robust tool to accurately predict breast cancer response on a patient-specific basis. Goal(s): We are applying our biology-based mathematical model to I-SPY 2 breast cancer patients to test if its predictive ability generalizes to multi-site data. Approach: Quantitative contrast-enhanced and diffusion-weighted MRI data collected early during treatment were used to calibrate a mathematical model describing tumor cell movement, proliferation, and response. After calibration, the model predicts tumor status after the treatment regimen. Results: The concordance correlation coefficient between the measured and predicted 9-week change was 0.91 for tumor cellularity and 0.88 for tumor volume. Impact: The high degree of agreement between measured and predicted changes in tumor cellularity and volume in the I-SPY 2 dataset indicates that our biology-based mathematical model can potentially make accurate predictions using MRI data from multiple clinical sites.