Accurate and early prediction of neoadjuvant therapy (NAT) response is essential to personalize treatment to increase pathological complete response (pCR) rates for locally advanced breast cancer (LABC) patients. Here, we use a biology-based mathematical model calibrated with quantitative MRI data to predict tumor status after a course of NAT.

91 patients of three unique molecular subtypes from the I-SPY 2 clinical trial received dynamic contrast-enhanced and diffusion-weighted MRI before (V1), 3 weeks into (V2), and after completion of (V3) a standard-of-care or experimental course of NAT. Using a biology-based mathematical model computing the voxel-wise change in number of tumor cells as a function of diffusion, proliferation, and death due to treatment, we calibrated parameters with V1 and V2 data to make patient-specific predictions at V3.

The concordance correlation coefficients between the observed and modeled change from V1 to V3 was 0.94 for total tumor cellularity and 0.90 for tumor volume. Our model predictions at V3 differentiated patients who achieved pCR from non-pCR patients with an area under the receiver operator characteristic curve of 0.78.

Our mathematical model can accurately predict tumor status for LABC patients after a course of treatment using only standard-of-care MRI data.