An Optimal Mass Transport Model for the Analysis of DCE-MRI and its Application to Breast Cancer Treatment Response

Abstract

Purpose: Dynamic contrast-enhanced MR imaging (DCE-MRI) is widely deployed in cancer care and research, but the methods conventionally used to quantify contrast agent kinetics do not account the cross-voxel movement characterized by advection and diffusion. We hypothesized that unbalanced optimal mass transport could be used to quantify and visualize such contrast agent flows across tumor volumes. Methods: We developed a computational fluid dynamics model termed the unbalanced regularized optimal mass transport (urOMT) model. We tested the urOMT on a multi-institutional dataset of 153 longitudinal DCE-MRI scans from 39 breast cancer patients treated with neoadjuvant chemotherapy (NACT.) Results: The urOMT model can quantify dynamic fluid transport properties such as net speed, flux and rates of contrast entering and leaving the tumor (influx and efflux). The urOMT model can also visualize the trajectories and directions of net fluid flows. Quantitative metrics from urOMT exhibited distinct patterns that may be relevant to predicting pathological complete response (pCR) to NACT. Conclusion: The urOMT model can be used to estimate and visualize local fluid flow in DCE-MRI breast cancer images. Model-based estimates of flux, influx and efflux should be tested as potential predictive imaging biomarkers to measure treatment effectiveness in patients treated with NACT. The urOMT model in principle has applicability to other cancer imaging use cases, but this will require further testing.