Fluid mechanics based quantitative transport mapping network (QTMnet) for predicting treatment response of breast cancer from DCE MRI

Motivation: There lacks a non-invasive breast cancer treatment response prediction method from MRI images. Goal(s): To test the feasibility of using perfusion parameter reconstructed from QTMnet processing of DCE MRI to predict breast cancer treatment response. Approach: QTMnet is trained on synthetically generated perfusion data and tested on 41 breast DCE MRI images. Results: $$$PS$$$ and $$Vₑ$ map from QTMnet can predict whether breast cancer has complete treatment responses. Impact: QTMnet can be coupled into clinical breast cancer practice to predict treatment response.