A voxel-wise composition ratio of DCE-MRI time-intensity curve profiles allows for visualizing and quantifying hemodynamic heterogeneity

We propose a novel model-free and data-driven approach, i.e., voxel-wise composition ratio on 19 dynamic contrast-enhanced MRI (DCE-MRI) time-intensity curve (TIC) profiles (Type-19) to visualize and quantify spatial hemodynamic heterogeneity. The proposed quantitative method for breast tumor was evaluated and compared with the two existing methods (qualitative and semi-quantitative methods) in 4 different clinical applications. In distinguishing malignancy on breast cancer lesions and predicting tumor proliferation status, we found that the machine learning model based on the Type-19 feature outperformed other two models in the validation set.