A task-based approach to parametric imaging with dynamic contrast enhanced MRI

In this paper, we propose a task-based approach to parametric imaging and apply the proposed method to an example problem of prostate cancer segmentation with dynamic contrast enhanced Magnetic Resonance Imaging (DCE MRI). Traditionally, the time activity curve obtained from dynamic series of MR images is modeled without considering a specific task in order to obtain the kinetic parameters and to construct the parametric images. This mostly consists of estimating parameters based on minimizing the error between the model and measurement. In this paper, we develop a new method for the estimation of kinetic parameters based on the maximization of tumor segmentation performance. We use Fisher Ratio as the criterion that quantifies the image's ability to classify tumor and normal pixels. Then, the kinetic parameters are estimated with a weighted approach such that the Fisher Ratio is maximized. The calculation of the Fisher Ratio requires the prior knowledge of the confirmed regions of the tumor. Therefore, we use a training dataset to determine the optimum set of parametric images. The proposed method results in parametric images with a considerable improvement in terms of classification power between tumor and the normal regions of the prostate.