Estimating pharmacokinetic parameter maps from breast DCE-MRI with implicit regularization by guided image filtering

The pharmacokinetic (PK) parameters estimated from DCE-MRI by the nonlinear least squares approach on each pixel individually may be biased and erroneous. To obtain more reliable estimation, we propose a method to utilize the information of anatomical structure contained in the DCE-MRI image sequence for the reconstruction of PK parameter maps through the guided image filtering. Using the DCE-MRI image sequence as the guidance image, the guided image filter conveys the information of anatomical structure to the PK parameter maps as the spatial regularization in an implicit way. In this way, the bias and variance of estimation can be reduced. Results on the simulated and real breast DCE-MRI images showed that the proposed method could improve the estimation precision and obtain more reasonable PK parameter maps.