Simultaneous estimation of kinetic parameters and the input function from DCE-MRI data: theory and simulation

DCE-MRI is a noninvasive functional imaging technique capable of assessing tumor microvasculature clinically. Major limitations associated with conventional region-of-interest (ROI) based compartmental methods include the requirement of invasive acquisition of the input function and labor-intensive identification of the ROIs. We propose a novel blind system identification approach for quantitative imaging by simultaneously estimating the input function and the kinetic parameters. New statistical model is on the pixel domain, whose parameters are initialized using a sub-space based algorithm, and further refined by an iterative maximum likelihood estimation procedure. The performances of the proposed scheme is examined extensively via simulations. The real breast tumor DCE-MRI data are examined by determining the time activity curves and the underlying factor images.