In Dynamic Contrast-Enhanced (DCE) MRI, inaccurate estimation of the arterial input function (AIF) is still a major cause of the low reliability of kinetic parameter estimates. We propose a new method of AIF estimation. It combines AIF measured from the blood-pool signal and multichannel blind deconvolution. The weights of the measured AIF are based on its analytically derived uncertainty and a model relating signal intensity and gadolinium concentration.
The method has been evaluated on simulated myocardial perfusion data, mimicking real noise and kinetic parameter distributions. The hybrid method gave better results compared to the blood-pool or blind-deconvolution approaches alone.