Automatic mask generation using independent component analysis in dynamic contrast enhanced-MRI

Studying image intensity change in each pixel in dynamic contrast enhanced (DCE)-MRI data enables differentiation of different tissue types based on their difference in contrast uptake. Pharmacokinetic modeling of tissues is commonly used to extract physiological parameters (i.e. K^trans and v_e) from the intensity-time curves. In a two compartmental model the intensity-time curve of the feeding blood vessels or arterial input function (AIF) as well as the signal from extravascular space (ES) is required. As direct measurement of these quantities is not possible some assumptions are made to approximate their values. Any error in measuring these quantities results in an error in the measured physiological parameters. We propose using Independent component analysis (ICA) to generate an automatic mask for separating the two spaces and extracting their intensity-time curves. An experimental phantom is constructed to mimic the behavior of real tissues and the actual intensity-time curves for the AIF and ES are measured from its DCE-MRI images. Then ICA is applied to the DCE dataset to separate these spaces. The result show high degree of agreement between the actual and ICA results.