Generalized matched filtering for time-resolved MR angiography of pulsatile flow

Abstract

Generating flow-specific images (arteriograms, venograms) with optimal signal-to-noise ratios for time-resolved MR angiography is a conditional maximum problem, and its solutions are generalized matched filters. We have investigated six matched filters, corresponding to all possible combinations of three flow suppression conditions and two signal-to-noise ratio maximization procedures. Four of these matched filters correspond to previously described methods: the subtractive matched filter, the standard deviation, the global venous eigenimage and the global arterial eigenimage. The two others are referred to here as the local venous eigenimage and the local arterial eigenimage. These six matched filters have been applied to 2D time-resolved phase contrast angiographic data. The local arterial eigenimage is found to be the most effective in suppressing undesired venous flow and preserving desired arterial flow.