GENERALIZED MATCHED FILTERING FOR TIME-RESOLVED MR-ANGIOGRAPHY OF PULSATILE FLOW

Generating flow-specific images (arteriograms, venograms) with optimal signal-to-noise ratios for time-resolved MR angiography is a conditio nal maximum problem, and its solutions are generalized matched filters . We have investigated six matched filters, corresponding to all possi ble combinations of three flow suppression conditions and two signal-t o-noise ratio maximization procedures. Four of these matched filters c orrespond to previously described methods: the subtractive matched fil ter, the standard deviation, the global venous eigenimage and the glob al arterial eigenimage. The two others are referred to here as the loc al venous eigenimage and the local arterial eigenimage. These six matc hed filters have been applied to 2D time-resolved phase contrast angio graphic data. The local arterial eigenimage is found to be the most ef fective in suppressing undesired venous flow and preserving desired ar terial flow.