Observer performance methodologies for evaluating blood vessel visibility in MR angiograms using accurate geometric registration to high resolution x-ray angiograms.

Abstract

Receiver operator characteristic (ROC) and two alternative forced choice (2AFC) methodologies for quantitatively assessing vessel visibility in MR angiography (MRA) were examined using x-ray angiography images as truth. The methodologies required MRA projection images with the same orientation and magnification as the x-ray images. Geometric distortions limited the quality of the registration. The observer performance measurements were compared to vessel contrast-to-noise ratio (CNR) measurements. Receiver bandwidth (RBW) and magnetization transfer (MT) effects in 3D time-of-flight MRA were evaluated. Overall, applying MT significantly increased all three measurements while decreasing the RBW significantly improved the 2AFC and CNR measurements. The relative importance of both effects on the 2AFC measure increased as vessel diameter decreased, although statistical significance was limited by small samples for diameters less than approximately 0.7 mm. These results demonstrate the usefulness of these observer performance methodologies for MRA technique assessment.