Evaluation of Measures of Technical Image Quality for Intracranial Magnetic Resonance Angiography

Abstract

We evaluate three measures of technical image quality for intracranial magnetic resonance angiography (MRA): (1) a two-alternative forced choice (2AFC) evaluation of vessel visibility, (2) vessel-to-background signal-difference-to-noise ratio (SDNR), and (3) observer ranking of the fidelity of vessel morphology compared to that in a gold standard image. The gold standard used for both the 2AFC and ranking measures is intraarterial catheter angiography. These measures are applied to healthy arterial segments. The 2AFC and SDNR measures directly evaluate the visibility of artery segments for which the existence is known from the gold standard images. We argue that (1) 2AFC evaluates the carrier signals on which any vascular disease process is modulated and provides an upper bound on the detectibility of vascular lesions, (2) SDNR is a predictor of 2AFC, and (3) ranking may be used to predict the relative performance of techniques in the detection of vascular lesions.