Objectives: Time-of-arrival (TOA) maps can be derived from high-resolution 4-dimensional (4D) contrast-enhanced magnetic resonance angiography (MRA) data sets to provide a quantitative description of contrast material arrival time in each voxel. This information can further be processed to create a compressed time evolution curve that virtually shortens the contrast bolus (virtual bolus [VB]). The purpose of this project was to determine whether TOA-enhanced 4D MRA and/or VB imaging improve the display of contrast kinetics in patients with vascular disease.
Methods: High-resolution whole-brain contrast-enhanced 4D MRA examinations with 1.2-second temporal reconstruction were acquired by using radial acquisition and highly constrained projection reconstruction (radial 4D contrast-enhanced HYPRFlow, abbreviated as HFMRA in this article) in 10 patients (8 patients with arteriovenous malformations [AVM], 1 patient with an arteriovenous fistula, and 1 patient with a high-grade intracranial stenosis). The TOA for each voxel was defined as the time point when the signal intensity reached 20% of its maximum. In the first method, TOA maps were generated, color-encoded, and then multiplied with the time-resolved contrast-enhanced MRA images at each time frame to form new 4D MRA images (TOA-enhanced HFMRA), which contains the contrast arrival times with defined color encoding. In the second method, each time frame was weighted by a Gaussian distribution in the time domain to form a virtual 4D bolus map. This 4D bolus map was then color-coded and multiplied with the HFMRA images to form a digital subtraction angiography (DSA)-like VB, where at each time frame, only vessels with certain TOA values within the defined bolus length appear. HFMRA, TOA maps, and VB images were scored qualitatively with regard to delineation of arteries, veins, and nidus, as well as artifacts. Furthermore, diagnostic confidence and arteriovenous overlap were evaluated and compared between techniques. A comparison with DSA was performed where DSA served as the reference standard in terms of number of arterial feeders, draining veins, and Spetzler-Martin score of AVMs. In addition, TOA maps were evaluated quantitatively.
Results: Overall, diagnostic confidence score of TOA was significantly higher compared with that of HFMRA (P = 0.03). Virtual bolus showed significantly higher scores for overall diagnostic confidence (P = 0.02) and reduced arteriovenous overlap (0.01) compared with HFMRA. Furthermore, VB-reduced arteriovenous overlap scores were significantly higher compared with TOA (P = 0.04). Agreement regarding AVM draining veins was lower between DSA and HFMRA (κ = 0.3) compared with TOA and VB (κ = 0.56). Agreement regarding Spetzler-Martin score was lower between DSA and HFMRA (κ = 0.56) compared with TOA and VB (κ = 0.74).
Conclusions: TOA-enhanced HFMRA provides serial images and time of arrival maps in one inclusive display. In this study, TOA mapping combined with Virtual Bolus imaging improved diagnostic confidence in AVM patients and facilitated arteriovenous separation. The VB method further reduced overlap of arterial and venous structures.