Aims: The accuracy of magnetic resonance angiography in detecting proximal coronary artery stenoses is unclear. We postulated that fast magnetic resonance angiography is capable of (1) imaging proximal coronary arteries, and (2) detecting stenoses of > or = 50% of their luminal diameter.
Methods and results: Thirty-five patients, referred for analysis of angina pectoris, underwent both conventional angiography and magnetic resonance angiography of coronary arteries. A fast k-space segmented gradient-echo technique was used during breath-holds. Two observers, blinded to the results of conventional angiography, independently analysed the magnetic resonance studies for (1) length of visualized segments, and (2) presence of signal voids indicative of stenoses. From 140 proximal arteries, 15 (11%) were excluded because of incomplete imaging or degraded image quality. Mean length of the visualized segments was 9 +/- 4 mm for the left main, 62 +/- 16 mm for the left anterior descending, 21 +/- 9 mm for the left circumflex and 89 +/- 32 mm for the right coronary artery. Sensitivity for detecting > or = 50% luminal diameter stenoses was 0.00 for the left circumflex, 0.53 for the left anterior descending coronary artery, 0.71 for the RCA and 1.00 for the left main artery. Specificity varied from 0.73 for the left anterior descending coronary artery to 0.96 for the left circumflex. Inter-observer agreement was 0.90.
Conclusion: Thus, segmented magnetic resonance angiography is capable of non-invasive imaging of proximal coronary anatomy. Its good accuracy in detecting left main coronary artery disease, intermediate accuracy in detecting right coronary artery and left anterior descending coronary artery stenoses, and low accuracy in detecting left circumflex lesions fit within a range of sensitivities and specificities found by others. Further technical advances are necessary to make the technique clinically robust.