Time-effectiveness, observer-dependence, and accuracy of measurements of left ventricular ejection fraction using 4-channel MDCT

Purpose: To evaluate the time-effectiveness, inter-observer variance, and accuracy of left ventricular ejection fraction (EF) measurements using retrospectively ECG-gated four-channel multi-detector row CT (MDCT) angiography in comparison with biplane cine-ventriculography. Materials and Methods: Twenty consecutive patients underwent retrospectively ECG-gated MDCT angiography and conventional coronary angiography with biplane ventriculography. Raw MDCT data were reconstructed at 0% - 90% of the cardiac cycle in increments of 10%. Ten geometrically identical multiplanar reformations parallel to the short axis of the heart were reconstructed in each patient. Three blinded readers segmented the left ventricle in the end-systolic and end-diastolic phase using standardized window settings in order to determine the EF. The EF was measured with biplane cine-ventriculography by two blinded readers and was compared with MDCT. The time needed for post-processing was recorded and the inter-observer agreement for both imaging techniques was assessed. Results: Mean post-processing time was 63 +/- 3 min per patient for MDCT and 5.5 +/- 1.2 min for ventriculography. MDCT and ventriculography showed a good correlation (r = 0.83, p < 0.0001) for measurement of the EF. Mean errors of EF measurements for the three MDCT readers compared with the mean of the ventriculography were - 6.3 +/- 6.6%, - 4.7 +/- 7.1% and - 4.6 +/- 5.7%, respectively. The mean differences between the three readers assessing MDCT were - 1.6 +/- 3.2% (reader 1 versus 2, r = 0.96), - 1.6 +/- 5.6% (1 versus 3, r = 0.95) and - 0.011 +/- 2.9 % (2 versus 3, r = 0.97, p < 0.0001). The mean differences between the two readers assessing ventriculography was 0.32 +/- 5.1 % (r = 0.88, p < 0.0001). Conclusions: MDCT correlates well with biplane cine-ventriculography but has the tendency to underestimate the left ventricular EF. Measurements using MDCT have a high inter-observer agreement, however, the time needed for additional MDCT data post-processing is still unacceptably long.