Evaluation of two "fast" software tools for the measurement of left ventricular volumes in retrospectively ECG-gated multidetector CT of the heart: biplane area-length method and "shape tracking" method]

Purpose: Evaluation of 2 "fast" software tools, a biplane area-length method, and a new automatic "shape tracking" method, for the calculation of left ventricle volumes (end-diastolic volume EDV, end-systolic volume ESV, stroke volume SV) and ejection fraction (EF) in retrospectively ECG-gated multidetector computed tomography. Materials and Methods: 60 contrast-enhanced cardiac CT datasets (16x0.75mm collimation, normal dose: 120-140 W, 400-500wmAs(eff) [n=44], low dose 120 W, 100mAs(eff) [n=16]) were acquired from eight intubated healthy pigs on different days. Images were reconstructed with a slice thickness and increment of 2 mm every 10% of the cardiac cycle. The LV function was evaluated via the common short axis method as the reference method, the biplane area-length method, and the automatic "shape tracking" method. In the latter a three-dimensional triangulated deformable surface model was used to segment the endocardial border of the left ventricle and to track its motion through the cardiac phases. The results were compared using the Bland-Altman-plot, the correlation coefficient, and the Wilcoxon test. Results: All 60 data sets could be evaluated with all three methods. Good correlations were found for left ventricular functional parameters for all data sets, the normal dose (ND), and low dose (LD) data sets between 0.65 and 0.89 for the "shape tracking" method and between 0.7 and 0.87 for the area-length method. The "shape tracking" method showed a mean overestimation of the EDV of 3.1 (LD, p = 0.38) to 4.3 ml (ND, p < 0.05), the SV of 4.0 (LD, p = 0.08) to 4.9 ml (ND, p < 0.05) and the EF of 1.3 (LD, p=0.16) to 2.0% (ND, p < 0.05). The EDV was underestimated between 0.3 (LD, p=0.7) and Uml (ND, p=0.08). The area-length method showed an overestimation of the EDV (6.6 to 6.7 ml [p < 0.05]), the SV (5.9 to 8.4 ml [p < 0.05]), the EF (1.2 to 3.0 % [p < 0.05]) and the normal dose ESV (0.6 ml [p = 0.74]). The low dose ESV was underestimated (1.3 ml [p = 0.21]). Conclusion: The left ventricular functional parameters acquired with both "fast" software methods correlate well with the common semi-automatic short axis method, and both methods significantly overestimate the EDV, SV and EF. Both methods can be used in low dose examinations.