Objective: Global cardiac function assessment using multidetector CT (MDCT) is time-consuming. Therefore we sought to compare an automatic software tool with an established semi-automatic method.
Methods: A total of 36 patients underwent CT with 64 × 0.5 mm detector collimation, and global left ventricular function was subsequently assessed by two independent blinded readers using both an automatic region-growing-based software tool (with and without manual adjustment) and an established semi-automatic software tool. We also analysed automatic motion mapping to identify end-systole.
Results: The time needed for assessment using the semi-automatic approach (12:12 ± 6:19 min) was reduced by 75-85% with the automatic software tool (unadjusted, 01:34 ± 0:29 min, adjusted, 02:53 ± 1:19 min; both p<0.001). There was good correlation (r=0.89; p<0.001) for the ejection fraction (EF) between the adjusted automatic (58.6 ± 14.9%) and the semi-automatic (58.0 ± 15.3%) approaches. Also the manually adjusted automatic approach led to significantly smaller limits of agreement than the unadjusted automatic approach for end-diastolic volume (±36.4 ml vs ±58.5 ml, p>0.05). Using motion mapping to automatically identify end-systole reduced analysis time by 95% compared with the semi-automatic approach, but showed inferior precision for EF and end-systolic volume.
Conclusion: Automatic function assessment using MDCT with manual adjustment shows good agreement with an established semi-automatic approach, while reducing the analysis by 75% to less than 3 min. This suggests that automatic CT function assessment with manual correction may be used for fast, comfortable and reliable evaluation of global left ventricular function.