The reliability of automatic measurement of left ventricular function with dual-source computed tomography datasets

• Published in: European radiology Vol. 19; no. 12; pp. 2919 - 2930
• Main Authors: de Jonge, G. J., van Ooijen, P. M. A., van der Vleuten, P. A., Lubbers, D. D., Kasemier, J. H., de Bock, G. H., Oudkerk, M.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer-Verlag 01.12.2009; Springer Nature B.V
• Subjects: Cardiac; Diagnostic Radiology; Female; Humans; Imaging; Internal Medicine; Interventional Radiology; Male; Medicine; Medicine & Public Health; Middle Aged; Neuroradiology; Observer Variation; Radiography, Dual-Energy Scanned Projection - methods; Radiology; Reproducibility of Results; Sensitivity and Specificity; Tomography, X-Ray Computed - methods; Ultrasound; Ventricular Dysfunction, Left - diagnostic imaging; Heart; Function; Computed tomography; Software package; Dual-source computed tomography; Left ventricular function; Software comparison
• ISSN: 0938-7994; 1432-1084
• DOI: 10.1007/s00330-009-1506-3

The purpose of the study was to assess the reliability of (semi-)automatic left ventricular (LV) function measurements using three different software packages on the same dual-source computed tomography (DSCT) datasets and to compare agreement among the software packages. Forty consecutive patients, undergoing cardiac DSCT were included (31 men, mean age 58 ± 14 years). LV function analysis was performed with all three software packages. ANOVA testing was used to determine the difference among the repeated measurements and the difference among the software packages. Bland–Altman plots were computed to describe the agreement among the software packages. No significant difference was found among the repeated measurements. In the comparison of the three software packages, a significant difference was observed when measurements were used with minimal user interaction. When end-diastolic and end-systolic phases were manually set, there was no overall significant difference, but in 12.5% of patients a large (>10%) difference in LVEF was found. All three software packages have good intraobserver variability, but the results of the three packages were significantly different. For clinical use, one should be aware of the clinical impact of possible segmentation flaws when (semi-)automatic LV function assessment is used.