Calcium score: semiautomatic calculation using different vendors versus fully automatic software

Recent advances in technology have resulted in a multitude of cardiac imaging postprocessing software products from vendors unrelated to the scanner on which the cardiac study was initially performed. A fully automatic calcium score software has also become available. We assess the intervendor variability of calcium score measurement using the semiautomatic software provided by the scanner vendor versus an unrelated vendor versus fully automatic software.

All consecutive patients who had a calcium score performed from March 2007 to January 2008 were included in this study. The studies were performed on two 64-slice computed tomographic scanners from 2 different vendors. The allocation of the patient to the specific scanner was done according to scanner and technologist availability. The studies were read twice by a cardiac radiologist and a general radiologist with at least 3 months' interval at a workstation specified by the computed tomographic scanner vendor and then at an unrelated workstation, with semiautomatic software. Calcium score was also independently performed by the fully automatic software, blinded to the results of previous readings. Agreement was tested with Pearson correlation coefficient, Bland-Altman graphs, and the Fleiss κ test.

The study population included 101 patients: 70 patients scanned at 1 scanner and 31 at a different scanner. Intervendor variability for the 2 groups had κ = 0.98 ± 0.01 and κ = 0.96 ± 0.02; interobserver variability had κ = 1. Semiautomatic versus automatic variability showed κ = 0.88 to 0.94.

Because of very strong agreement between the calcium score measurements obtained by semiautomatic and fully automatic software by different vendors, calcium score measurements can be performed robustly at vendor-specific software, nonrelated software, or fully automatic software.