Damiano Caruso MD, Presenter: Nothing to Disclose

Marco Rengo MD, Abstract Co-Author: Nothing to Disclose

Davide Bellini MD, Abstract Co-Author: Nothing to Disclose

Marco Maria Maceroni MD, Abstract Co-Author: Nothing to Disclose

Carlo Nicola De Cecco MD, Abstract Co-Author: Nothing to Disclose

Andrea Laghi MD, Abstract Co-Author: Research Consultant, Bracco Group Research Consultant, Vital Images, Inc Research Consultant, General Electric Company Research Consultant, iCAD, Inc

To compare left ventricular (LV) function assessment using three different software tools on the same dual source computed tomography (DSCT) datasets with the results of MRI as a standard references.

Eighty-seven patients (63 male, 24 female, mean age : 56.7 ± 12.9) undergoing DSCT and MRI, were included in this study. Reconstruction were made retrospectively at every 5% from 0% to 95% of the RR interval with 2,5mm slice thickness. Functional analysis were made on three different workstation. In all software we performed both automatic and semi-automatic method. Pearson's correlation coefficient was calculated to assess the correlation between the different DSCT software tools and MRI. Repeated measurements were performed by two blinded radiologists to determine intra-observer and inter-observer variability.

A low correlation between the automatic method and MRI results was found for all workstations. Correlation and Bland and Altman plots between DS-CT and MRI, obtained with the semiautomatic analysis on all WS were good (r2 = 0.81, 0.86 and 0.83 respectively). A constant error was found for all WS (+5%, -3% and +4%) for the automatic evaluation of EF. The mean time for DSCT automatic analysis was 285 ± 34 s, the mean time for semi-automatic analysis was 517 ± 67 s. Intra-reader agreement was good for all WS (wK=0.83, 0.79 and 0.81). Inter-reader agreement was significantly better for Leonardo and Aquarius than for Vitrea (wK=0.85 and 0.91 Vs wK=0.75).

We demonstrate that there were no significant differences between MRI values and DSCT values using a semi-automated analysis. Automated analysis instead was easier and time-saving than semi-automated analysis and it could be done either by radiologist than fellows; but left ventricular function values, using automatic analysis, showed significant differences with MRI.

The LV function is a predictive marker for morbidity and mortality in patients with CAD. This study investigates the variability of LV function assessment on three commercially available WS.

Caruso, D, Rengo, M, Bellini, D, Maceroni, M, De Cecco, C, Laghi, A, Evaluation of Global Left Ventricular Function with DS-CT Comparison of Three Workstations with Magnetic Resonance as Gold Standard.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034607.html