Cardiac Image Modeling Tool for Quantitative Analysis of Global and Regional Cardiac Wall Motion

Abstract

To evaluate the Cardiac Image Modeling (CIM 4.6; University of Auckland, Auckland, New Zealand) tool's ability to assess cardiac function via quantitative calculations of global and regional ejection fraction (EF) from magnetic resonance imaging in comparison with a current method of global analysis with Argus (Siemens Medical Solutions) and regional analysis with visual analysis. Global cardiac function is commonly assessed quantitatively by post processing tools that calculate global EF. Currently, regional cardiac function is assessed by subjective visual analysis of wall motion, which can have significant interobserver variability. CIM is a tool that may reduce variability by generating a semi-automated 3-dimensional heart model to calculate quantitative global and regional EF. Thirty-one patients (22 men, 9 women; mean age 55.1 +/- 17.5 years) were selected based on global EFs calculated at the time of the clinical visit with the Argus postprocessing tool (Siemens Medical Solutions). Patients were then placed into 2 predetermined categories of normal: EF >or=50% and abnormal: EF <50%. Regional EF was calculated for each segment of a 16-segment cardiac model. Three blinded reviewers used the standard of care assessment of regional function, which was a qualitative grading of the 16 segments into categories of normal or abnormal regional wall motion by visual analysis. CIM quantitatively analyzed global EF and regional EF for each segment. These segments were then sorted into the predetermined categories of normal (EF >or=50%) and abnormal (EF <50%). Level of agreement was conducted via Pearson correlation coefficient and Bland-Altman analysis for global EF analysis and observed proportion of agreement (p(a)), sensitivity, and specificity for regional EF analysis. Global EF analysis showed a high correlation (r2 = 0.85; y = 0.94x + 4.85, P < 0.001) between the Argus and CIM analyses. Sixteen-segment regional EF analysis showed p(a) averages >0.60. Regional wall motion by short axis slices showed pa averages >0.75, and combined analyses of all 3 reviewers' 16-segment regional data showed an overall total p(a) = 0.79 (sensitivity = 72%, specificity = 88%). Interobserver and intraobserver variability were low (p(a) > 0.65) in this study. Global EF analysis of cardiac magnetic resonance imaging by CIM showed high agreement with the commonly used Argus postprocessing tool. Furthermore, CIM is capable of evaluating regional EF with good agreement in comparison with the current visual method. In addition to determining abnormal versus normal cardiac wall motion, CIM is able to add to the analysis a quantitative regional EF for each given segment. As a semi-automated tool, CIM has the potential to reduce reviewer variability and decrease the time required for analysis. In the future, CIM can potentially quantitatively track global and regional changes in patients with heart disease and aid the clinical management throughout the course of the disease.