On the value of geometry-based models for left ventricular volumetry in magnetic resonance imaging and electron beam tomography: a Bland–Altman analysis

Abstract 

Objective: Methodological comparison of ellipsoid model-based approaches and Simpson method to evaluate left ventricular volumetric parameters by magnetic resonance (MR) and electron beam tomography (EBT) and analysis of the origin of possible discrepancies. Methods and material: 100 subjects (87 patients, 13 healthy volunteers) were studied in MR in various cardiac views and EBT long axis view to determine left ventricular volumes and masses by applying (rotational) ellipsoid and Simpson model. Observer variation and method agreement was quantified by means of variance component and Bland–Altman analysis. Results: Simpson approach showed smaller observer variability than all ellipsoid approaches. All geometry-based models gave smaller left ventricular volumes than Simpson approach, the bias in mass determination was minimal. Whereas high correlation coefficients (typically 0.85–0.95) for left ventricular volume and mass measurements indicated satisfying correspondence between methods, large 95% limits of agreement made a transfer of results for single subjects between Simpson and ellipsoid approaches difficult and between different geometry-based models almost impossible. Because 95% limits of agreement and observer variability of geometry-based approaches were of equal order, the latter could be identified as main limiting factor of methodological agreement. Conclusion: MR Simpson approach is superior to all ellipsoid model-based approaches, because observer variability is smaller.

Keywords:  Heart, Ventricular function, Magnetic resonance imaging, cine, Tomography, computed, Statistics