Bioautomation (Apr 2006)
A New Parameter for Improved Assessment of the Tracing Repeatability for Area Measurements of the Left Ventricle
Abstract
In order to validate an automated border detection (ABD) algorithm for obtaining a cross-sectional area of the left ventricle of heart, the areas obtained by ABD need to be compared with the areas obtained by a trained tracer as the latter's measurements are regarded as gold standard. Before performing an agreement analysis between the areas obtained by the ABD and that of the tracer, we need to know how well the tracer can repeat the tracing of the same image. A new parameter is proposed to assess the repeatability of this tracing. The results of three sets of 30x2 measurements have been analyzed. It was found that this new parameter can reveal the true difference in two tracings of the same image both spatially and numerically on a scale of 0 to 1. It can also yield the area differences between the ABD and traces.
