Improving the applicability of myocardial densitometry and parametric imaging by extended automated densogram analysis.

In clinical applications the analysis of X-ray contrast densograms acquired in regions of interest (ROI's) over the myocardium is disturbed by many complex factors. For this reason we acquire redundant densogram information for quality control before extracting densitometric parameters. In our approach, initially some stable measures of quality for densograms are used to lower the influence of poor quality densograms by a quality weighted averaging. For example a shape quality measure, Q1, is calculated using regions of optimal and minimal acceptable quality defined with respect to a prototype densogram. Not a few myocardial ROI's yield densograms that differ from single-source densograms (SSD's) due to e.g. superposition of different perfusion beds or the position of the ROI relative to the coronary sinus or stenoses. This might result in a densogram shape with oscillating or plateau behavior. For densograms of a such general shape many parameters defined in the usual way do not depend smoothly on the densogram values. The conventional definitions of some parameters (appearance time, rise time) are therefore extended for application to multi-maxima densograms as well as to SSD's. These new methods are evaluated using digitized clinical angiocardiograms and are applied to parametric imaging (pixeldensograms) in a slightly modified way. Taking into account the densogram quality, its shape and its origin results in a considerable improvement both for densitometry and parametric imaging of myocardial perfusion.