On‐line evaluation of systolic performance by densitometry in digital left ventriculography

The angiocardiographic evaluation of left ventricular end-diastolic (LVEDV) and end-systolic (LVESV) volumes and ejection fraction (EF) is routinely performed by the area-length method (ALM) but may lead to erroneous results. Digital imaging in real time allows densitometric procedures of determining left ventricular (LV) performance to be applied alternatively. In this study, we present densitometric algorithms for the analysis of LVEDV, LVESV, and EF from digital image data, establish accuracy and reproducibility, and determine value and limitations in comparison with ALM in single-plane 30 degrees right anterior oblique (RAO) projection. A linear relationship between iodine depth and measured densities is mainly burdened with scatter radiation and beam hardening which reduce primary radiation and suppress iodine depth. However, facilities such as deconvolution and correction algorithms are capable of reducing these sources of error. In the present study, computer-analyzed contrast images of iodine-filled wedges and spheres showed a near-linear relationship between iodine depth between 50-100 mg/cm2 and measured densities. Contrast images of heart casts and LV angio-grams of 54 patients were obtained with a digital image acquisition and processing system, and evaluated by two independent observers. The phantom study resulted in significantly (p < or = 0.01) better densitometric standard errors of estimate for volumes 3.3 ml densitometry (DENS) vs. 8.9 ml (ALM) and simulated EF 4.3% (DENS) vs. 7.8% (ALM) than ALM. The standard error of estimate for the comparison between both methods was 8.4 ml for volumes and 7.5% for EF. Densitometric volumes tended to underestimate volumes calculated by ALM. The angiographic study of patients demonstrated significant correlations between both methods (LVEDV r = 0.78, LVESV r = 0.83, total volumes: r = 0.89; EF r = 0.88). The standard errors of estimate can be ascribed to systematic, method-related errors of both DENS and ALM (LVEDV +/- 28.9 ml, LVESV +/- 23.4 ml, total volumes (EDV and ESV) +/- 27.1 ml; EF +/- 8.1%). The intra- and interobserver variability, respectively, exhibited significantly smaller (p < or = 0.01 and p < or = 0.05, respectively) standard errors of estimate for densitometric EF 4.6% (DENS) vs. 8.5% (ALM) and 7.1% (DENS) vs. 10.3% (ALM), respectively. Inclined but not significant differences were found for LVEDV and LVESV. In conclusion, the data presented indicate that the calculation of LV volumes and EF in digital left ventriculography may be performed accurately by densitometric calculation in single-plane 30 degrees RAO projection. Minor underestimations in densitometric volume determination may be anticipated in the evaluation of LV geometry.