Search

A simple algorithm is developed to enhance the resolution of components of multimodal pulmonary radiohistograms. Based on the assessed right ventricular ejection fraction (RVEF), it deconvolves out the effect of indicator smearing in the right ventricle on the heart rate samples of the pulmonary radiohistogram, c(k). The output is the ideal curve, c(k)'=c(k)/RVEF - (1- RVEF).c(k-1)/RVEF, that would be obtained if RVEF equalled unity. Since the formula is not a recursive one, it does not suffer from error propagation pertinent to numerical deconvolution. Once RVEF is known, implementation of the algorithm is simple. If RVEF is not known accurately it may be replaced with its upper estimate, RVEF(up)>RVEF, yielding a partial deconvolution of the curve. When applied to studies of patients with left-to-right shunt the method improves the accuracy and lessens the interobserver variation of the Maltz-Treves method.

Radionuclide-angiography (RNA) left-to-right intracardiac-shunt quantification algorithms, based on the part-by-part fit technique and the use of a so-called gamma variate model function (GVF), were tested via simulation analysis using data obtained from normal subjects. A good bolus of radioindicator was obtained by administering it directly into the vena subclavia. Normal subjects were defined as those having pulmonary histograms (PH) with no visible distortion caused by a shunt. Pure, non-superimposed data on the downslope of the PH curves, which are lost in presence of a shunt, proved to be appropriate reference values for testing the accuracy of results of standard shunt quantification algorithms. A generalized four-parameter GVF was introduced in order to extend the flexibility of the model function. The use of the three-parametric GVF to reconstruct the downslope of the PH curve out of the upslope data proved to be inadequate. This reveals an evident source of error in algorithms that calculate the shunt contribution by fitting GVF parameters to so-called difference-curve data. It is concluded that the inherent restricted statistical weight of RNA data prevents accurate results being obtained from standard RNA-shunt-assessment algorithms.