Modeling of indicator dilution curves for measurment of bidirectional cardiovasculr shunts

A method for measurement of bidirectional cardiovascular shunt fractions from a single indicator dilution curve is presented. Linear system models corresponding to four basic anatomic configurations of bidirectional shunts are devised to represent the indicator dilution response of the cardio-pulmonary circulation. The specific congenital heart defects and persistent fetal pathway in which bidirectional shunts exist are discussed in terms of the basic anatomic configurations. The computer algorithms and digital processing techniques for analyses of the indicator dilution curves are presented in detail. Forty-one thermodilution curves recorded during 13 pediatric cardiac catheterization procedures were analyzed an the results compared to simultaneous oximetric determinations. The comparison of L-R shunt fractions (expressed in terms of pulmonary plus R-L shunt flows) as calculated by the two methods a standard error of 8.3 percent and a maximum difference of 15 percent. Error analysis of the indicator dilution method resulted in a standard deviation of 6 percent. The reproducibility of multiple determination showed a standard deviation of 5.1 percent. Comparison of the R-L shunt fractions (expressed in terms of systemic plus L-R shunt flows) as calculated by the present method and oximetry gave a standard error of 8.3 percent and a maximum difference of 13 percent. The standard deviation of differences in multiple determinations was 2.9 percent. Twenty thermodilution curves and 8 fiberoptic dye curves were recorded in 5 newborn lamb experiments. The analyses of the data showed the applicability of the method for measurement of shunts in cases involving persistent fetal pathways. In several of the catheterization procedures additional indicator dilution data was recorded enabling validation of the method by internal consistency consideration. A method for separation of the R-L component of the bidirectional shunt curve using deconvolution by an exponential function was validated using data from two individual case studies. A method for extrapolation of the R-L component in which cardiac artifact is present was derived in a third case. The applicability of the linear system modeling approach was demonstrated in a case involving a type IV truncus arteriosus. The patients from, which data was obtained in these studies ranged in age from 10 to 19 years, Congenital heart defects and accompanying pulmonary disorders representing all of the basic anatomical bidirectional shunt configuration were present in these patients. Proper model selection is important for accurate shunt quantification by either the thermodilution or oxygen methods. The error associated with improper selection is discussed. In cases where a single model for shunt configuration can not be clearly established, a range of possible shunt fractions must be calculated from two models. Two addition features of this dissertation are the derivation of a complete set of equations for calculation of bidirectional shunt fractions using oximetry and the development of a method for deconvolution of indicator dilution curves using a Wiener filter technique.