Deconvolution is a mathematical technique used to improve radionuclide angiocardiography after suboptimal injection of radiopharmaceutical. A new deconvolution algorithm designed to be relatively insensitive to the random errors that occur in experimental data was tested. First-pass radionuclide angiocardiography using iridium-191m was performed to quantitate left-to-right shunting in normal dogs and dogs with atrial septal defects. Deconvolution was used to correct for injection shape. Four quantitation techniques were studied (good injection/no deconvolution, good injection/deconvolution, fragmented injection/no deconvolution, fragmented injection/deconvolution). The mean (p less than .001) and standard deviation (p less than .0001) of the fragmented injection/no deconvolution technique were significantly different from the other three techniques, which were not significantly different from each other (mean or standard deviation) at the p = 0.05 level. This deconvolution method made it possible to accurately quantitate left-to-right shunts even with fragmented injections.