Quantification of myocardial tissue kinetics from serial tomographic images is limited because of bidirectional cross-contamination of recorded counts between myocardium and blood for metabolic tracers with relative slow blood clearance. We have developed and validated a new deconvolution technique that permits calculation of spillover fractions derived from geometric measurements of the imaged cross section (wall thickness, chamber diameter) and the intrinsic resolution of the tomograph. Serial gated positron-emission computerized imaging (PCT) and a-v blood sampling across the heart were performed in five dogs for 45 min after i.v. C-11 palmitate (CPA) and in five dogs for 3 hr after i.v. F-18 deoxyglucose (FDG). Tracer concentrations in myocardial tissue and arterial blood were also measured in vitro. Uncorrected PCT tissue and blood concentrations correlated poorly with in vitro measurements. After correction for count crossover, the correlation for FDG in tissue was r = 0.99, for FDG in blood r = 0.97, and for CPA in blood r = 0.99. Deconvolution techniques applied to serial PCT images provide accurate noninvasive measurement of myocardial tracer concentrations and direct determination of the arterial input function required for measurements of myocardial metabolism.