The location and extent of myocardial infarction (MI) are important predictors of patient course. The current study tests the hypothesis that MI size could be measured accurately using rotating gamma camera single-photon emission computed tomography ( SPECT ) and technetium-99m pyrophosphate (PPi) and that the accuracy of these measurements was independent of MI location and transmural or nontransmural distribution. SPECT was performed in 38 dogs 48 hours after ligation of the left anterior descending coronary artery (14 dogs) or left circumflex coronary artery (LC) (24 dogs) at the mid-level or below. Projection images were corrected for center-of-rotation and field nonuniformity and processed with a 1-dimensional low-pass filter to diminish rib activity. Sixteen 0.5-cm-thick transverse sections, including the entire left ventricle, were reconstructed by filtered backprojection , low-pass filtered, contrast enhanced and processed with a 3-dimensional boundary enhancement operator. The boundary of PPi uptake in each slice was marked automatically using an algorithm that combined a directional derivative and a threshold, and required continuity of the boundary in 3 dimensions. The total number of volume elements that showed abnormal tracer uptake were summed, corrected to absolute volume, and multiplied by the specific weight of cardiac muscle. Scintigraphic MI weight was compared with pathologic MI weight. There was an excellent correlation between scintigraphic and pathologic MI weight. The poorer correlation for nontransmural compared with transmural MIs is most likely a function of size alone, since MIs that weighed less than 10 g (n = 12, range 1.3 to 9.5 g), both transmural and nontransmural, showed a similar correlation: S = 1.07 X P + 0.56 (r = 0.81, standard error of the slope = 0.245).(ABSTRACT TRUNCATED AT 250 WORDS)