To determine if the detection of coronary artery disease by dipyridamole-thallium imaging is improved by quantitative versus qualitative analysis, and combining quantitative variables, 80 patients with chest pain (53 with and 27 without coronary artery disease) who underwent cardiac catheterization were studied. Segmental thallium initial uptake, linear clearance, monoexponential clearance and redistribution were measured from early, intermediate and delayed images acquired in three projections. Normal values were determined from 13 other clinically normal subjects. When five segments per view were used for quantitative analysis, sensitivity and specificity were 87 and 63%, respectively, for uptake, 77 and 67% for linear clearance, 60 and 60% for monoexponential clearance and 62 and 56% for redistribution. Of the four variables, uptake and linear clearance were the most sensitive (p less than 0.01) and specificity did not differ significantly. Using three segments per view, the specificity of uptake increased (p less than 0.05) to 78% without a significant change in sensitivity (85%). With this approach, sensitivity and specificity did not differ from those of qualitative analysis (85 and 78%, respectively). Stepwise logistic regression analysis demonstrated that the best quantitative thallium correlate of the presence of coronary artery disease was a combination variable of "either abnormal uptake or abnormal linear clearance, or both." Using five segments per view, the model's specificity (85%) was greater than that of uptake alone (p less than 0.02), with similar sensitivity (92%). Using three segments per view, the model's specificity (93%) was greater than that of uptake alone (p less than 0.05) and of qualitative analysis (p less than 0.05), with similar sensitivity (85%).(ABSTRACT TRUNCATED AT 250 WORDS)