RESUMO
A quantitative method for the analysis of 201thallium myocardial scintigrams, developed in an experimental infarcted dog heart model, has been compared with two nonquantitative methods for interpretation of stress myocardial scintigrams in two groups of patients studied with coronary angiography: 11 with normal coronary arteries and 14 with coronary artery disease. Three independent observers interpreted scintigrams which were 1) not computer processed; 2) corrected for background activity in lungs and chest wall; and 3) processed by a computer method which uses a uniform threshold of counts determined from the dog model to define perfusion defects. Interobserver variability as well as sensitivity and specificity of detecting coronary disease were examined. In patients with coronary artery disease interobserver variability was improved by using the computer technique: observers agreed as to the existence of a perfusion defect in 93% of the scintigrams as compared to 55% and 81% for the unprocessed and background-subtracted images respectively. No false positive indications of coronary disease were obtained by any of the three techniques. Use of the computer method did not improve the sensitivity of detecting coronary disease, however--71% compared to 64% for unprocessed images and 79% for background-substracted images. The advantages of this quantitative computer method are increased consistency of interpretation and lack of false positive diagnoses of coronary disease. An improved sensitivity of detection may be gained by varying thallium count thresholds according to anatomic location in the heart.