Thallium-201 myocardial scintigraphy: improved sensitivity, specificity and predictive accuracy by application of a statistical image analysis algorithm.

Exercise and 3 hour delayed redistribution thallium-201 myocardial perfusion imaging was performed in 107 patients including 87 patients with documented coronary artery disease and 20 patients with normal coronary arteriograms. A computer algorithm for statistical analysis and redisplay of the analog scintillation data was developed to augment the visual analysis of the standard analog image. The basic algorithm identified the myocardial area (pixel) with the most absolute counts and developed an appropriate standard deviation range from a table. All pixels in the area of the myocardium containing absolute counts below three standard deviation ranges (or approximately 6 standard deviations) were deleted from the final image. The remaining pixels were redisplayed in a digitized bimodal format and the image photographed. Two experienced observers compared visual analysis of the analog image alone and the analog image in conjunction with the computer-analyzed image for sensitivity, specificity and predictive accuracy in the detection of perfusion defects in patients with coronary artery disease and in normal subjects. Sensitivity in patients with coronary artery disease and in normal subjects. Sensitivity in patients with coronary artery disease for the analog scintillation image alone was 79 percent (69 of 87) and with computer analysis 95 percent (83 of 87). Specificity in the patients with no coronary disease was 100 percent (20 of 20) for both techniques. The predictive accuracy of the test was 83 percent (89 of 107) for the analog image alone and 96 percent (103 of 107) for the two images combined. It is concluded that use of a computer statistical analysis algorithm of thallium-201 analog myocardial perfusion images improves the accuracy of detection of perfusion defects in patients with coronary artery disease.