A Robust Edge Detection Method for Gated

We present a myocardial edge detection technique that was devel oped for fast, reproduciblemeasurementsof left ventricularejection fraction in the clinical setting. Methods This myocardial edge detection method compares three edge parameters—countampli tude and first and second count derivatives—inthree consecutive locations along a radius to a predetermined template of these values.Eachofthe radii,definedat 10-degreeintervals,hasdifferent templatevaluesthat permitaccurateedgedetectioneventhough adjacent structures, such as the left atrium and the right ventricle, alteredge parameters.The template for edge detection is based on either the average edge parameters determined from manually defined edges in 15 patients (automatic method) or an operator defined edge in the first frame (semiautomatic method). Results: The edge detection methods were tested in 100 patients, and intraobserverand interobservervariabilitiesas wall as comparison with clinicallyobtained ejectionfractions were calculated.The stan dard error of the estimate was less than 3.1 % for all observer comparisons. In 15 patients with both high-count (400,000counts per image) and low-count (50,000 counts per image) studies, the mean absolute difference in ejection fraction was 2.6 % for intraob server comparisons. Conclusion: A robust myocardial edge detec tion technique was developed that is applicable for routine clinical use. Key Words: gated blood-pool scintigraphy; left ventricular ejection fraction; reproducibilit