Rationale and objectives: The study compared the performance of conventional endocardial and epicardial centroid algorithms with the new "myocardial" centroid algorithm in patients with anterior myocardial infarction. "Floating" endocardial or epicardial centroid algorithms, commonly used in tomographic imaging methods to assess regional motion, may misrepresent left ventricular regional myocardial function in the presence of markedly asymmetric left ventricular contraction.
Methods: A new centroid algorithm based on regional myocardial mass distribution was tested in 29 patients with a first anterior myocardial infarction and was compared with conventional centroid algorithms. Direct comparisons in 60 equal sectors at one midventricular level per patient were performed between electron beam computed tomography and technetium-99m sestamibi single-photon emission computed tomography. The thresholds of regional myocardial function used to define infarction were varied for regional ejection fraction from 20% to 40% and for regional wall thickening from 0 to 4 mm. Regression and Bland-Altman analysis were used to compare infarct size by regional myocardial function with infarct size by sestamibi single-photon emission computed tomography.
Results: The new myocardial centroid showed the least shift toward infarcted myocardium from diastole to systole and had the highest amplitudes of the measurement curves for regional ejection fraction and regional wall thickening. The optimal regional myocardial function thresholds for each centroid algorithm for regional ejection fraction were endocardial, 30% (R = 0.62; mean difference to sestamibi, -0.5% +/- 22.1% tomographic infarct size points); epicardial, 30% (R = 0.79; mean difference, 2.2% +/- 13.1% tomographic infarct size points); and new myocardial, 25% (R = 0.88; mean difference, -0.6% +/- 9.5% tomographic infarct size points). The optimal thresholds for regional wall thickening were endocardial, 1 mm (R = 0.70; mean difference, -2.2% +/- 14.3% tomographic infarct size points); epicardial, 1 mm (R = 0.78; mean difference, -4.6% +/- 12.7% tomographic infarct size points); and new myocardial, 2 mm (R = 0.71; mean difference, 2.1% +/- 14.1% tomographic infarct size points). The best agreement (R = 0.88) between electron beam computed tomography infarct size and sestamibi single-photon emission computed tomography infarct size was achieved with regional ejection fraction and the new myocardial centroid algorithm.
Conclusions: In asymmetrically contracting left ventricles, the new myocardial centroid algorithm is superior to conventional methods for tomographic analysis of regional myocardial function.