Comparison of four motion correction techniques in SPECT imaging of the heart: A cardiac phantom study

The aim of this study was to evaluate the accuracy of four different motion correction techniques in SPECT imaging of the heart. Methods: We evaluated three automated techniques: the crosscorrelation (CC) method, diverging sq uares (DS) method and two-dimensional fit method and one manual shift techn ique (MS) using a cardiac phantom. The phantom was filled with organ concen trations of Tc-99m closely matching those seen in patient studies. The phan tom was placed on a small sliding platform connected to a computer-controll ed stepping motor. Linear, random, sinusoidal and bounce motions of magnitu de up to 2 cm in the axial direction were simulated. Both single- and dual- detector 90 degrees acquisitions were acquired using a dual 90 degrees dete ctor system. Data were acquired over 180 degrees with 30 or 15 frames/detec tor (single-/dual-head) at 30 sec/frame in a 64 x 64 matrix. Results: The s imulated single-detector system, CC method, failed to accurately correct fo r any of the simulated motions. The DS technique overestimated the magnitud e of phantom motion, particularly for images acquired between 45 degrees le ft anterior oblique and 45 degrees left posterior oblique. The two-dimensio nal and MS techniques accurately corrected for motion. The simulated dual 9 0 degrees detector system, CC method, only partially tracked random Or boun ce cardiac motion and failed to detect sinusoidal motion. The DS technique overestimated motion in the latter half of the study. Both the two-dimensio nal and MS techniques provided superior tracking, although no technique was able to accurately track the rapid changes in cardiac location simulated i n the random motion study. Average absolute differences between true and ca lculated position of the heart on single- and dual 90 degrees detectors wer e 1.7 mm and 1.5 mm for the two-dimensional and MS techniques, respectively . The corresponding values for the DS and CC techniques were 5.7 and 8.9 mm , respectively. Conclusion: Of the four techniques evaluated, manual correc tion by an experienced technologist proved to be the most accurate, althoug h results were not significantly different from those observed with the two -dimensional method. Both techniques accurately determined cardiac location and permitted artifact-free reconstruction of the simulated cardiac studie s.