OBJECTIVE: The aim of the study was to develop and evaluate a registration-based motion correction algorithm as a method of reducing respiratory motion artefacts in myocardial perfusion imaging. MATERIALS AND METHODS: The NCAT software was used to build nine male and nine female computer simulations of myocardial perfusion imaging data, with different respiratory motions and left ventricular ejection fractions. Imaging data were generated at various time points throughout each cardiac cycle. The data were summed over each cardiac cycle, forward projected, normalized, noise added and reconstructed with and without motion correction. Motion correction was performed using an algorithm that aligns images within a projection using nonlinear registrations. A standard simulation with no respiratory motion was also generated for comparison. The algorithm was applied to the standard to determine its effect on images with no respiratory motion. RESULTS: The median difference in mean segmental counts compared with the standard was calculated for each simulation. The mean (range) of these values was 3% (1-6%), 14% (12-16%) and 28% (28-29%) for displacements of 1, 2 and 3 cm, respectively. The largest changes occurred inferiorly and anteriorly. Motion correction reduced these differences to 2% (0-4%), 5% (2-7%) and 7% (7-7%), respectively. The process of correcting for motion reduced the mean counts in all segments by 3% (1-5%). CONCLUSION: Artefacts resulting from respiratory motion are improved using our algorithm when motion is 2 cm or greater.
OBJECTIVE: The aim of the study was to develop and evaluate a registration-based motion correction algorithm as a method of reducing respiratory motion artefacts in myocardial perfusion imaging. MATERIALS AND METHODS: The NCAT software was used to build nine male and nine female computer simulations of myocardial perfusion imaging data, with different respiratory motions and left ventricular ejection fractions. Imaging data were generated at various time points throughout each cardiac cycle. The data were summed over each cardiac cycle, forward projected, normalized, noise added and reconstructed with and without motion correction. Motion correction was performed using an algorithm that aligns images within a projection using nonlinear registrations. A standard simulation with no respiratory motion was also generated for comparison. The algorithm was applied to the standard to determine its effect on images with no respiratory motion. RESULTS: The median difference in mean segmental counts compared with the standard was calculated for each simulation. The mean (range) of these values was 3% (1-6%), 14% (12-16%) and 28% (28-29%) for displacements of 1, 2 and 3 cm, respectively. The largest changes occurred inferiorly and anteriorly. Motion correction reduced these differences to 2% (0-4%), 5% (2-7%) and 7% (7-7%), respectively. The process of correcting for motion reduced the mean counts in all segments by 3% (1-5%). CONCLUSION: Artefacts resulting from respiratory motion are improved using our algorithm when motion is 2 cm or greater.