Abstract
107
Objectives Motion and partial volume (PV) artifacts due to cardiac motion are well known. These artifacts may lead to quantification errors in PET imaging. Therefore, to use a correction scheme to utilize all PET information whereas avoiding the artifacts is advisable. In this study a new combined method of cardiac motion and PV correction on 3D-PET data is presented which is based on mass conserving optical flow.
Methods The law of mass conservation holds true for cardiac PET images as the total activity uptake in the heart muscle remains constant during the cardiac cycle. Adding a smoothing term to this law to solve the under-determined system of linear equations leads to an optical flow functional. The minimization of this functional by using the corresponding Euler-Lagrange equations enables estimation of optical flow vectors. These can be used to correct the data for cardiac motion and PV effects.
Results 1) Correlation coefficient (CC), 2) myocardial thickness and 3) mean blood pool activity, were used to assess the quality of the method. A software phantom and 14 patient datasets were included in the study. The CC between the images target and all other phases increased from 0.83 to 0.998 on average. The myocardial thickness, which varies with the cardiac phase, was estimated by taking the FWHM of the profile curves through the heart muscle. After correction it was found to be corrected to within 0.1 mm of the target phase for all image volumes. Similarly, activity in the blood pool varies due to PV effects. The standard deviation in the blood pool activity, measured in a 100 voxel large ROI, among different cardiac phases was reduced from 410 to 27.
Conclusions A new method for combined PV and cardiac motion correction was presented. The method accurately functioned on all patient and phantom datasets as measured with three different criteria