The purpose of this work was to develop a physical phantom for testing the accuracy of cardiac wall motion estimation algorithms, and to use the phantom to evaluate several motion estimation and reconstruction methods, including a simultaneous image reconstruction/wall motion estimation algorithm we have developed. Our approach was to attach radioactive markers to the myocardial wall of a dynamic cardiac phantom and to trace the motion of the markers throughout the cardiac cycle via gated SPECT acquisition. Then, without moving the phantom, and after the markers were allowed to decay to negligible levels, the myocardium was injected with 99mTc and a gated SPECT scan was acquired. From the gated myocardial emission data, two wall motion estimation methods were evaluated. The first method was by applying an optical flow algorithm to an optimized OSEM reconstruction of the myocardial emission data. The second was by applying our simultaneous image reconstruction/motion estimation algorithm to the myocardial emission data. The error in the estimated motion fields was described by the average Euclidean distance between the motion of the markers and the estimated motion. Values of 0.15 and 0.14 were found for the average Euclidean distance for the optical flow method applied to OSEM and the simultaneous method, respectively. Image quality was also evaluated and, in agreement with our previous findings, the simultaneous method produced myocardial images with improved noise characteristics and better uniformity in terms of the activity distribution within the myocardium.
Available at: http://works.bepress.com/jason_parker/10/