Cardiac Contraction Motion Correction in Gated Myocardial Perfusion SPECT Projection Domain

Abstract

Introduction: Cardiac contraction and respiratory movement are two main factors which degrade gated myocardial perfusion SPECT images quality by inducing image blurring and also cause quantification inaccuracies. We propose a non-rigid motion correction step before image summation to eliminate the influence of cardiac motion on MPS images.
Material and Method: In this work we studied the effect of our motion correction method on projection data both using simulation and patient data. The mathematical four-dimensional NURBS-based Cardiac-Torso phantom with  two different heart sizes (124 ml and 100 ml left ventricular cavities) were constructed for male and female, respectively.  SIMIND Monte Carlo simulation package was used to simulate a clinical Tc99-mibi perfusion SPECT acquisition protocol on the dual headed gamma camera (Philips Medical Systems, Cleveland, Ohio), with Low Energy High Resolution (LEHR) collimators. We have warped each frame from, with respect to end diastolic frame, using a 3D dimensional non-parametric diffeomorphic algorithm based on Thirion demon registration technique in MATLAB 2015. 4D sequence of motion corrected projections and original projection data were reconstructed using FBP and OSEM algorithm in AutoRecon application in Cedars Sinai package. Myocardial thickness, myocardium to blood pool contrast and CNR were measured in summed images before correction and motion corrected (BC and MC respectively). Quantitative study of Image sequences were conducted after importing image sequences to cardiac SPECT system’s work station. BC and MC summed image parameters were automatically derived and calculated on the basis of sex specific normal limits obtained from the healthy population using standard Cedars-Sinai software.
Result: Myocardial wall thickening between end-diatolic to end systolic frame, in lateral wall is reduced 11% from BC images to MC images in patient study which is statistically significant. Myocardium to blood pool contrast in MC images are greater than the corresponding summed images before correction (BC) (27.10±11.07 vs. 32.55±12.07), CNR has increased 31% in MC images compared to corresponding BC images. TPD (Total Perfusion Defect) measure before and after correction were highly correlated.
Conclusion: Motion correction in projection domain may increases image quality by reducing blurring due to cardiac motion and possibly intra frame motion. Higher myocardial to blood pool contrast and CNR obtained in this study. It is highly recommended that with correction we can improve defect detectability and diagnostic value of images, further assessment with large patient population is under evaluation.