Realistic Simulation of Local Image Appearance of Cardiac Magnetic Resonance Imaging Using a Virtual Phantom Population

Abstract : Magnetic resonance imaging (MRI) simulators have been largely applied to brain studies. However, cardiac applications of these simulators are only emerging. This paper focuses on the realistic simulation of local appearance of cardiac MRI datasets. Simulations are obtained from the MRISIM simulator with XCAT phantom (formerly NCAT) as input. Phantoms are further extended to increase realism of the local appearance of the simulated datasets. The extension is based on: resemblance of partial volume effect by using a higher resolution phantom as input to the simulator, addition of intensity variability of each tissue by increasing the number of labels of the phantom, and inclusion of trabeculae in the ventricular cavities. The clinical database included 40 patients for anatomical measurements and 5 healthy athletes for local grey value statistics. The virtual database included 20 digital phantoms. Histograms from different tissues were obtained from the real datasets and compared to histograms of the simulated datasets by means of Chi-square dissimilarity metric. The addition of sublabels and trabeculae improved the matching of real histograms in 8 out of 11 comparisons. Simulated intensity distributions were improved up to 76% with respect to the original distributions. Our methodology obtained a higher dissimilarity metric for lung and pericardial tissue.