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

CETIR Sant Jordi, Barcelona, Spain. Abstract. Magnetic resonance imaging (MRI) simulators have beenlargely applied to brain studies. However, cardiac applications of thesesimulators are only emerging. This paper focuses on the realistic sim-ulation of local appearance of cardiac MRI datasets. Simulations areobtained from the MRISIM simulator with XCAT phantom (formerlyNCAT) as input. Phantoms are further extended to increase realism ofthe local appearance of the simulated datasets. The extension is basedon: resemblance of partial volume eect by using a higher resolutionphantom as input to the simulator, addition of intensity variability ofeach tissue by increasing the number of labels of the phantom, and in-clusion of trabeculae in the ventricular cavities. The clinical databaseincluded 40 patients for anatomical measurements and 5 healthy ath-letes for local grey value statistics. The virtual database included 20digital phantoms. Histograms from dierent tissues were obtained fromthe real datasets and compared to histograms of the simulated datasetsby means of Chi-square dissimilarity metric. The addition of sublabelsand trabeculae improved the matching of real histograms in 8 out of 11comparisons. Simulated intensity distributions were improved up to 76%with respect to the original distributions. Our methodology obtained ahigher dissimilarity metric for lung and pericardial tissue.