PSF model simulation study using a cardiac phantom

Partially separable functions (PSF) model has been proposed for cardiac MRI due to its ability to get high spatial-temporal resolution dynamic images without the need of breath holding and triggering. During the PSF reconstruction, the choice of model order (L) is very important. It directly affects image quality and the model's ability to capture the motion. However, how to choose the value of L accurately has never been investigated. In this work, we use a numerical cardiac phantom to study the effect of the model order, L, in image reconstruction, and try to find out the optimal model order by different approaches. Simulation result shows that the model order is determined by the number of object motional states and it can be directly calculated by using the ratios of the differences of singular values from the navigator dataset. This simulation results may give a guidance of PSF research in the future.