Research On Rapid MRI

Magnetic resonance imaging(MRI) is an imaging technology using magnetic resonance signal to form internal images of a subject. As an essential technology in medical imaging, MRI has advantages of no harmful radiation, multi-parameter imaging and high tissue sensitivity. However, MRI has a slow imaging speed and high susceptibility to motion artifacts. Hence, how to implement a rapid MRI is a key problem to MRI technology development. Compressed sensing(CS) is a theory based on the premise that signals are sparse or compressible. CS can reconstruct signals from significantly fewer measurements than were required by traditional Nyquist criterion. Because MR images are highly compressible and MR signals can be encoded, CS has a broad application prospect in MRI.In this article, the sparse representation based on wavelet, curvelet and contourlet are described and the CS reconstruction based on the nonlinear conjugate gradient(CG), the fast iterative shrinkage-thresholding algorithm(FISTA) and the fast composite splitting algorithm(FCSA) are studied. A new method which implements FCSA based on dual-tree wavelet transform is proposed for MRI. The combination of dual-tree wavelet provides an approximate shift-invariance and directional selectivity. Experiments show that the proposed method has a better performance on both the reconstruction quality and time.For rapid MRI technology based on K space, the combined MRI framework PF-CS-SENSE which combines CS, parallel imaging and partial fourier imaging is studied. Two new combined MRI methods which based on dual-tree wavelet and contourlet are proposed. Experiments show that both these two proposed method can provide a better reconstruction quality and need less reconstruction time.In addition, a new rapid combined MRI framework is proposed in this article. While the proposed framework uses efficient CS reconstruction algorithms to process all the sampled data, it uses directly zero-filled inverse fourier transform to process the symmetric low frequency data. Experiment show that the proposed rapid combined MRI framework can significantly reduce the reconstruction time but still maintain the reconstruction quality.