Study On Rapid Cardiac Magnetic Resonance Imaging Based On Parallel Reconstruction And Deep Learning

Cardiac magnetic resonance（CMR）imaging technology can non-invasively obtain cardiac anatomical structure,cardiac function and other data,and is widely used in non-invasive clinical examination.However,the long scanning time has become an important reason for hindering the development of cardiac magnetic resonance technology.Therefore,accelerating the imaging speed while ensuring the quality of image reconstruction is the key issue of cardiac magnetic resonance imaging research.This paper introduces the principle of magnetic resonance imaging,wave coding technology and related fast imaging algorithms,and provides background knowledge for the application of parallel imaging in Wave-CAIPI and the online reconstruction of deep low-rank plus sparse network（L+S-Net）.In view of the problems existing in rapid cardiac magnetic resonance imaging,three solutions are proposed to speed up the imaging speed and improve the image reconstruction quality.The main work and contributions of this paper are as follows:In this paper,a parallel magnetic resonance imaging model based on iterative self-consistent parallel imaging reconstruction（SPIRi T）algorithm to reconstruct wave-coded balanced steady-state free precession（b SSFP）sequence is proposed.The combination of wave coding technology and SPIRi T image reconstruction method not only preserves the unique advantages of wave coding method,but also obtains robust reconstruction results based on SPIRi T framework.Using the reconstruction method based on sensitivity coding（SENSE）,the wave-coded b SSFP sequence was used for high-resolution single myocardial imaging.This paper quantifies and evaluates the reconstruction results of wave-coded b SSFP sequence in anatomical imaging,T₂ prepared b SSFP sequence imaging and T₁ quantitative imaging.Compared with the reconstruction results of traditional b SFFP sequence,the reconstruction results of wave-coded b SSFP sequence show higher signal-to-noise ratio and less aliasing artifacts in cardiac imaging.In dynamic cardiac magnetic resonance imaging,many networks are limited in practical clinical applications due to problems such as slow data processing and imaging speed.This paper proposes a method of using CPU and GPU to accelerate depth L+S-Net imaging based on SigPy architecture,and deploys the reconstruction network on the reconstruction platform based on Gadgetron for online testing.Compared with other methods,this method can further reduce the reconstruction time while maintaining good image quality.In order to accelerate the speed of cardiac magnetic resonance imaging and improve the quality of reconstructed images,this paper proposes three solutions,which provide new methods and ideas for the clinical application of cardiac magnetic resonance imaging.