Research And Online Dynamic Magnetic Resonance Imaging Technology Based On Pseudo-Polar Sampling

In the applications of Magnetic Resonance Imaging(MRI),the imaging time and the imaging quality are usually restricted to each other.Therefore,the problem of shorten the MRI time and improving the image quality has been a hot study issue in the MRI domain for many years.The new field of online dynamic MRI requires less imaging time.In this paper,the online dynamic magnetic resonance imaging technology is mainly studied.Two types of online MRI methods based on Kalman filter and predictive compensation are studied.The Pseudo-Polar trajectory was proposed to under-sampling the k-space for online dynamic MRI.GPU acceleration was used to fast the reconstruction and to achieve online imaging purpose.With dynamic MRI data for the heart and the larynx,the feasibility of the proposed method was verified through simulation experiments.This paper mainly includes below aspects:(1)Study the dynamic magnetic resonance imaging technique based on a Kalman filter model.The Kalman filter model was simplified based on the covariance diagonalization hypothesis applied on the parameters of the model.The Kalman filter model was combined with the SLAM(Selective Line Acquisition Mode)technology,and applied in the dynamic magnetic resonance imaging domain for the heart and larynx.The simulation experimental results show that,under the same under-sampling ratios,the online dynamic MRI method based on Kalman filter model has better imaging quality and longer imaging time than earlier view shared method.(2)Study the online magnetic resonance imaging technique based on MPOD(Motion Predicted Online Dynamic)algorithm.MPOD algorithm includes two parts,one is Motion Estimation(ME)and the other is Motion Compensation(MC).Using previous reconstructed three frames to predict the current frame,and the ARPS(Adaptive Rood Pattern Search)algorithm was used to match motion blocks.In order to suppress the block artifacts in ME,the MC algorithm named as OBMC(Overlapped Block Motion Compensation)was used to suppress the block artifacts.The residuals between current frame and predicted frame were reconstructed using an IST(Iterative Soft Thresholding)algorithm.The current frame was reconstructed using the predicted frame and residuals.The simulation experiment results show that the method based on the prediction method has better reconstruction quality and shorter imaging time than the previous method based on the direct difference calculation.(3)Study the online dynamic magnetic resonance imaging technology based on Pseudo-Polar sampling.Studying Pseudo-Polar sampling trajectory on Non-Cartesian coordinates for data acquisition.The sampling points on Pseudo-Polar sampling trajectory are located on the Cartesian grid,which can avoid traditional time-consuming gridding operation,and thus the Pseudo-Polar Fourier transform can be directly applied without gridding operation to reconstruct the images.The reconstruction time was shortened through reducing the iteration number of IST from the previous 10 times to 5 times,and the speed of the imaging could be further accelerated through the GPU acceleration.The experimental results show that,under same under-sampling rate,the proposed method can obtain better imaging quality than that of the Kalman filter model combined with SLAM method,MPOD method,or polar coordinates acquisition method.