| Magnetic resonance imaging (MRI) has been considered as one of the greatest inventions in the last century. Compared with computer tomography (CT), MRI can provide images with high soft tissue contrast and spatial resolution and without harmful radiation. Therefore, MRI has been widely applied in clinics and maybe the most promising non-invasive diagnostic tool in medicine.However, long data acquisition time makes MRI susceptible to patient motion. As a result, blurring and ghosting caused by patient motion may reduce anatomic detail and limit the detection of pathological findings. It is reported that approximately 14% of subjects require sedation or general anesthesia in order to obtain satisfactory images, both of which carry significant risks. In functional MRI (fMRI), the observed blood oxygenation level dependent (BOLD) effect can be small in comparison to motion related intensity changes. Consequently, the extreme motion sensitivity of fMRI necessitates correction for motion effects in order to extract the proper signal from activated brain area. In addition, motions such as heart-beating, respiration, blood flowing and wriggling of stomach or intestines are more severe obstacles for MRI.PROPELLER (Periodically Rotated Overlapping ParallEl Lines with Enhanced Reconstruction) MRI is proposed in 1999 for motion correction by James G Pipe of St. Joseph's Hospital. This method has the advantages of oversampling near the center of k-space and extracting motion information from the overlapped data between strips. This method has been evaluated clinically for quantification and correction for head motion and found to be able to reduce motion artifact and improve image quality of T2-weighted MR images. In recent papers, PROPELLER has been extended to suppress motion in diffusion-weighted imaging (DWI) and diffusion tensor imaging (DTI) respectively. Furthermore, the method has been adopted by General Electrics (GE) Company. Nowadays, PROPELLER MRI is one of the mostadvanced techniques on its MRI devices and of promising future.However, PROPLLER MRI can only suppress motion in T2-weighted images with data collected by fast spin-echo (FSE) sequence etc. When used in Tl-weighted MRI, current reconstruction algorithm cannot produce satisfactory result due to imprecise estimation of motion. Additionally, the compensation for sampling density in final gridding reconstruction holds a high computation load.The data acquisition is accomplished on the OPEMMARK 4000 MRI system manufactured by Shenzhen Anke High-tech Company. The current reconstruction algorithm are implemented on the collected Tl-weighted PROPLLER data and it is found that new artifacts are created even for stationary objects. The reason is imprecise motion information estimated by registration algorithm based on correlation in frequency (CF). When perform rotation estimation, the phase of k-space data is not taken into account in order to exclude influence of translation, which account for the drop of accuracy and robustness.In fact, an image can be reconstructed by zero-padding each strip and performing the Fourier transform. Then the motion information can be extracted through registration of images. Therefore, a new algorithm for motion estimation is proposed based on maximization of the correlation in frequency domain and mutual information in spatial domain (CF-MIS) sequentially. Initial and coarse estimation is obtained by maximizing correlation in frequency. Then estimation is further improved by the registration based on mutual information in spatial domain.Fuzzy image enhancement is proposed to preprocess the temp images before motion estimation with MIS algorithm. With fuzzy image enhancement, the pixels with much higher or much lower intensity in the image are suppressed and the contrasts between different tissues are extended. As a result, the estimation accuracy is further improved.After motion compensation, PROPLLER data are resampled onto a Cartesian grid. During the gridding reconstruction, the current algorithm for sampling density compensation is computationally intensive. Postcompensation approach by gridding a unity data vector is proposed in this paper. Consequently, the time complexity is decreased greatly.In conclusion, the proposed CF-MIS algorithm is shown to be of higher accuracy than the CF algorithm and succeed in suppressing motion artifacts in Tl-weighted PROPELLER MRI. The proposed density estimation method can perform the sampling density compensation much more rapidly while achieving equivalent quality of reconstruction.
|
PDF Full Text Request