Image acquisition and reconstruction methods for fast magnetic resonance imaging.

摘要

Magnetic resonance imaging (MRI) is a medical imaging modality of great interest due to its non-invasive nature and excellent soft tissue contrast. However, one of the major challenges in MRI is image acquisition speed. To enable faster imaging speed using MRI, several different image acquisition and reconstruction methods were developed.;A variable-density k-space sampling approach has been used to reduce the sampling rate beyond the limit imposed by the Nyquist criterion by under-sampling in the high-spatial-frequency region. Variable-density sampling was applied to stack-of-spirals trajectories for fast 3D imaging. Experimental results show the effectiveness of this variable-density sampling method for 3D imaging.;The technique was then applied to the clinical problem of assessing perfusion in the lower limb, which requires fast time-resolved volumetric imaging to track the first-pass contrast enhancement. Normal volunteer and patient study results confirm the technique's capability to assess perfusion. A new processing method that deconvolves the arterial input function using convex optimization also shows promise for more quantitative perfusion assessment.;Although variable-density sampling has been used in various imaging schemes for fast imaging, the optimal sampling pattern has not been previously determined. Given prior information of the object, which is characterized by an orthogonal basis set, optimal sampling locations were determined and tested on phantom and in-vivo data sets.;Partial k-space reconstruction, which exploits the Hermitian symmetry of MRI data, is widely used for faster imaging in Cartesian sampling schemes. However, partial k-space reconstruction for spiral trajectories has long been difficult. A new method using projection onto convex sets (POCS) allows partial k-space reconstruction for spiral trajectories.;PILS is a fast imaging technique that utilizes the localized sensitivities of multiple receiver coils. Here, this method is improved by allowing under-sampling in the readout direction. Furthermore, multiple demodulation hardware is proposed for the implementation of this improved technique.;High resolution imaging using MR is very challenging due to problems such as low signal-to-noise ratio and motion. By using small surface coils with fast spiral imaging sequences, these difficulties can be overcome. The proposed technique provides excellent in-vivo skin imaging results with resolution approaching 78 x 78 x 500 mum 3.