Magnetic Resonance Imaging (MRI) is a non-ionising imaging modality which can provide excellent soft-tissue contrast because of a large number of flexible contrast parameters. One major limitation of MRI is its long acquisition time. Parallel MRI provides a framework to reduce the scan time.
The aim of this thesis is to investigate and develop new methods to improve the performance of Parallel MRI. A new GUI (Graphical User Interface) based platform is developed using Matlab which provides an interactive environment to apply different Parallel MRI algorithms as well as helps to analyse the results.
Regularization based reconstruction in Parallel MRI utilizes some prior information about the image to achieve better reconstruction results. The use of regularization in Parallel MRI is investigated and a new algorithm is proposed which uses wavelet-denoising of the coil sensitivity estimates before applying SENSE (a Parallel MRI algorithm). The results show that the proposed method is computationally efficient and offers a good alternative to regularization for lower acceleration factors (AF) in Parallel MRI.
A good choice of the regularization parameter in regularization based Parallel MRI reconstructions plays a pivotal role to have good results. A new algorithm to choose the regularization parameter efficiently has been developed. This method uses the g-Factor (noise amplification parameter in Parallel MRI) as a regularization parameter and provides better reconstruction results than the contemporary methods. The proposed algorithm improves the computational efficiency of regularization based reconstructions in Parallel MRI.
The use of Parallel MRI in interventional imaging can greatly reduce the time required for imaging. A novel catheter based phased array coil, composed of two independent coil elements has been developed. This phased array receiver coil can implement Parallel MRI. Some initial imaging experiments using this coil system have been performed and the results show a successful implementation of Parallel MRI on the acquired data.