Wavelet transform in MRI data reconstruction
Glenn Elliott Yeager · ThinkTech (Texas Tech University) · 2015
Magnetic Resonance Imaging (MRI) is becoming a widely used method for non-invasively imaging biological tissues. The MRI technique, however, is slower and much costlier than its competing medical imaging technique, computed tomography (CT), which uses ionizing radiation. In the last 10 years, many improvements in parallel MRI (pMRI) techniques have been developed for fast acquisition of MRI data for making MRI a versatile research as well as clinical diagnostic tool. These pMRI techniques have the disadvantage of reducing the signal-to-noise ratio (SNR) and thus the quality of the reconstructed image. MRI data acquisition is an extremely complex process where radio frequency pulse sequences in a magnetic field allow recording of changes in the magnetic field from the protons in biological tissues in the Fourier domain known as k-space. A hybrid wavelet and Fourier encoding of the k-space has been shown to be successful in reconstructing high quality sparse images. Based on the compressibility of images in the hybrid wavelet domain, an average wavelet coefficient significance map can be generated and fast acquisition of any MRI data may be accomplished when combined with an appropriate pMRI technique. The feasibility of generating an average significance map from 20 brains from McGill University Simulated Brain Database has been demonstrated and validated.