Field Map Estimation in MRI using Compressed Sensing Algorithm

For non-cartesian magnetic resonance imaging, like spiral imaging, field inhomogeneity could cause image blurring, especially for long readout time. General correction method required field map estimation. However, when images are in low spin density, the estimated field map suffers from noise. A regularized method which utilizes the physical feature that field map is spatial smoothing, is proposed to estimate field map with little noise. The field map estimated by regularized method only have good performance while the images in low noise level. Once image suffers from severe noise, an accurate field map is still hard to obtain. In reality, to shorten scan time in spiral imaging, we would decrease the number of interleaves of sampling. As results of that, Signal-to-noise Ratio (SNR) of image gets lower, and effect of B0 inhomogeneity becomes serious problem. In such situation, a better way to calculate field map is required. In this paper, we propose optimized field map estimation method which employs compressed sensing algorithm. Actually, recovery expected signal of compressed sensing (CS) algorithm is noise reduction process, which could be used to estimate field map when images are in low SNR. The experiments show that using Wavelet transform as regularization term could perform better when images are in low Signal-to-Noise Ratio (SNR). To improve calculated field map further, both Total Variation (TV) term and Waveform term as regularization term are adapted. The method in this paper promises great field map estimation.