An Efficient Algorithm for MR Image Reconstruction Without Low Spatial Frequencies

In conventional Fourier transform (FT) magnetic resonance (MR) imaging, the dynamic ranges of the electronic circuits, including the receiver, the preamplifier, and the analog-to-digital converter (ADC), are not used efficiently and may introduce signal distortion and quantization noise. If the image to he reconstructed is known to have some zero-valued pixels, the dynamic ranges can be better used by disregarding the largest signals and using signal restoration methods such as that proposed by Jackson and Macovski. We present here an improved method based on a least squares error (LSE) technique, and an efficient algorithm for signal restoration when the low-frequency components are discarded. In our method, the regions of support in both the image domain and the frequency domain can have arbitrary shapes, and all zero pixels in the image domain can be taken into account. The algorithm has been tested on simulated and experimental data. © 1990 IEEE