A Comparison of the Noise Characteristics of Projection Reconstruction and Two-Dimensional Fourier Transformations in NMR Imaging

Human images of excellent quality have been obtained with Nuclear Magnetic Resonance (NMR), using two different techniques: projection reconstruction (PR) and two-dimensional Fourier transformations (2DFT). It has been shown in CT that the reconstruction algorithms will multiply the noise in the projections and change the noise spectrum. A comparison of the noise of the two techniques is presented under the constraint of equal imaging conditions using both computer simulations and data obtained with the UCSF human imager. The computer simulations show that the two techniques produce approximately equal signal to noise values when S/N ~ 15. Below that value small differences are observed. For both simulations and actual data, the 2DFT shows a uniform spatial frequency response while the PR method shows a roll-off at low spatial frequencies. The 2DFT method is shown to be more resistant to artifact formation.