A constrained Gauss-Seidel method for correction of point spread function effect in MR spectroscopic imaging

Magnetic resonance spectroscopic imaging is Limited by a low signal-to-nois e ratio, so a compromise between spatial resolution and examination time is needed in clinical application. The reconstruction of truncated signal int roduces a Point Spread Function that considerably affects the spatial resol ution. In order to reduce spatial contamination, three methods, applied aft er Fourier transform image reconstruction, based on deconvolution or iterat ive techniques are tested to decrease Point Spread Function contamination. A Gauss-Seidel (GS) algorithm is used for iterative techniques with and wit hout a non-negative constraint (GS+). Convergence and noise dependence stud ies of the GS algorithm have been done. The linear property of contaminatio n was validated on a point sample phantom. A significant decrease of contam ination without broadening the spatial resolution was obtained with GS+ met hod compared to a conventional apodization. This post-processing method can provide a contrast enhancement of clinical spectroscopic images without ch anges in acquisition time. (C) 2000 Elsevier Science Inc. All rights reserv ed.