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

Area of Science:

• Medical Imaging
• Spectroscopy
• Image Reconstruction

Background:

• Magnetic resonance spectroscopic imaging (MRSI) suffers from low signal-to-noise ratio, necessitating a trade-off between spatial resolution and acquisition time.
• Truncated signal reconstruction in MRSI introduces a Point Spread Function (PSF), which degrades spatial resolution and causes signal contamination.

Purpose of the Study:

• To evaluate post-processing techniques for reducing PSF contamination in MRSI.
• To assess the effectiveness of iterative deconvolution methods, specifically the Gauss-Seidel (GS) algorithm with and without non-negative constraints (GS+), in improving image quality.

Main Methods:

• Three post-processing methods were tested after Fourier transform reconstruction: deconvolution and iterative techniques.
• The Gauss-Seidel (GS) algorithm, with and without a non-negative constraint (GS+), was implemented and analyzed for convergence and noise dependence.
• The linear nature of PSF contamination was validated using a point sample phantom.

Main Results:

• The GS+ method demonstrated a significant reduction in PSF contamination compared to conventional apodization.
• This reduction in contamination was achieved without compromising or broadening the spatial resolution of the spectroscopic images.
• The study validated the linear property of contamination on a point sample phantom.

Conclusions:

• The GS+ iterative method is an effective post-processing technique for reducing PSF contamination in MRSI.
• This approach enhances contrast in clinical spectroscopic images without requiring modifications to the acquisition protocol or increasing examination time.
• The findings suggest that GS+ can improve the diagnostic utility of MRSI by providing clearer, higher-resolution images.