Adaptive reconstruction of phased array MR imagery.

Area of Science:

• Medical Imaging
• Signal Processing
• Magnetic Resonance Imaging (MRI)

Background:

• Multicoil Magnetic Resonance (MR) imaging is crucial for high-resolution scans.
• Current reconstruction methods like sum-of-squares can introduce noise and artifacts.
• A need exists for advanced reconstruction techniques that improve image quality without requiring prior coil information.

Purpose of the Study:

• To develop and evaluate an adaptive spatial matched filter for phased array MR image reconstruction.
• To assess the filter's ability to improve signal-to-noise ratio (SNR) and reduce noise in MR images.
• To demonstrate the technique's effectiveness in handling artifacts without a priori knowledge.

Main Methods:

• An adaptive spatial matched filter was implemented using locally relevant array correlation statistics.
• Sample correlation matrices were derived from individual coil images for NMR signal and noise.
• Eigen-analysis was employed to determine an optimal filter vector based on estimated correlation statistics.

Main Results:

• The adaptive filter achieved SNR performance close to the optimal matched filter.
• Root-mean-square (RMS) noise in dark regions was reduced by up to the square root of N (number of coils).
• The method effectively suppressed localized motion and flow artifacts in MR images.

Conclusions:

• The adaptive spatial matched filter offers near-optimal reconstruction of multicoil MR imagery.
• It significantly reduces noise and suppresses artifacts, outperforming traditional methods like sum-of-squares.
• This technique provides a robust solution for MR image reconstruction without requiring prior knowledge of coil parameters.