Instrument Variables for Reducing Noise in Parallel MRI Reconstruction.

Area of Science:

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

Background:

• Generalized Autocalibrating Partially Parallel Acquisition (GRAPPA) is a standard parallel MRI technique.
• Image quality in GRAPPA is often compromised by noise, especially at higher acceleration factors.
• Scanner-generated noise propagates through GRAPPA's fitting and interpolation, distorting reconstructed images.

Purpose of the Study:

• To analyze the noise generation mechanism in GRAPPA from a system identification viewpoint.
• To develop a novel framework for noise reduction in GRAPPA.
• To propose a new reconstruction method, instrument variables (IV) GRAPPA, to mitigate noise.

Main Methods:

• Analysis of GRAPPA noise using a noisy input-output system model.
• Development of a new framework based on the errors-in-variables (EIV) model.
• Implementation and validation of the instrument variables (IV) GRAPPA reconstruction algorithm.

Main Results:

• The proposed errors-in-variables (EIV) framework accurately models GRAPPA noise generation.
• Instrument variables (IV) GRAPPA demonstrates superior noise removal compared to conventional GRAPPA.
• Experimental validation on phantom and in vivo brain data confirms improved image quality.

Conclusions:

• The EIV framework offers a robust approach to understanding and addressing noise in GRAPPA.
• The IV GRAPPA method provides a significant improvement in image quality by reducing noise.
• The proposed methods have the potential for broader application in solving EIV problems for noiseless GRAPPA reconstruction.