A method to assess spatially variant noise in dynamic MR image series

Area of Science:

• Medical Imaging
• Biophysics
• Signal Processing

Background:

• Accurate noise measurement in parallel Magnetic Resonance Imaging (MRI) is crucial but challenging, especially in vivo.
• Existing methods like subtraction or multiple acquisitions are often impractical due to motion and dynamic contrast variations.

Purpose of the Study:

• To develop and validate a novel, image-based method for accurate noise variance assessment in MR image series with temporal redundancy.
• To enable retrospective evaluation of noise levels and g-factor maps from multiframe MR data.

Main Methods:

• Utilized Karhunen-Loeve transform and random matrix theory.
• Fitted the probability density function of eigenvalues from the temporal covariance matrix to the Marcenko-Pastur distribution.
• Validated the method using numerical simulations, phantom experiments, and in vivo cardiac and brain imaging.

Main Results:

• The proposed method accurately assessed noise variance in simulated and experimental MR data.
• Successfully derived g-factor maps for phantoms, consistent with the multiple acquisition method.
• Demonstrated agreement with established subtraction and multiple acquisition methods in in vivo cardiac and brain datasets.

Conclusions:

• The novel method provides a practical and accurate approach for assessing spatially variant noise and g-factor maps in parallel imaging MRI.
• This technique is applicable to multiframe image series, offering retrospective noise evaluation capabilities for in vivo studies.