Testing the distribution of nonstationary MRI data

Area of Science:

• Medical Imaging
• Signal Processing
• Statistical Modeling

Background:

• Magnetic Resonance (MR) image noise is commonly modeled as a Gaussian distribution in complex-valued image components.
• Validating this noise model is crucial for accurate image analysis and interpretation.
• Existing validation methods face challenges with time-varying signal intensities in MR images.

Purpose of the Study:

• To investigate and validate the accepted Gaussian noise model for MR images.
• To develop a procedure for hypothesis testing of the noise model.
• To extend the noise model to accommodate time-varying signal components and address drift.

Main Methods:

• Employed repeated hypothesis testing to validate the noise model.
• Extended the Gaussian noise model to include a time-varying mean to account for signal drift.
• Implemented a procedure for modeling, estimating, and removing the drift component.

Main Results:

• Demonstrated a procedure for validating the Gaussian noise model in MR images.
• Successfully extended the noise model to handle time-varying signal intensities.
• The implemented procedure showed consistency with the proposed time-varying noise model.

Conclusions:

• The Gaussian noise model is validated for MR imaging, even with time-varying signals.
• The developed procedure effectively models and removes signal drift for accurate noise analysis.
• This work provides a robust method for MR image noise characterization.