Truncation artifact reduction in magnetic resonance imaging by Markov random field methods

Area of Science:

• Statistics
• Signal Processing
• Medical Imaging

Background:

• Truncation artifacts are a common problem in Fourier series reconstructions.
• Existing methods may not adequately account for signal characteristics or experimental noise.
• Magnetic Resonance Imaging (MRI) data often suffers from these artifacts.

Purpose of the Study:

• To develop a novel statistical method for reducing truncation artifacts in Fourier series reconstructions.
• To incorporate Bayesian principles to account for signal characteristics and experimental noise.
• To provide an automated parameter selection solution for the proposed model.

Main Methods:

• A Bayesian approach using Markov random fields for function modeling.
• Incorporation of errors from Fourier series truncation and experimental noise.
• Application of Monte Carlo Markov chain methods for statistical inference.
• Development of an automatic parameter selection strategy.

Main Results:

• Successfully reduced truncation artifacts in both simulated and real magnetic resonance images.
• Demonstrated the ability to model specific function characteristics.
• Showcased the integration of experimental noise into the reconstruction process.
• Validated the effectiveness of the automatic parameter selection.

Conclusions:

• The proposed Bayesian statistical method offers an effective solution for reducing truncation artifacts.
• The method is robust and applicable to real-world data, such as in MRI.
• The integration of function characteristics and noise handling improves reconstruction quality.