K-Bayes reconstruction for perfusion MRI II: modeling and technical development

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Magnetic resonance imaging (MRI) is widely used, but perfusion MRI has limited resolution, hindering disease detection.
• Artifacts like partial volume effects and aliasing further degrade perfusion MRI quality due to k-space sampling limits and discrete Fourier transform (DFT) reconstruction.

Purpose of the Study:

• To develop a novel Bayesian modeling procedure (K-Bayes) for reconstructing perfusion MRI data.
• To improve the resolution, precision, and diagnostic utility of perfusion MRI by addressing limitations of standard DFT reconstruction.

Main Methods:

• Developed the K-Bayes approach, integrating an MRI signal k-space process model with a Markov random field prior.
• Incorporated high-resolution segmented structural MRI information into the prior distribution.
• Validated K-Bayes using simulation studies and in vivo human brain perfusion MRI data.

Main Results:

• K-Bayes reconstructed images showed significant qualitative and quantitative improvements over DFT.
• Demonstrated reduced bias, increased precision, and greater effect sizes in K-Bayes reconstructed perfusion MRI.
• Achieved higher resolution in K-Bayes images compared to standard DFT reconstruction.

Conclusions:

• The K-Bayes method offers a substantial advancement for perfusion MRI reconstruction.
• This technique enhances the ability to detect subtle changes in brain perfusion, improving diagnostic capabilities for neurological conditions.