Synthetic magnetic resonance imaging revisited

Area of Science:

• Medical Imaging
• Computational Physics
• Statistical Modeling

Background:

• Synthetic magnetic resonance (MR) imaging aims to predict MR images across various parameters from observed scans.
• Existing methods suffer from poor performance without regularization or are computationally impractical for 3-D imaging.
• A need exists for computationally feasible and accurate synthetic MR imaging techniques.

Purpose of the Study:

• To develop a computationally practical and high-performing method for synthetic MR image generation.
• To incorporate spatial context into MR image synthesis.
• To address the limitations of existing synthetic MR imaging approaches.

Main Methods:

• Proposed a method incorporating Gaussian Markov random field (MRF) structure on transformations of relaxation times and proton density.
• Specified MRF structure using a matrix normal distribution.
• Modeled observed magnitude images with the Rice distribution and employed a one-step-late expectation-maximization algorithm for computational practicality.

Main Results:

• The proposed method is computationally feasible for implementation.
• Excellent predictive performance was achieved in generating synthetic MR images.
• The approach effectively accounts for spatial context in MR images.

Conclusions:

• The developed method offers a computationally practical solution for synthetic MR imaging.
• The integration of MRF and Rice distribution modeling leads to excellent predictive performance.
• This approach advances the field of synthetic MR image generation in clinical settings.