Super-resolution Reconstruction for Tongue MR Images

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Anatomy

Background:

• Magnetic resonance (MR) imaging is crucial for studying tongue structure and motion in clinical medicine and research.
• Acquiring images in three orthogonal planes (axial, sagittal, coronal) provides detailed views but results in anisotropic resolution.
• Anisotropic resolution limits automatic volumetric analyses (segmentation, registration) and visualization of oblique slices.

Purpose of the Study:

• To develop a super-resolution reconstruction method for generating isotropic MR tongue image volumes.
• To improve the quality of tongue MR images for enhanced volumetric analysis and visualization.

Main Methods:

• A novel super-resolution reconstruction technique combining three orthogonal MR image stacks.
• Preprocessing steps including intensity matching and image registration.
• Markov random field optimization for data combination to achieve an isotropic volume.

Main Results:

• The proposed method successfully generated isotropic image volumes from orthogonal MR tongue datasets.
• Demonstrated superior performance on five clinical datasets compared to conventional reconstruction methods.
• Enhanced image quality suitable for advanced volumetric analyses and visualization.

Conclusions:

• The developed super-resolution method effectively overcomes the limitations of anisotropic resolution in MR tongue imaging.
• This technique offers improved volumetric analysis and visualization capabilities for clinical and research applications.
• The method shows significant potential for advancing the study of tongue biomechanics and related disorders.