Application of partial differential equation-based inpainting on sensitivity maps

Area of Science:

• Medical Imaging
• Image Processing
• Computational Science

Background:

• Partial differential equation (PDE)-based inpainting is a powerful image interpolation technique.
• Coil sensitivity maps are crucial for Magnetic Resonance (MR) parallel imaging.
• Existing methods for sensitivity map generation often struggle with extrapolation and hole filling.

Purpose of the Study:

• To introduce a novel PDE-based inpainting model and numerical method for MR parallel imaging.
• To apply this model for improving the accuracy and completeness of coil sensitivity maps.
• To address challenges of extrapolation and hole filling in MR image reconstruction.

Main Methods:

• Development of a novel PDE-based inpainting model and its numerical solution.
• Application of the inpainting model to generate and refine coil sensitivity maps.
• Comparison of inpainted sensitivity maps with Thin-Plate Spline (TPS) and Gaussian Kernel Smoothed (GKS) methods using phantoms and cardiac MR images.

Main Results:

• The proposed inpainting technique accurately determines coil sensitivity maps for phantoms and cardiac MR images.
• Images reconstructed using inpainted sensitivity maps show superior quality compared to those using GKS.
• The inpainting method achieves comparable results to TPS but with significantly reduced computation time.

Conclusions:

• PDE-based inpainting offers an effective solution for generating accurate coil sensitivity maps in MR parallel imaging.
• The novel model successfully handles extrapolation and hole filling, improving image reconstruction quality.
• This approach presents a more efficient alternative to existing methods like TPS for sensitivity map estimation.