New 3D phase-unwrapping method based on voxel clustering and local polynomial modeling: application to quantitative...

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Image Processing

Background:

• Quantitative Susceptibility Mapping (QSM) requires accurate 3D phase unwrapping.
• Existing methods struggle with noise, disconnected regions, and rapid phase changes.

Purpose of the Study:

• To develop a novel, accurate, and robust 3D phase-unwrapping algorithm.
• To address limitations of current methods in challenging QSM data.

Main Methods:

• Developed CLOSE3D, a 3D phase-unwrapping method using voxel clustering and local polynomial modeling.
• Explored 26-neighborhood for phase variation analysis and data clustering.
• Employed region-growing local polynomial modeling for intra-block, inter-block, and residual voxel unwrapping.

Main Results:

• CLOSE3D achieved <0.39% mean error ratio in simulations with severe noise and rapid phase changes.
• Significantly lower error ratios compared to region-growing and region-expanding methods at high noise levels (≥60%).
• Accurate phase unwrapping and quantitative susceptibility reconstruction in *in vivo* brain QSM data, even with artifacts.

Conclusions:

• CLOSE3D demonstrates high accuracy and robustness in 3D phase unwrapping.
• The method effectively handles challenging conditions like noise and open-ended cutlines.
• Enhances the reliability of 3D MRI applications, including QSM and susceptibility weighted imaging.