Phase unwrapping with graph cuts optimization and dual decomposition acceleration for 3D high-resolution MRI data

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Processing

Background:

• Spatial phase unwrapping is crucial for magnetic resonance imaging (MRI) but faces challenges from low signal-to-noise ratio (SNR) regions and rapid phase variations.
• Global optimization methods offer accuracy but are computationally intensive, limiting their practical application.
• Greedy methods are faster but often less accurate, necessitating more efficient algorithms.

Purpose of the Study:

• To introduce dual decomposition acceleration to a three-dimensional (3D) graph cut-based spatial phase unwrapping algorithm.
• To significantly improve the computational efficiency of 3D graph cut-based phase unwrapping.
• To address the challenges posed by low SNR and rapid phase variations in MRI.

Main Methods:

• The phase unwrapping problem was framed as a global discrete energy minimization problem.
• Dual decomposition was employed to split the problem into overlapping subproblems, enhancing computational efficiency.
• The accelerated 3D graph cut method was validated using 3D multiecho gradient echo images from an agarose phantom and five brain hemorrhage patients, compared against an unaccelerated method.

Main Results:

• The proposed dual decomposition acceleration achieved up to an 18-fold increase in computation speed.
• The method demonstrated significant improvements in computational efficiency for 3D graph cut-based phase unwrapping.
• Experimental results confirmed the effectiveness of the acceleration technique.

Conclusions:

• Dual decomposition is a highly effective technique for accelerating 3D graph cut-based spatial phase unwrapping algorithms.
• This acceleration significantly enhances the computational efficiency, making advanced phase unwrapping methods more practical for MRI applications.
• The findings contribute to faster and more efficient image processing in medical imaging.