Bilateral filter regularized accelerated Demons for improved discontinuity preserving registration.

Area of Science:

• Medical Image Analysis
• Computational Imaging
• Biomedical Engineering

Background:

• Classical accelerated Demons algorithm uses L2 norm regularization with Gaussian smoothing.
• Gaussian smoothing struggles with discontinuities in deformation fields.
• Discontinuities are common in medical imaging, particularly in abdominal scans due to organ motion.

Purpose of the Study:

• To improve the accuracy of the accelerated Demons algorithm by addressing limitations in regularization.
• To introduce a novel approach using bilateral filtering to enhance deformation field smoothing.
• To achieve more realistic image registration by better handling discontinuities.

Main Methods:

• Replaced Gaussian filter in accelerated Demons with a bilateral filter.
• Bilateral filter incorporates image intensity information alongside displacement fields.
• Adjustable parameters in bilateral filter allow content-aware smoothing.

Main Results:

• Demonstrated significant improvements in Target Registration Error (TRE) on the POPI dataset.
• Achieved more realistic deformations, particularly in the presence of discontinuities.
• Validated on both 2D and 3D datasets, showing superior performance for abdominal data.

Conclusions:

• Bilateral filtering offers a more effective regularization strategy for accelerated Demons.
• The proposed method enhances registration accuracy by preserving important image features and discontinuities.
• This approach is particularly beneficial for medical datasets with complex motion, such as sliding organs.