Adaptive Diffeomorphic Multiresolution Demons and Their Application to Same Modality Medical Image Registration with...

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Diffeomorphic demons ensure smooth and reversible image registration but require manual iteration setting, impacting results.
• Manual iteration control in diffeomorphic demons is a significant limitation for accurate medical image registration.

Purpose of the Study:

• To develop an adaptive diffeomorphic multiresolution demons method for automated iteration control in nonrigid medical image registration.
• To enhance the robustness and accuracy of medical image registration, particularly in cases of large deformations.

Main Methods:

• Implemented an optimized framework incorporating nonrigid registration and a diffeomorphism strategy.
• Designed a grey-value-based similarity energy function and incorporated adaptive iteration stopping criteria.
• Validated the method using synthetic and same-modality medical images, simulating large deformations (rotational distortion, extrusion).

Main Results:

• The proposed adaptive method achieved superior registration accuracy, evidenced by the highest normalized cross-correlation coefficient and structural similarity, and the lowest mean square error.
• Successfully performed medical image registration with large deformations, maintaining stable evaluation indexes despite increasing deformation strength.
• Outperformed active demons, additive demons, and standard diffeomorphic demons in quantitative analyses for same-modality medical image registration.

Conclusions:

• The adaptive diffeomorphic multiresolution demons method is effective and robust for nonrigid medical image registration, especially with large deformations.
• Automated iteration control significantly improves registration outcomes compared to manual settings.
• The proposed method offers a reliable solution for challenging medical image registration tasks.