Isotropic Total Variation Regularization of Displacements in Parametric Image Registration

Area of Science:

• Medical Imaging
• Computational Anatomy
• Image Registration

Background:

• Image registration is crucial for analyzing medical image time-series, particularly for respiratory motion.
• Traditional Tikhonov regularization struggles with non-smooth displacements at organ interfaces.
• Accurate registration near sliding interfaces is challenging but vital for motion quantification.

Purpose of the Study:

• To develop and validate an isotropic Total Variation (TV) regularization method for accurate parametric image registration.
• To enable precise motion quantification and sliding motion detection in dynamic medical imaging.
• To overcome limitations of existing methods, such as the need for organ masks.

Main Methods:

• Isotropic Total Variation (TV) regularization applied to parametric displacement fields.
• Efficient numerical solution using the Alternating Directions Method of Multipliers (ADMM).
• Validation on four clinical databases of breathing motion (CT lung, MR liver).

Main Results:

• Achieved accurate whole-volume registration for breathing motion across diverse datasets.
• Demonstrated superior performance (10-40% reduction in TRE) compared to existing methods.
• Provided precise motion quantification and sub-pixel sliding motion detection (mean TREs < 1 mm).

Conclusions:

• The proposed TV-regularization method effectively handles non-smooth displacements at sliding interfaces.
• The method is robust, parameter-independent, and does not require organ masks.
• This approach offers a significant advancement in medical image registration for dynamic processes like respiration.