A symmetric 4D registration algorithm for respiratory motion modeling.

Area of Science:

• Medical Imaging
• Computational Anatomy
• Image Processing

Background:

• Dynamic volumetric lung images present challenges for accurate analysis due to respiratory motion.
• Existing registration methods may struggle with the complexities of 4D (3D+Time) data and non-rigid deformations.

Purpose of the Study:

• To develop and validate an effective 4D image registration algorithm for dynamic volumetric lung CT images.
• To accurately model respiratory motion and enable precise interpolation of anatomical regions over time.

Main Methods:

• A deforming 3D model with continuous trajectory and smooth spatial deformation is constructed.
• The non-rigid transformation is represented using two 4D B-spline functions (forward and inverse parameterization).
• An objective function is minimized to penalize errors in intensity matching, feature alignment, spatial-temporal non-smoothness, and inverse inconsistency.

Main Results:

• The algorithm successfully constructs a deforming 3D model interpolating regions in 4D CT images.
• Demonstrated efficacy in respiratory motion estimation on public benchmarks and clinical lung CT data.

Conclusions:

• The proposed 4D image registration algorithm is effective for dynamic volumetric lung images.
• The method provides accurate respiratory motion estimation, showing potential for clinical applications.