Scan-based volume animation driven by locally adaptive articulated registrations

Area of Science:

• Medical imaging and biomechanics
• Computer graphics and geometric modeling
• Computational anatomy

Background:

• Creating accurate 3D models of the human body from medical scans is challenging.
• Deforming and animating these models, especially articulated regions, requires robust correspondence and registration methods.

Purpose of the Study:

• To develop a complete system for anatomically accurate, example-based volume deformation and animation of articulated body regions.
• To address the correspondence problem across multiple in vivo volume scans of an individual.

Main Methods:

• A novel registration method using hierarchical volume structures and biharmonic clamped splines to constrain nonlinear optimization.
• Volume Blend Deformation (VBD) for pose variation approximation and initialization.
• Data-driven deformation in the volume domain facilitated by established cross-scan correspondence.

Main Results:

• Generation of occlusion-free, person-specific 3D human body models.
• Asymptotically accurate inner tissue deformations and realistic volume animation.
• Robust performance demonstrated on complex articulated regions like the hand and knee.

Conclusions:

• The developed system provides a robust and efficient method for creating patient-specific 3D models and animations from volume scans.
• The novel registration approach effectively handles complex articulated body deformations, overcoming limitations of previous methods.
• This work has significant implications for medical imaging, biomechanical analysis, and virtual reality applications.