Real-time motion compensated patient positioning and non-rigid deformation estimation using 4-D shape priors

Area of Science:

• Medical Imaging
• Computational Anatomy
• Image-Guided Therapy

Background:

• Range imaging (RI) is used for patient positioning and respiration analysis in motion compensation.
• Current RI methods for positioning use rigid transformations, ignoring respiratory motion deformations.
• Existing RI respiration analysis requires slow non-rigid registration (several seconds).

Purpose of the Study:

• To develop a real-time framework using RI for joint respiratory motion compensated positioning and non-rigid surface deformation estimation.
• To improve accuracy and speed in patient alignment during procedures affected by breathing.
• To enable real-time free-form deformation tracking using 4-D shape priors.

Main Methods:

• A novel real-time framework based on range imaging (RI) is proposed.
• The method utilizes pre-procedurally obtained 4-D shape priors to guide intra-procedural alignment.
• It simultaneously estimates rigid-body table transformation and free-form deformation for respiratory motion.
• A GPU-based implementation achieves rapid computation times.

Main Results:

• The proposed method significantly outperforms conventional alignment strategies.
• Achieved improvements of 3.0x in rotation accuracy and 2.3x in translation accuracy.
• Real-time performance was demonstrated with computation times of 40 ms.
• The framework successfully performs joint positioning and non-rigid deformation estimation.

Conclusions:

• The developed RI framework enables real-time, accurate, and motion-compensated patient positioning.
• It effectively addresses the limitations of rigid-body transformations in current RI positioning techniques.
• The joint estimation of rigid and non-rigid transformations offers a significant advancement for image-guided interventions.