Lung respiration motion modeling: a sparse motion field presentation method using biplane x-ray images.

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• Respiratory motion introduces significant uncertainty in lung lesion localization for biopsies.
• Existing statistical models struggle to accurately capture complex, localized respiratory motion patterns.

Purpose of the Study:

• To develop and validate a novel statistical respiratory motion model using biplane X-ray images.
• To enhance the accuracy of respiratory motion field estimation by preserving local motion details for individual patients.

Main Methods:

• Constructed respiratory motion fields from CT datasets of 18 healthy subjects.
• Generated a lung contour motion repository based on boundary control point displacements.
• Utilized a statistical sparse motion field presentation (SMFP) method, applied twice, for motion field approximation.
• Fine-tuned non-zero coefficients to match reconstructed volumetric image projections with X-ray images.

Main Results:

• The SMFP-based method demonstrated a lower maximum average target registration error (2.9(1.6) mm) compared to the PCA method (3.1(2.0) mm).
• The SMFP method also achieved a smaller maximum average symmetric surface distance (2.4(1.3) mm) than the PCA method (2.5(1.6) mm).

Conclusions:

• The proposed statistical respiratory motion model effectively improves the accuracy of respiratory motion field estimation.
• This approach offers a promising solution for reducing tumor location uncertainty in lung biopsies.