On a PCA-based lung motion model.

Area of Science:

• Radiotherapy Physics
• Medical Imaging
• Computational Anatomy

Background:

• Respiration-induced lung motion is a major uncertainty in lung cancer radiotherapy.
• Previous models focused on single-point motion, neglecting the entire lung's dynamics.
• Accurate modeling of lung motion is crucial for effective radiotherapy planning.

Purpose of the Study:

• To understand why principle component analysis (PCA) effectively models lung motion.
• To determine the optimal number of PCA coefficients for accurate lung motion modeling.
• To develop a method for deriving entire lung motion from a single internal marker using PCA.

Main Methods:

• Theoretical analysis of PCA for lung motion modeling.
• Validation using mathematical respiratory phantoms with known motion patterns.
• Application to 11 clinical 4DCT scans from eight lung cancer patients.

Main Results:

• PCA model demonstrated significant modeling power and generalization performance on clinical data.
• Average 3D modeling error using PCA was within 1 mm (0.7 ± 0.1 mm).
• Deriving lung motion from a single internal marker yielded an average 3D error within 2 mm (1.8 ± 0.3 mm).

Conclusions:

• PCA provides a robust framework for modeling entire lung motion.
• A single internal marker can effectively drive the PCA-based lung motion model.
• Two PCA coefficients are generally sufficient for accurate lung motion modeling in most patients.