Dynamic lung modeling and tumor tracking using deformable image registration and geometric smoothing

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiotherapy Physics

Background:

• Accurate lung modeling and tumor tracking are crucial for effective image-guided radiotherapy.
• Dynamic changes in lung volume and tumor position during respiration pose challenges for treatment planning.

Purpose of the Study:

• To develop and evaluate a fully automatic deformable image registration algorithm for dynamic lung modeling and tumor tracking.
• To assess the algorithm's accuracy in delineating and tracking tumor motion using 4D CT images.

Main Methods:

• A greyscale-based optical flow method with geometric smoothing was employed for deformable image registration.
• The algorithm utilizes 4D CT images, extracting lung models from exhale phases and representing them using signed distance functions and tetrahedral meshes.
• Registration was performed by calculating displacement based on CT image intensity values.

Main Results:

• The developed algorithm successfully registered dynamic lung models across 10 phases of 4D CT scans.
• Evaluation using lung volume change and maximum tissue movement indicated the algorithm's accuracy.
• Testing demonstrated the effectiveness of deformable registration for tumor motion delineation and tracking.

Conclusions:

• The proposed automatic deformable image registration algorithm is effective for dynamic lung modeling and tumor tracking.
• This approach enhances precision in image-guided radiotherapy by accurately accounting for respiratory motion.
• The method provides a robust tool for delineating and monitoring tumor movement during radiation treatment.