A Method for Assessing Ground-Truth Accuracy of the 5DCT Technique

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiology

Background:

• Accurate modeling of respiratory motion is crucial for effective lung cancer treatment planning.
• Current techniques for generating breathing phase-specific CT images require validation against true patient motion.

Purpose of the Study:

• To develop and validate a technique for assessing the accuracy of patient-specific breathing motion models.
• To use original free-breathing CT scans as ground truth for evaluating image generation accuracy.

Main Methods:

• Constructed patient-specific motion models from 25 free-breathing helical CT scans using deformable image registration.
• Generated deformation vector fields to simulate breathing motion from a reference image.
• Assessed image similarity between simulated and original scans using deformable registration.

Main Results:

• Qualitative image overlays showed excellent agreement between simulated and original CT scans.
• The technique accurately modeled significant diaphragm displacements (up to 2 cm) and lung-chest wall sliding motion.
• Mean and 95th percentile errors were 1.15 ± 0.37 mm and 2.47 ± 0.78 mm, respectively.

Conclusions:

• The proposed ground truth-based technique enables voxel-by-voxel accuracy analysis for motion modeling errors.
• The 5-dimensional CT technique accurately reproduced original helical CT scans across diverse breathing patterns.
• This method is applicable for identifying organ- or tumor-specific motion errors in treatment planning.