Extending CT Ventilation Imaging to 4D Using Time-Varying Velocity Fields Validated with Galligas PET
Abstract
Purpose
CT ventilation imaging (CTVI) is a promising functional imaging modality. However, conventional 3D approaches using Displacement Vector Fields (DVF) assume a linear trajectory between peak-exhale and peak-inhale phases, potentially underestimating actual respiratory motion. The purpose of this study is to evaluate the feasibility of 4D-CTVI using Time-Variant Fields (TVF), which model deformation as a temporal velocity field, and to compare it with conventional DVF-based 3D-CTVI.
Methods
4DCT and Galligas PET datasets of 5 lung cancer patients were obtained from The Cancer Imaging Archive (TCIA). Lung masks were generated to exclude the trachea and major vessels. To generate CT ventilation images, deformable image registration (DIR) was performed between peak-exhale and peak-inhale CT images using Advanced Normalization Tools (ANTs). Two approaches were compared: 3D-CTVI, calculated from Displacement Vector Fields (DVF) using the Symmetric Normalization (SyN) algorithm, and 4D-CTVI, derived from Time-Varying Velocity Fields (TVF) which model deformation as a continuous spatiotemporal velocity field. For both methods, ventilation images were calculated using the Jacobian determinant of the resulting deformation fields. Spatial accuracy was evaluated by calculating Spearman's rank correlation coefficients with Galligas PET images based on block-averaged analysis.
Results
The mean Spearman's rank correlation coefficients with Galligas PET were 0.38 ± 0.31 for 3D-CTVI and 0.51 ± 0.30 for 4D-CTVI. The 4D-CTVI demonstrated a significantly higher correlation compared to the 3D-CTVI (p < 0.05).
Conclusion
4D-CTVI using TVF is feasible and demonstrated improved spatial correlation with functional PET compared to the conventional DVF-based 3D approach. Further accuracy improvements and validation are expected through the utilization of all respiratory phases, optimization of DIR parameters, and an increase in the number of cases.