Main Session
Sep 29
PQA 05 - Physics

3214 - Respiratory Cycle-Resolved Ventilation Imaging Based on Dynamic CBCT for Quantitative Analysis of Lung Function

12:30pm - 01:45pm ET
Poster Hall - Exhibit Hall A
Screen: 19
POSTER

Presenter(s)

Tingliang Zhuang, PhD - University of Texas Southwestern Medical Center, Dallas, Texas

T. Zhuang1,2, R. Zuo1, H. C. Shao1, S. J. Domal1, S. Stojadinovic1, K. D. Westover1, S. B. Jiang1,2, and Y. Zhang1,2; 1Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX, 2MAIA Laboratory, Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX

Purpose/Objective(s): Ventilation imaging (VI) from motion-resolved tomography is increasingly used for noninvasive lung function assessment, but current approaches mainly rely on 4D-CT/CBCT, which suffer from phase-sorting artifacts, intra-phase residual motion, and only provide a nominal, motion-averaged breathing cycle. As a result, temporal motion variability and true lung function dynamics are not captured. In this study, we propose dynamic CBCT for VI. By eliminating phase sorting and reconstructing a CBCT volume from each X-ray projection, dynamic CBCT enables time-resolved volumetric imaging to capture both regular and irregular motion cycle-to-cycle. We hypothesize that dynamic CBCT-derived VI can quantify cycle-resolved lung function and improve correlation between functional changes and radiation dose.

Materials/Methods:

Four patients with locally advanced lung cancer treated to 60 Gy in 30 fractions were retrospectively analyzed. Dynamic CBCT was reconstructed using a recently PMF-STINR technique to derive cycle-resolved VI (CR-VI). PMF-STINR reconstructs a reference CBCT and an eigenvector-based motion model from a standard CBCT scan. Concurrently-solved projection-specific weighting factors scale the motion eigenvectors to generate time-resolved deformation vector fields (DVFs), which deform the reference CBCT into a dynamic CBCT series.

From the solved dynamic DVFs, voxel-wise Jacobians were computed for each breathing cycle, and their mean (J-mean), standard deviation (J-std), and probability distributions were obtained across cycles. Lung function changes were quantified by differences in J-mean and J-std, and by the Wasserstein distance (WD) between voxel-wise Jacobian distributions at the beginning and end of treatment course. Conventional 4D-CBCT-derived Jacobian metrics were calculated using the same scan data.

Correlations between Jacobian metrics (J-mean, J-std, and WD) and planned dose were evaluated for voxels within high functional lung receiving 15-40 Gy. Dose-function relationships across patients were analyzed using a linear mixed-effects model.

Results:

Among J-mean, J-std and WD in dynamic CBCT-derived CR-VI, J-mean changes showed statistically significant (p<0.05) correlations with dose in both individual patients and in the linear mixed-effects model, that indicating greater functional loss at higher radiation doses. Overall, no significant correlations were observed using 4D-CBCT-derived metrics across the whole dataset.

Conclusion:

We demonstrated the feasibility of quantifying lung function changes using dynamic CBCT-derived, respiratory cycle-resolved ventilation imaging, achieving improved correlation with planned dose compared with 4D-CBCT-based ventilation imaging.

Dynamic CBCT (J-mean ratio between fx30 and fx1) 4D CBCT (J ratio between fx30 and fx1)
P1 -0.97 -0.54(p>0.1)
P2 -0.86 -0.81
P3 -0.99 +0.94
P4 -0.78 -0.84
Linear Mixed-effect Model (slope/p-value) -0.03/0.02 -0.0007/0.97