Presenter(s)
V. Sharma1, K. Shrestha2, Z. Davidson3, L. Rabizadeh2, E. Lee2, E. Anderson4, R. Krochmal5, J. A. Haas6, T. J. Carpenter6, N. Aghdam7, D. Pang1, S. Rudra1, K. R. Unger1, N. Paudel8, and J. W. Lischalk9; 1Department of Radiation Medicine, MedStar Georgetown University Hospital, Washington, DC, 2Georgetown University School of Medicine, Washington, DC, 3George Washington School of Medicine and Health Sciences, Washington, DC, 4Department of Radiation Oncology, Cedars-Sinai Medical Center, Los Angeles, CA, 5Division of Pulmonary and Critical Care Medicine Georgetown University Hospital, Washington, DC, 6Department of Radiation Oncology, Perlmutter Cancer Center at New York University Langone Hospital - Long Island, Mineola, NY, 7Department of Radiation Oncology, Beth Israel Deaconess Medical Center, Boston, MA, 8Northwestern Feinberg School of Medicine, Chicago, IL, 9Department of Radiation Oncology, Perlmutter Cancer Center at New York University Langone Hospital - Long Island, New York, NY
Purpose/Objective(s): Four-dimensional CT (4DCT) is widely used to quantify respiratory motion for SBRT planning, yet it may underestimate intrafraction excursions during treatment delivery. We conducted a paired, target-level comparison of intrafraction motion measured from CyberKnife (CK) treatment log corrections versus 4DCT-derived centroid motion in the superior–inferior (SI), left–right (LR), and anterior–posterior (AP) axes.
Materials/Methods: Fifteen lung targets were included. Median age was 78 years (range 66–96); 80% were male and 87% had a smoking history. Most lesions were early-stage NSCLC (73%) treated with standard SBRT regimens (48–60 Gy in 3–10 fractions). Targets were located in the upper lobes (53%), lower lobes (27%), and right middle lobe (20%), and most patients were medically inoperable with substantial cardiopulmonary comorbidity. CK motion was derived from time-resolved translational correction logs sampled during beam delivery across all fractions for each target. 4DCT motion was derived from phase-resolved fiducial centroid positions across the respiratory cycle. For both modalities, motion was expressed relative to the target-specific mean position (mean-centered). Three-dimensional (3D) displacement was calculated as the Euclidean magnitude of the orthogonal components v[(?SI)2+(?LR)2 + (?AP)2]. Endpoints included axis-specific maxima (Max SI, Max LR, Max AP) and three-dimensional displacement summarized as the mean, 95th percentile, and maximum. Paired target-level comparisons were performed using two-sided Wilcoxon signed-rank tests. CK sampling was dense, with a median of 286 timepoints per target (IQR 206.5–466.5) across a median of 5 fractions.
Results: Across targets, CK log-derived intrafraction motion was greater than 4DCT centroid motion across all key endpoints, with the most pronounced separation in high-percentile and maximum excursions. The median per-target mean 3D displacement was 4.824 mm for CK versus 1.255 mm for 4DCT (p<0.05). Median 95th percentile 3D displacement was 10.209 mm versus 3.104 mm (p<0.05), and median maximum 3D displacement was 20.540 mm versus 3.243 mm (p<0.05). Axis-specific maxima were likewise higher with CK: Max SI 11.177 vs 2.324 mm, Max LR 11.302 vs 0.911 mm, and Max AP 11.469 vs 1.527 mm (all p<0.05).
Conclusion: In this paired, target-level analysis, CyberKnife treatment logs captured markedly greater intrafraction motion than 4DCT centroid estimates, with the widest gaps in the 95th percentile and maximum excursions. These results suggest that 4DCT centroid motion can underestimate clinically important motion “tails” during SBRT delivery and emphasize the need to account for intrafraction variability when judging the adequacy of motion management.