3125 - 4D CT Underestimates Intrafraction Thoracic Tumor Motion During Robotic Radiosurgical Treatments
Presenter(s)
V. Sharma1, Z. Davidson2, K. Shrestha3, E. Lee3, L. Rabizadeh3, E. Anderson4, R. Krochmal4, T. J. Carpenter5, J. A. Haas5, N. Aghdam6, D. Pang1, S. Rudra1, K. R. Unger1, N. Paudel7, and J. W. Lischalk8; 1Department of Radiation Medicine, MedStar Georgetown University Hospital, Washington, DC, 2George Washington School of Medicine and Health Sciences, Washington, DC, 3Georgetown University School of Medicine, Washington, DC, 4Division of Pulmonary and Critical Care Medicine Georgetown University Hospital, Washington, DC, 5Department of Radiation Oncology, Perlmutter Cancer Center at New York University Langone Hospital - Long Island, Mineola, NY, 6Department of Radiation Oncology, Beth Israel Deaconess Medical Center, Boston, MA, 7Northwestern Feinberg School of Medicine, Chicago, IL, 8Department of Radiation Oncology, Perlmutter Cancer Center at New York University Langone Hospital - Long Island, New York, NY
Purpose/Objective(s): Respiratory motion estimates used for lung SBRT planning are typically derived from 4DCT, but phase-binned imaging provides only a discrete snapshot and may miss intrafraction extremes during treatment delivery. We therefore performed a paired, target-level temporal comparison of CyberKnife (CK) log–derived motion against the 4DCT centroid motion envelope, quantifying mean-centered deviation and time-weighted dwell beyond clinically relevant displacement thresholds and beyond the target-specific 4DCT maximum.
Materials/Methods: Fifteen lung targets were included (median age 78 years [66–96]; 80% male; 87% with a smoking history). CK log files were processed as timestamped time series of translational corrections annotated by fraction; timestamps as a surrogate for beam-on delivery time. For 4DCT, centroid coordinates across respiratory phases were converted to a centroid-motion envelope using 3D displacement from the phase-mean centroid (median, p95, max). CK intrafraction motion was quantified per fraction as mean-centered 3D deviation (median, p95, max). Time-weighted dwell beyond thresholds (>5 mm, >10 mm, and > target-specific 4DCT maximum) was computed from timestamp-derived intervals. Paired target-level CK versus 4DCT metrics were compared using Wilcoxon signed-rank testing.
Results: CK exceeded the 4DCT centroid envelope across central tendency and tail metrics. Across the averaged cohort, the median 3D deviation was 4.824 mm (IQR 2.44–5.46) for CK compared with 1.26 mm (0.88–2.31) for 4DCT. Median CK p95 was 10.209 mm (6.29–12.64) versus 4DCT p95 of 3.10 mm (1.69–4.30) and median CK maximum was 20.77 mm (14.49–36.95) versus 4DCT maximum 3.24 mm (1.76–5.19); all 15/15 targets had CK maximum exceeding the target-specific 4DCT maximum. Across the cohort, CK deviation above the 4DCT maximum accounted for 33.7% of total beam-on time, corresponding to approximately 92.4 minutes per patient based on average treatment time. CK deviation >5 mm above the 4DCT maximum accounted for 4.0% of beam-on time (18.1 minutes per patient), while deviations >10 mm accounted for 0.9% of beam-on time (4.0 minutes per patient).
Conclusion: In this paired temporal analysis, CyberKnife demonstrated sustained intrafraction motion beyond the 4DCT centroid-defined envelope. A substantial proportion of beam-on time occurred outside the target-specific 4DCT maximum excursion, while large-magnitude deviations were brief and uncommon. These findings suggest that phase-binned 4DCT centroid sampling may underestimate the duration of intrafraction motion during lung SBRT and highlight the value of time-resolved motion assessment during treatment delivery.