Time Is of the Essence: Precise Quantification of Intrafraction Baseline Positional Drift across Brain, Head-and-Neck, Thorax, Breast, and Pelvis Sites By Leveraging 100,000 IGRT Sessions
Abstract
Purpose
Numerous studies have quantified target baseline positional drift rates across anatomic sites; however, these investigations are often limited by surrogate localization methods (e.g., surface-guidance), single-site analyses, or limited datasets. Here, we leverage the largest dataset of image–couch shift pairs to precisely quantify time-dependent baseline positional drift rates across five sites.
Methods
The radiation oncology information system database was queried to obtain couch positions, shifts, and timestamps from five years of conventionally-fractionated IGRT treatments delivered on four linear accelerators, totalling 100,019 treatment sessions. For sessions that contained consecutive image pairs of the same modality (CBCT or orthogonal kV pair), the applied couch shift after the second image was binned by elapsed time between images. Linear baseline positional drift rates were quantified as a function of motion percentile for brain, head-and-neck, thorax, breast, and pelvis, defined as the fraction of sessions with equal or lower motion.
Results
A total of 26,399 sessions (26%) contained at least one subsequent image pair of the same imaging modality within the sites of interest, yielding 27,602 image pairs from 3,894 treatment plans and 3,256 patients (brain:6,873; head and neck:6,221; thorax:6,446; breast:3,768; pelvis:4,294). At the 80th motion percentile, baseline positional drift rates were 0.6, 0.8, 1.2, 1.6, and 1.5 mm/min for brain, head and neck, thorax, breast, and pelvis, respectively. These rates correspond to 3-mm PTV margins being reached within 5.0 and 3.8 minutes for brain and head and neck, and 5-mm margins being reached within 3.8, 3.1, and 3.3 minutes for thorax, breast, and pelvis, respectively. The median coefficient-of-determination (R²) for linear drift-rate fits across all sites and motion percentiles (5th–95th) was 0.96 (interquartile range: 0.92–0.97).
Conclusion
These results suggest that random positioning uncertainty is dominated by time-dependent baseline positional drift, emphasizing the need for appropriate margin selection, time-based re-imaging, or active intrafraction motion management.