266 - Time-Dependent Solution to the Van Herk PTV Margin Formalism for Brain, Head-and-Neck, Thorax, Breast, Abdomen and Pelvis Sites via 33,000 Image Pairs

Purpose/Objective(s):

Numerous studies have 1) demonstrated target baseline positional drift and 2) used the van Herk formalism for PTV assessment. However, few studies derive time-dependent PTV margins and are limited to single sites and smaller datasets. Under the hypothesis that random set-up uncertainty s is time-dependent, we analyzed, to our knowledge, the largest dataset of image–couch shift pairs to quantify time-dependent PTV margins for brain, head-and-neck (HN), thorax, breast, abdomen, and pelvis sites.

Materials/Methods:

Our entire institutional database was queried for couch positions, shifts, and timestamps from 5 years of conventional IGRT (100,019 sessions). Random set-up uncertainty s was calculated using the van Herk formalism as the root mean square (RMS) of the per-patient standard deviation (SD) of random error, defined as shifts after subsequent imaging of the same modality (kV/CBCT), i.e., repeat imaging after initial alignment. For time-dependent analysis, per-patient SD was evaluated versus mean time between images in the lat, lng, and vrt directions. RMS was calculated in sequential (increasing time) 40-patient groups to obtain s_Lat, s_Lng, and s_Vrt, then combined in quadrature to derive s_Total. Group-wise s_Total values were assigned to the mean inter-image time and fit linearly versus time to obtain s_Total(t). Time-dependent isotropic PTV margins were then calculated as 2.5·S + 0.7·s_Total(t), where S = 0.5 mm represents a systematic imaging–treatment isocenter offset.

Results:

33,792 shifts from 3,243 patients were analyzed (Table 1). s_Lat, s_Lng, s_Vrt, and s_Total were strongly correlated to time, with R² = 0.88 for brain and R² = 0.95 for other sites. Brain had the lowest per-patient dispersion of random errors, and thus the lowest slope of s_Total vs. time (0.7 mm/min); breast had the largest (1.5 mm/min). Instantaneous PTV margins reached 3 mm at 3.7 / 3.1 min for brain / HN, and 5 mm at 4.9 / 3.6 / 4.8 / 4.7 min for thorax / breast / abdomen / pelvis. The average margin required from time 0?T equaled 3 mm at 7.3 / 6.1 min for brain / HN, and 5 mm at 9.8 / 7.2 / 9.6 / 9.4 min for thorax / breast / abdomen / pelvis.

Conclusion:

Findings support our hypothesis — PTV margins are time-dependent across sites, with standard margins lasting ~6 min for HN, ~7 min for brain and breast, and ~10 min for thorax, abdomen, and pelvis, including time required for imaging and alignment. Clinics should consider: 1) active monitoring, 2) time-dependent re-imaging criteria, 3) increased margins, or 4) increased delivery speed when small margins are required.

Abstract 266 - Table 1.

	Brain	HN	Thorax	Breast	Abdomen	Pelvis
# Patients	569	585	663	493	364	569
# Shifts / Image-Pairs	7,219	7,313	7,276	3,903	3,492	4,589
Slope of sTotal vs. time (mm/min)	+0.7	+0.8	+1.1	+1.5	+1.1	+1.1
R2 of sTotal vs. time	0.90	0.99	0.99	0.98	0.98	0.98
Time, Instantaneous PTV margin = 3 mm (min)	3.7	3.1
Time, Instantaneous PTV margin = 5 mm (min)			4.9	3.6	4.8	4.7
Time, Average PTV margin 0 ? T = 3 mm (min)	7.3	6.1
Time, Average PTV margin 0 ? T = 5 mm (min)			9.8	7.2	9.6	9.4