Sensitivity of Cumulative Dose Distributions to Deformable Image Registration Uncertainties Associated with Tumor Regression Dynamics
C Dial*, G Hugo, J Siebers, Virginia Commonwealth University, Richmond, VATH-E-218-7 Thursday 1:00:00 PM - 2:50:00 PM Room: 218
Purpose: To demonstrate the sensitivity of cumulative dose distributions to deformable image registration (DIR) uncertainties associated with tumor regression dynamics.
Methods: For 9 patients receiving definitive radiation therapy, structures are delineated by a qualified physician on 4-7 weekly helical CTs acquired during the course of radiotherapy. Plans based on 3 different margins (7mm sup/inf, 5mm axial; 3mm sup/inf, 2mm axial; 0mm expansions of clinical target volume) are generated using planning criteria specified in RTOG protocol 0839. Dose calculated on weekly images is accumulated to the planning data-set using a surface-based deformable image registration algorithm driven by two distinct contour sets; each simulating a presumed extreme of regression dynamics: (1) simulates a scenario where displaced tissue tracks a regressing gross tumor volume (GTV) border; (2) assumes infiltrative disease where surrounding tissue remains fixed as visible tumor regresses. Volume of ipsilateral lung receiving at least 20 Gy (V20) and dose to 95% of the planning target volume (D95 PTV) for each regression-scenario and margin are compared. GTV regression is also reported.
Results: Mean percent decrease in GTV for the population is 40.7% (5.2-80.5%). Changes in lung V20 resulting from assumed regression dynamics are < 3% (0.3-2.9%) for all patients and margins. 7 of 9 patients exhibit < 3% differences (0.4-2.6%) in D95 PTV for all margins. An average change of 52.2% (50.7-54.2%) and 5.4% (4.6-6.3%) in D95 PTV for all margins is observed for the remaining 2 patients respectively. The patient with the greatest change in D95 PTV also exhibited a large amount of tumor-volume regression (80.5%).
Conclusions: For 3D conformal plans, V20 of ipsilateral lung and D95 PTV are relatively insensitive to DIR deviations associated with regression dynamics. However, deviations in D95 PTV can be large in the presence of large regression.