Room: AAPM ePoster Library
Deformable image registration (DIR) based dose accumulation between treatment modalities and fractions is utilized for treatment planning and outcome prediction. DIR geometric target registration errors (TRE) and dose accumulation error (DAE) caused by TRE both need to be evaluated. TRE can be evaluated using existing approaches, e.g., landmarks. In this study we propose a new approach to assess DAE based on a 3D dose gradient and TRE 3D vector field.
Dose accumulation is accomplished by summation of the dose volumes transformed by the deformation vector field (DVF) generated from the registration of images of two different fractions. If TRE is known, DAE is computed as TRE • Gdose where Gdose is the dose gradient, and • is the vector dot operator. Both TRE and Gdose are 3D vectors. Therefore, DAE is the magnitude of TRE projected on the Gdose direction (|TRE//Gdose|) then multiplied by the Gdose magnitude. Only TRE along Gdose direction contributes to dose accumulation errors but not TRE perpendicular to Gdose. Geometric TRE in dose accumulation only needs to be evaluated along the dose gradient direction. The clinical goal of DAE < 5% of Rx (prescribed dose) yields the dose gradient-dependent TRE threshold |TRE//Gdose|<0.05*Rx/|Gdose|.
A GYN HDR case was used for demonstration. MR scans from two fractions were registered on uterus contours using a Gaussian Mixture Model DIR method. Surface TREs were calculated and tested against TRE tolerances (0.05*Rx/|Gdose|). The maximum tolerance within 2mm of the evaluation point was used to account for HDR applicator localization uncertainty. Results showed that surface TRE = 0.2±0.6 mm and 91% of registered area was below the dose gradient dependent TRE tolerance.
A method was developed to access image registration errors and dose accumulation errors by considering directional TRE and dose gradient vector fields.