Investigation of tilted dose kernels for portal dose prediction in a-si electronic portal imagers Krista Chytyk MSc student Supervisor: Dr. Boyd McCurdy
Introduction The objective of cancer radiotherapy is to destroy the malignancy by maximizing the dose to the tumour while minimizing the dose to the normal tissues The accuracy of this delivery directly affects patient outcome because small variations in dose may lead to large discrepancies in tissue response dose errors of greater than 3% can have a considerable effect on patient outcome * * A. Brahme, Dosimetric precision requirements in radiation therapy, Acta. Radiologica. Oncology 23 Fasc. 5, 379-391 (1984).
Geometric verification of treatments reduces the uncertainty in the delivery location carried out with radiographic film, but has been largely replaced by electronic portal imaging devices (EPIDs) a-si EPIDs have shown potential as a radiation dosimeter * * http://www.varian.com
Performing dosimetric verification while the treatment is administered allows for appropriate adjustments to be made for the remainder of the treatment or treatments Measured and predicted portal images can be compared to determine whether the plan is delivering correct dose distributions
There had been no investigation into the accuracy of a-si portal dose image prediction using the parallel kernel assumption applying parallel dose kernels in a divergent clinical geometry where radiation fluence is not truly parallel Parallel (convolution) Tilted (superposition) Incident diverging fluence Parallel kernels Tilted kernels
Methods and Materials Monte Carlo Simulations Dose kernels for the EPID were obtained at a number of incident angles for a range of monoenergetic photon energies using EGSnrc 0 to 14 every 2 0.1, 2, 6 and 18 MeV user code DOSXYZnrc Incident tilted tilted photon beams Incident perpendicular photon beams back -x-axis EPID case +x-axis Phosphor Proximal half Distal half
Simulated detector design specifications provided by Varian Medical Systems (Palo Alto, CA) uniform epoxy front cover followed by a 22 mm air gap and the imaging cassette 1 mm thick copper buildup plate, a Lanex Fast-B phosphor (gadolinium oxysulphide) screen and an a-si photodiode array 3 cm slab of uniform water buildup was added directly upstream of the detector front cover simulated area of the detector for dose kernel generation was approximately 40 x 40 cm 2
Dose Kernel Analysis The effect of angle on the symmetry of the dose kernel about the incident pencil beam was investigated kernels were divided into the negative-x half (proximal half) and the positive-x half (distal half), as in figure
Tilted vs. Parallel Kernels Clinical impact was investigated by applying the parallel and tilted kernels to a step function representing incident fluence The convolution (parallel kernels) and superposition (tilted kernels) methods of dose calculation were employed for two field sizes and two SDDs 27.2 x 20.4 and 30 x 30 cm 2 at isocenter 105 and 140 cm D = ## $ detector ( x', y') " k( x! x', y! y') dx' dy'
For the superposition method, the appropriate tilted kernel for an angle of incidence for a voxel was interpolated from the available kernels This kernel was then rotated azimuthally and translated to the position of the voxel of interest, then regridded to the EPID image pixel coordinates The results of the convolution versus full superposition approach were evaluated by the χ- comparison test * with Δd = 0.784 mm (the detector resolution) and ΔD = 1.0% % = D # D c 2 ( r) $ D + # d 2 " r ( r)! D r 2 * A Bakai, M Alber and F Nusslin, A revision of the gamma-evaluation concept for the comparison of dose distributions, Phys. Med. Biol. 48, 3543-53 (2003).
Results Dose Kernel Comparison 1E-15 Dose/Fluence (Gy*cm 2 /N) 1E-16 1E-17 1E-18 1E-19 1E-20 1E-21 0 kernel 14 kernel (distal) 14 kernel (proximal) 1E-22 1E-23 0 5 10 15 20 Position (cm)
Superposition vs. Convolution Fraction of Voxels Fraction of Voxels 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0.30 0.25 0.20 0.15 0.10 0.05 0.1 MeV 2 MeV 0.00-2 -1 0 1 2!-Comparison Value 6 MeV 18 MeV (a) (b)
Discussion and Conclusions Increasing asymmetry in the χ-comparison distributions with increasing incident photon beam energy Positive values of the χ-comparison indicate the parallel kernel assumption underestimates the tilted kernel superposition results slightly greater tendency for the parallel kernel assumption to underestimate the superposition results with increasing incident photon energy This study validates the practice of applying parallel dose kernels in a divergent clinical geometry to determine the dose deposited in an a-si EPID a full superposition is not required for dosimetric accuracy of better than 1.5% for extreme cases and 1.0% for clinically relevant scenarios
Questions?