Experimental Validation of Dose Warping Accuracy From Deformable Image Registration
Varadhan1*, S Hui2, (1) Minneapolis Radiation Oncology, Fridley, MN, (2) Univ Minnesota, Minneapolis, MNSU-E-J-204 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
The purpose of this work is to verify accuracy of deformable dose arising from deformable image registration (DIR) algorithms where the mass and density of tissue is conserved. This was done by using custom built deformable phantom mimicking bladder geometry with implanted MOSFETs that can be positioned in multiple locations thereby enabling direct measurement of dose delivered in different deformation states.
The phantom was made using viscoelastic polymer which is nearly tissue equivalent with high tensile strength and elasticity. The phantom was deformed by positioning it with in a compression plate. 5 parallel air canals that run along the organ were used for positioning MOSFET detectors at multiple locations. Dose calculation was performed in undeformed geometry with varying degree of dose gradients. DIR was performed using both B-Splines algorithm using 3D Slicer and commercial MimVista platform which uses intensity based free form algorithm. The resultant DVF was applied to the original dose distribution and compared with directly measured dose in deformed geometry at multiple locations. For IMRT fields the accuracy of dose warp was also validated by recalculating the original fluence map on the deformed data set and comparing the resultant dose distribution with the deformed dose from DIR.
For non IMRT type fields, both B-Splines and MimVista yielded excellent point dose agreement to directly measured dose (within 2 cGy or 2%). For IMRT fields with large modulation since the inherent uncertainty in MOSFET measurement is 4.6%, agreement was less satisfactory. However comparison between IMRT doses recalculated in the deformed geometry with deformed dose from DIR yielded γ3%/3mm greater than 90%.
We have verified that DIR based dose warping can yield accurate results for the algorithms studied. The magnitude of deformation and degree of dose modulation has the greatest impact on accuracy of dose warp.