Room: Exhibit Hall
Purpose: Recently, deformable image registration (DIR) technology is widely used to evaluate inter/intra organ deformation during pre- and post-irradiation in adaptive radiotherapy. However, the deformation accuracy is unclear in several situations and its suitable evaluation method has not been established yet. The purpose of this study is to propose validation method of deformation accuracy with three-dimensional (3D) phantom.
Methods: We created several homogenous 3D phantoms by 3D printer. Two different size and same shape 3D phantoms with hole were designed. With these 3D phantoms, stationary images were obtained by Aquilion LB (Toshiba Medical System) in several situations with intentional settings with considering three-dimensional Affine transformation parameters: Translation, Scaling, Rotation, and Sheer. All CT images were obtained for the phantoms with/without water inside hole. Then DIRs among different stationary images were performed by Raystation Ver.4.74 (Raysearch). Quantitative geometric assessments including the image deformation amount error (TRE), dice similarity coefficient (DSC) and consistency were evaluated in two algorithms (Intensity-based and Hybrid-based DIR). For evaluation in Hybrid-based algorithm, shape of phantom was added as the structure by dosimetrist before DIR. All contents were evaluated following AAPM TG-132.
Results: A total of 16 DIRs between reference image and target image were evaluated in this study. The larger TREs were observed in Intensity-based DIR compared to Hybrid-based DIR. Especially over 0.2 cm TREs were observed except for Translation condition in Intensity-based DIR. On the other hand, all DSC scores were over 0.9 (ranging from 0.903 to 0.969) with regardless of DIR algorithm in each condition.
Conclusion: We proposed simple validation method for deformation accuracy with 3D phantom. This method is able to evaluate quantitative geometrical accuracy in DIR. With this method, deformation inaccuracy in Hybrid-based DIR was smaller than that in Intensity-based DIR.
Deformation, Radiography, Validation