Premium
WE‐E‐BRC‐10: Quality Assurance of Deformation Algorithms Using a Two‐Dimensional Deformable Phantom
Author(s) -
Kirby N,
Chuang C,
Pouliot J
Publication year - 2011
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3613388
Subject(s) - imaging phantom , quality assurance , deformation (meteorology) , ground truth , computer science , similarity (geometry) , computer vision , artificial intelligence , algorithm , orientation (vector space) , metric (unit) , biomedical engineering , physics , optics , mathematics , geometry , medicine , image (mathematics) , operations management , external quality assessment , pathology , meteorology , economics
Purpose: Clinical implementation of deformation algorithms requires dependable quality assurance techniques. A two‐dimensional deformable phantom that can objectively verify the accuracy of the algorithms throughout an entire slice of the anatomy is proposed. Methods: The phantom represents a single plane of the anatomy for a head and neck patient. Inflation of a balloon catheter inside the phantom simulates tumor growth. CT and camera images of the phantom are acquired before and after its deformation. Non‐radiopaque markers reside on the surface of the deformable anatomy and are visible through an acrylic plate, which enables an optical camera to measure their positions; thus, establishing the ground truth deformation. This measured deformation can be directly compared to the predictions of deformation algorithms, using several similarity metrics, and it can be applied to create a simulated deformation for a patient CT, which can also be used to test algorithm accuracy. The ratio of the number of points with more than a 3 mm deformation error over those that are deformed by more than 3 mm was used for an error metric. A comparison of the deformation algorithm accuracy for the phantom CTs and the simulated CTs evaluates the adequacy of the phantom electron density heterogeneity. Results: The balloon catheter deforms 32 out of the 54 surface markers by more than 3 mm. Different deformation errors result from the different similarity metrics. The most accurate deformation results for the phantom CTs had an error of 75%, compared to 25% for the simulated CTs. Conclusions: The developed phantom demonstrates its utility for verifying deformation algorithms and determining which is the most accurate. The reduction of the deformable anatomy to a two‐dimensional system allows for the use of non‐radiopaque markers, which do not influence deformation algorithms. This is the fundamental advantage of this verification technique.