Open AccessEvaluation and Comparison of Deformable Image Registration Algorithms for 4D CT Images
Author(s) -
Xiaokun Hu,
Guangpu Shao,
Jimin Yang,
Juan Yang
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1453/1/012038
Subject(s) - image registration , metric (unit) , algorithm , artificial intelligence , displacement (psychology) , computer science , computer vision , point (geometry) , image (mathematics) , motion (physics) , mathematics , geometry , engineering , psychology , operations management , psychotherapist
Deformable image registration (DIR) is crucial in adaptive radiation therapy. However, the validation is a challenging work due to the lack of gold-standard. This study proposed an evaluation framework by using point-to-point displacement vector field (DVF). Three DIR algorithms including original flow, active Demons and symmetric force Demons were validated for ten lung 4D CT images with landmarks. DVFs derived from DIR algorithms (dDVF) and manually measured according to landmarks (mDVF) were analyzed and compared. Their target registration errors (TRE) and the relationship of lung motion and three-dimensional TRE were explored. The distance discordance metric (DDM) values were calculated. For all cases, the active Demons algorithm had the smallest TRE value and DDM values and it showed the poor correlation between 3D TRE and lung motion, which indicated that this algorithm outperformed other DIR algorithms enrolled in this study. Preliminary results demonstrated that the proposed evaluation framework had a potential ability of providing clinical guidance for the selection of appropriate algorithm in radiation therapy.