Premium
SU‐E‐J‐81: A Contour‐Guided Deformable Image Registration Scheme for Organ Surface Deformation
Author(s) -
Gu X,
Pompos A,
Zhong Z,
Wang J,
Guo X,
Jia X,
Dong B,
Jiang S,
Solberg T
Publication year - 2013
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 - 10.1118/1.4814293
Subject(s) - image registration , affine transformation , artificial intelligence , medical imaging , computer vision , surface (topology) , mathematics , algorithm , computer science , geometry , nuclear medicine , image (mathematics) , medicine
Purpose: Accurate deformation vector field (DVF) acting on organ surface is vital for surface dose summation of deformable cavitary organs in adaptive radiotherapy. The aim of this study is to develop a contour‐guided deformable image registration (DIR) scheme to establish accurate DVF on an organ surface. Methods: A critical cavitary organ, for example bladder, is delineated by a clinician on fraction 1 and fraction 2 CT images used for cervical high‐dose rate (HDR) brachytherapy planning. Both CT sets are preprocessed by creating binary images by setting one inside of the organ under evaluation and zero outside. The organ surface is subsequently extracted and discretized with a triangular mesh. The DVF on designed vertices is estimated through an inverse‐consistent demons‐DIR between the paired binary images. Due to shortcomings in DIR algorithms, the DVF derived through DIR always has limited accuracy. To increase the DVF accuracy, an iterative closest point (ICP) algorithm is adopted in this study to match corresponding vertices of the paired surface meshes through an affine registration. The accuracy of the resulting registration is then evaluated. Results: Using contours drawn by a clinician as ground truth we evaluated DVF accuracy through organ surface‐to‐surface distance (SSD), defined as the shortest Euclidean distance between the vertices on the target and the deformed surfaces. In our evaluation case, the maximum SSD is 27.8mm before registration, decreases to 15.0mm following demons registration, and 8.33mm after further affine alignment. The 95 percentile SSD decreases from 23.9mm to 10.8mm and to 3.8mm, respectively. Conclusion: The proposed contour‐guided DIR scheme substantially improves the accuracy of DVF on organ surfaces. Results from the evaluation case demonstrate that the proposed contour‐guided DIR scheme can provide accurate DVF making it a useful approach to be applied in adaptive radiotherapy.