Premium
TH‐A‐BRA‐05: Linear Regression Analysis of 2D Projection Image Data of 6 Degrees‐ Of‐Freedom Transformed 3D Image Sets for Stereotactic Radiotherapy
Author(s) -
Lin C,
Winey B
Publication year - 2012
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 - 10.1118/1.4736254
Subject(s) - transformation (genetics) , projection (relational algebra) , mathematics , artificial intelligence , principal component analysis , computer vision , pixel , image registration , algorithm , computer science , image (mathematics) , biochemistry , chemistry , gene
Purpose: To improve accuracy and speed of identifying the 6 degrees‐of‐ freedom (DoF) transformation during patient positioning alignment for stereotactic radiotherapy by determining the linear relationship between two real‐time orthogonal 2D projection images and two digitally reconstructed radiographs (DRRs) generated from a 3D volume image. Methods: Each 2D image representing an independent transformation was characterized and defined by a shape function, a function parameterized by distance and pixel value. A principal component analysis was developed to find the linear relationship between the 2D projection images and the two DRRs generated from the 3D volume image for small‐scale transformations by matching the 2D images to a set of base functions established from singular transformations. The patient‐positioning algorithm calculates the 6 DoF transformation of the patient based upon two orthogonal real‐time 2D images and from the linear relationship established. The algorithm has been validated against classical image registration methods on patient data for runtime and positioning accuracy. Results: The algorithm has accuracy to at least 1 pixel (equivalent to 0.5 mm accuracy given this image resolution). The shape function characterizes each 2D image on an average of 1.78 seconds. The algorithm interprets the evaluated transformation as a linear combination of separate 6 DoF transformations. The principal component analysis allows for rapid and accurate position matching of the images, and provides the transformation needed to align the orthogonal images to the volume on an average of 3.32 seconds. Conclusions: The linear relationship developed from the shape function algorithm affords speed and accuracy of patient positioning during stereotactic radiotherapy treatment. The algorithm is fast and accurate for small‐order 6 DoF transformations. This algorithm indicates the potential for high precision of patient positioning from the interpolation and extrapolation of the linear relationships. This project is supported by the NCI Federal Share of program income earned by Massachusetts General Hospital on C06 CA059267, Proton Therapy Research and Treatment Center.