Abstract #4152
Automatic Fiducial Detection in T2 Weighted MRI in a Manifold Learning and Gaussian Mixture Modeling Framework
S. Ghose 1 , J. Mitra 1 , D. Rivest Henault 1 , A. Fazlollahi 1 , P. Stanwell 2 , P. Greer 3 , P. Pichler 3 , J. Fripp 1 , and J. Dowling 1
1
Australian e-Health Research Centre, CSIRO
Digital Productivity Flagship, Herston, QLD, Australia,
2
University
of Newcastle, NSW, Australia,
3
Department
of Radiation Oncology, Calvary Mater Newcastle Hospital,
NSW, Australia
Gold seeds or fiducials implanted in the prostate prior
to radiation treatment are frequently used to enable the
rigid registration of the two modalities required for
the transfer of the prostate contours from MRI to CT. An
automatic efficient detection method for the fiducials
from MRI is necessary to automate the procedure. This
work proposes Gaussian mixture modeling (GMM) and
spectral clustering based methods for fiducial candidate
selection and a similarity score based fiducial
detection. The proposed approach detects fiducials with
an accuracy of 95% when compared to the manual
detection.
This abstract and the presentation materials are available to members only;
a login is required.
Join Here
