Patient specific dosimetry phantoms using multichannel LDDMM of the whole body

  • Authors:

  • Daniel J. Tward;Can Ceritoglu;Anthony Kolasny;Gregory M. Sturgeon;W. Paul Segars;Michael I. Miller;J. Tilak Ratnanather

  • Affiliations:

  • The Center for Imaging Science, The Johns Hopkins University, Baltimore, MD;The Center for Imaging Science, The Johns Hopkins University, Baltimore, MD;The Center for Imaging Science, The Johns Hopkins University, Baltimore, MD;Carl E. Ravin Advanced Imaging Laboratories, Duke University, Durham, NC and Department of Biomedical Engineering, University of North Carolina, Chapel Hill, NC;Carl E. Ravin Advanced Imaging Laboratories, Duke University, Durham, NC and Department of Radiology, Duke University Medical Center, Durham, NC;The Center for Imaging Science, The Johns Hopkins University, Baltimore, MD and Institute for Computational Medicine, The Johns Hopkins University, Baltimore, MD;The Center for Imaging Science, The Johns Hopkins University, Baltimore, MD and Institute for Computational Medicine, The Johns Hopkins University, Baltimore, MD

  • Venue:
  • Journal of Biomedical Imaging - Special issue on Parallel Computation in Medical Imaging Applications
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper describes an automated procedure for creating detailed patient-specific pediatric dosimetry phantoms from a small set of segmented organs in a child's CT scan. The algorithm involves full bodymappings fromadult template to pediatric images using multichannel large deformation diffeomorphic metric mapping (MC-LDDMM). The parallel implementation and performance of MC-LDDMM for this application is studied here for a sample of 4 pediatric patients, and from 1 to 24 processors. 93.84% of computation time is parallelized, and the efficiency of parallelization remains high until more than 8 processors are used. The performance of the algorithm was validated on a set of 24 male and 18 female pediatric patients. It was found to be accurate typically to within 1-2 voxels (2-4mm) and robust across this large and variable data set.