Segmentation of biological target volumes on multi-tracer PET images based on information fusion for achieving dose painting in radiotherapy

  • Authors:

  • Benoît Lelandais;Isabelle Gardin;Laurent Mouchard;Pierre Vera;Su Ruan

  • Affiliations:

  • LITIS EA4108, QuantIF, University of Rouen, Rouen Cedex, France;LITIS EA4108, QuantIF, University of Rouen, Rouen Cedex, France, Dept. of Nuclear Medicine, Centre Henri-Becquerel, Rouen Cedex 1, France;LITIS EA4108, QuantIF, University of Rouen, Rouen Cedex, France;LITIS EA4108, QuantIF, University of Rouen, Rouen Cedex, France, Dept. of Nuclear Medicine, Centre Henri-Becquerel, Rouen Cedex 1, France;LITIS EA4108, QuantIF, University of Rouen, Rouen Cedex, France

  • Venue:
  • MICCAI'12 Proceedings of the 15th international conference on Medical Image Computing and Computer-Assisted Intervention - Volume Part I
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Medical imaging plays an important role in radiotherapy. Dose painting consists in the application of a nonuniform dose prescription on a tumoral region, and is based on an efficient segmentation of Biological Target Volumes (BTV). It is derived from PET images, that highlight tumoral regions of enhanced glucose metabolism (FDG), cell proliferation (FLT) and hypoxia (FMiso). In this paper, a framework based on Belief Function Theory is proposed for BTV segmentation and for creating 3D parametric images for dose painting. We propose to take advantage of neighboring voxels for BTV segmentation, and also multi-tracer PET images using information fusion to create parametric images. The performances of BTV segmentation was evaluated on an anthropomorphic phantom and compared with two other methods. Quantitative results show the good performances of our method. It has been applied to data of five patients suffering from lung cancer. Parametric images show promising results by highlighting areas where a high frequency or dose escalation could be planned.