MAP-MRF segmentation of lung tumours in PET/CT images

  • Authors:

  • Hugh Gribben;Paul Miller;Gerard G. Hanna;Kathryn J. Carson;Alan R. Hounsell

  • Affiliations:

  • The Institute of Electronics, Communications and Information Technology, Queen's University Belfast, UK;The Institute of Electronics, Communications and Information Technology, Queen's University Belfast, UK;Centre for Cancer Research and Cell Biology, Queen's University Belfast, UK;Northern Ireland Regional Medical Physics Agency, Royal Victoria Hospital, Belfast, UK;Centre for Cancer Research and Cell Biology, Queen's University Belfast and Northern Ireland Regional Medical Physics Agency, Cancer Centre, Belfast City Hospital, UK

  • Venue:
  • ISBI'09 Proceedings of the Sixth IEEE international conference on Symposium on Biomedical Imaging: From Nano to Macro
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

The unsupervised maximum a posterior - Markov random field labelling technique for lung tumour segmentation in registered PET/CT imagery is proposed. The technique was applied to a range of PET/CT scan clinical datasets obtained from patients with non-small cell lung cancer. The technique was then extended to use a vector approach to take into account the CT datasets along with the corresponding PET. The performances of both the scalar and vector algorithms were in this case then compared to manual outlines obtained from the four clinicians' gross tumour volume outlines. Results showed comparable variability with that of the clinicians, with slightly better results returned for the vector technique.