A segmentation problem in quantitative assessment of organ disposition in radiotherapy

Giovanni Naldi, Barbara Avuzzi, Simona Fantini, Mauro Carrara, Ester Orlandi, Elisa Massafra, Stefano Tomatis

Research output: Contribution to journalArticlepeer-review

Abstract

Radiotherapeutic treatment of cancer is best conducted if the prescription dose is given to the tumor while surrounding normal tissues are maximally spared. With the aim to meet these requirements the complexity of radiotherapy techniques have steadily increased under a strong technological impulse, especially in the last decades. One problem involves the rate of the particular disposition of the structures of interest in a patient. Recently the authors (Tomatis et al., 2010; 2011) have proposed a computational approach in order to represent quantitatively the geometrical features of organs at risk, summarized in characteristics of distance, shape and orientation of such organs in respect to the target. A basic problem to solve before to compute the risk index, is the segmentation of the organs involved in the radiotherapy planning. Here we described a 3D segmentation method by using the clinical computed tomography (CT) data of the patients. Our algorithm is based on different steps, a preprocessing phase where a nonlinear diffusion filter is applied; a level set based method for extract 2D countours; a postprocessing reconstruction of 3D volume from 2D segmented slices. Some comparisons with manually traced segmentation by clinical experts are provided.

Original languageEnglish
Pages (from-to)179-186
Number of pages8
JournalImage Analysis and Stereology
Volume30
Issue number3
DOIs
Publication statusPublished - 2011

Keywords

  • 3D reconstruction
  • Geometrical features
  • Level set
  • Radiotherapy planning
  • Segmentation

ASJC Scopus subject areas

  • Physics and Astronomy (miscellaneous)

Fingerprint Dive into the research topics of 'A segmentation problem in quantitative assessment of organ disposition in radiotherapy'. Together they form a unique fingerprint.

Cite this