Skip to Main content Skip to Navigation
Theses

Vérification automatique de la position d'un patient pendant les traitements en radiothérapie externe par recalages d'images

Abstract : This research deals with treatment quality control in external radiotherapy, and concern more specifically automatic verification of patient set-up at the beginning of the fractions. Treatment images, or portal images, obtained during treatment are compared to a reference image that can be a radiograph or a DRR (digitally reconstructed radiograph). Patient positioning is checked in three dimensions by the measurement of anatomical structures displacement between the treatment images and the reference image. The main problem met by automation of this task is the very poor quality of portal images. Their contrast is very low, images are often oisy and edges are blurred. Our first aim was than the improvement of portal images quality, using filtering and restoration, in order to further the segmentation. A generic modelling of shapes likely to be found in both treatment and reference images was introduced as a priori knowledge in the segmentation of the simulation images. Shape recognition was realised by deformable matching of the generic modelling. Shape of patient specific structures result from this procedure and is the a priori knowledge introduced in the segmentation of portal images. Knowing that patient position is near the reference, reference structure and portal image superposition allows elimination, from the segmented portal image, of the segments that does not heng to the shapes of interest, by means of an interval simulating possible discrepancy. Moreover, information about edge's local orientation and the gray level local transition indicating the passage from bone to soft tissues, improve the results. Patient position is determined by a rigid matching of the segmented treatment and reference images. The results points out that the overall accuracy of patient position measurement is less than 1 pixel for translations and less than 1° for rotation.
Document type :
Theses
Complete list of metadata

https://hal.univ-lorraine.fr/tel-01750486
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 12:47:28 PM
Last modification on : Friday, October 23, 2020 - 5:00:28 PM

File

INPL_T_2000_HILT_G.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01750486, version 1

Collections

Citation

Gisèle Hilt. Vérification automatique de la position d'un patient pendant les traitements en radiothérapie externe par recalages d'images. Autre. Institut National Polytechnique de Lorraine, 2000. Français. ⟨NNT : 2000INPL085N⟩. ⟨tel-01750486⟩

Share

Metrics

Record views

139

Files downloads

61