Intra-patient semi-automated segmentation of the cervix-uterus in CT-images for adaptive radiotherapy of cervical cancer.
Phys Med Biol
; 58(15): 5317-32, 2013 Aug 07.
Article
em En
| MEDLINE
| ID: mdl-23863718
ABSTRACT
For online adaptive radiotherapy of cervical cancer, fast and accurate image segmentation is required to facilitate daily treatment adaptation. Our aim was twofold (1) to test and compare three intra-patient automated segmentation methods for the cervix-uterus structure in CT-images and (2) to improve the segmentation accuracy by including prior knowledge on the daily bladder volume or on the daily coordinates of implanted fiducial markers. The tested methods were shape deformation (SD) and atlas-based segmentation (ABAS) using two non-rigid registration methods:
demons and a hierarchical algorithm. Tests on 102 CT-scans of 13 patients demonstrated that the segmentation accuracy significantly increased by including the bladder volume predicted with a simple 1D model based on a manually defined bladder top. Moreover, manually identified implanted fiducial markers significantly improved the accuracy of the SD method. For patients with large cervix-uterus volume regression, the use of CT-data acquired toward the end of the treatment was required to improve segmentation accuracy. Including prior knowledge, the segmentation results of SD (Dice similarity coefficient 85 ± 6%, error margin 2.2 ± 2.3 mm, average time around 1 min) and of ABAS using hierarchical non-rigid registration (Dice 82 ± 10%, error margin 3.1 ± 2.3 mm, average time around 30 s) support their use for image guided online adaptive radiotherapy of cervical cancer.
Texto completo:
1
Base de dados:
MEDLINE
Assunto principal:
Processamento de Imagem Assistida por Computador
/
Neoplasias do Colo do Útero
/
Colo do Útero
/
Radioterapia Guiada por Imagem
Idioma:
En
Ano de publicação:
2013
Tipo de documento:
Article