Venous tree separation in the liver: graph partitioning using a non-ising model.
Inf Process Med Imaging
; 22: 197-207, 2011.
Article
em En
| MEDLINE
| ID: mdl-21761657
ABSTRACT
Entangled tree-like vascular systems are commonly found in the body (e.g., in the peripheries and lungs). Separation of these systems in medical images may be formulated as a graph partitioning problem given an imperfect segmentation and specification of the tree roots. In this work, we show that the ubiquitous Ising-model approaches (e.g., Graph Cuts, Random Walker) are not appropriate for tackling this problem and propose a novel method based on recursive minimal paths for doing so. To motivate our method, we focus on the intertwined portal and hepatic venous systems in the liver. Separation of these systems is critical for liver intervention planning, in particular when resection is involved. We apply our method to 34 clinical datasets, each containing well over a hundred vessel branches, demonstrating its effectiveness.
Buscar no Google
Coleções:
01-internacional
Base de dados:
MEDLINE
Assunto principal:
Veia Porta
/
Reconhecimento Automatizado de Padrão
/
Inteligência Artificial
/
Flebografia
/
Interpretação de Imagem Radiográfica Assistida por Computador
/
Tomografia Computadorizada por Raios X
/
Imageamento Tridimensional
Tipo de estudo:
Diagnostic_studies
/
Prognostic_studies
Limite:
Humans
Idioma:
En
Revista:
Inf Process Med Imaging
Assunto da revista:
DIAGNOSTICO POR IMAGEM
/
INFORMATICA MEDICA
Ano de publicação:
2011
Tipo de documento:
Article
País de afiliação:
Estados Unidos