Quantification of cardiac patterns using a spatio-temporal analysis of the right ventricle in magnetic resonance imaging
Abstract. An accurate quantification of the right ventricular function is important to support the evaluation, diagnosis and prognosis of several cardiac diseases, as well to complement the analysis of the left ventricular function. Traditionally, this quantification is performed by the manual delin...
- Autores:
-
Atehortúa Labrador, Angelica María
- Tipo de recurso:
- Fecha de publicación:
- 2014
- Institución:
- Universidad Nacional de Colombia
- Repositorio:
- Universidad Nacional de Colombia
- Idioma:
- spa
- OAI Identifier:
- oai:repositorio.unal.edu.co:unal/52020
- Acceso en línea:
- https://repositorio.unal.edu.co/handle/unal/52020
http://bdigital.unal.edu.co/46269/
- Palabra clave:
- 0 Generalidades / Computer science, information and general works
61 Ciencias médicas; Medicina / Medicine and health
Right ventricle segmentation
Local motion model
Structural information
Non rigid registration
Cardiac patterns
Segmentación del ventrículo derecho
Modelo de movimiento local
Información estructural
Registro no rígido
Patrones cardíacos
- Rights
- openAccess
- License
- Atribución-NoComercial 4.0 Internacional
| Summary: | Abstract. An accurate quantification of the right ventricular function is important to support the evaluation, diagnosis and prognosis of several cardiac diseases, as well to complement the analysis of the left ventricular function. Traditionally, this quantification is performed by the manual delineation of the right ventricle (RV), a time consuming task that introduces high inter and intra observer variability. Different semi-automatic and fully-automatic a- pproaches have been proposed to delineate the RV. Nevertheless, most of these methods, e.g multi-atlas strategies and approaches with training of parameters, are strongly dependent on RV shapes present in the training set utilized. This thesis presents a fully automatic segmentation strategy of the RV in MRI-cardiac sequences to quantify cardiac patterns. Unlike multi-atlas methods, the proposed strategy estimates the RV using exclusively information from the sequence itself. The core of this work is a spatio-temporal analysis that is per- formed by using a motion descriptor and anatomical heart information, taking advantage of natural heart dynamic. This strategy was improved from two perspectives: 1) by introducing a prior regularizer term that smooths the obtained delineation and 2) by propagating the spatio-temporal segmentation from the basal slice to apex using a non-rigid registration. The proposed approach achieves an average Dice Score of 0.78 evaluated in a set of 48 patients. |
|---|