Métodos matemáticos y computacionales para elastografía médica
Loading...
Official URL
Full text at PDC
Publication date
2021
Defense date
2021
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
La elastografía médica tienen como objetivo reconstruir las posibles anomalías de un tejido blando del cuerpo analizando su rigidez. En este trabajo proponemos una aproximación determinista y otra Bayesiana que dan solución a este tipo de problemas cuando el número de anomalías es conocido: dados unos datos medidos por mecanismos de elastografía, tratamos de determinar la elasticidad, ubicación y forma de las anomalías que mejor se ajustan a los datos medidos minimizando una función de error cuadrática. Este problema de optimización en su formulación determinista reconstruye de forma razonable los parámetros involucrados, sin embargo no nos proporciona información estadística de los resultados obtenidos. Es por ello que se plantea también la formulación Bayesiana del problema. Para el método Bayesiano, en primer lugar se obtiene el máximo a posteriori minimizando una función de coste con términos regularizadores y después se aproxima la distribución a posteriori mediante la aproximación de Laplace.
Medical elastography aims to reconstruct the possible anomalies of a soft tissue of the body by analyzing its stiffness. In this work we propose a deterministic and a Bayesian approach to solve this type of problem when the number of anomalies is known: given the data measured by elastography techniques, we try to determine the elasticity, location and shape of the anomalies that best fit the measured data by minimizing a quadratic error function. This optimization problem in its deterministic formulation reconstructs in a reasonable way the parameters involved, however it does not provide us statistical information of the results obtained. This is why the Bayesian formulation of the problem is also proposed. For the Bayesian method, we calculate the maximum a posteriori minimizing a regularized cost function and then the posterior distribution is approximated by the Laplace approximation.
Medical elastography aims to reconstruct the possible anomalies of a soft tissue of the body by analyzing its stiffness. In this work we propose a deterministic and a Bayesian approach to solve this type of problem when the number of anomalies is known: given the data measured by elastography techniques, we try to determine the elasticity, location and shape of the anomalies that best fit the measured data by minimizing a quadratic error function. This optimization problem in its deterministic formulation reconstructs in a reasonable way the parameters involved, however it does not provide us statistical information of the results obtained. This is why the Bayesian formulation of the problem is also proposed. For the Bayesian method, we calculate the maximum a posteriori minimizing a regularized cost function and then the posterior distribution is approximated by the Laplace approximation.