Thesis MODELADO DE CONECTIVIDAD CEREBRAL EN REPOSO CON MODELOS AUTOREGRESIVOS MULTIVARIADOS Y ESTIMACIÓN EFICIENTE DE SU ORDEN
Date
2018
Authors
GUERRA PINTO, VICTOR DANIEL
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
En el area de modelado de funciones cognitivas, uno de los problemas de interés esel modelado de conectividad cerebral en estado de reposo, donde se busca poder identificarpatrones de actividad cerebral, cuando el sujeto experimental no se encuentra haciendoninguna tarea cognitiva en particular, esta indentificación normalmente es realizadapor un experto que pueda identificar los patrones de actividad viendo por ejemplo, unelectroencefalograma (EEG), esta tarea es hecha con una muy baja resolución temporal,con un etiquetado de señales por segmentos de 30 segundos (a este proceso de etiqutadopor segmentos se le dice segmentación). Una forma de facilitar el estudio de modeloscerebrales es hacer esta identificación automáticamente. Actualmente se han propuestovarios algoritmos y modelos en la literatura de series de tiempo y machine learning pararesolver problemas de segmentación de señales, entre estos, los modelos escondidos deMarkov Autoregresivos Multivariados (HMM-MAR en inglés), han mostrado en la literaturaque reuelven de manera eficaz la segmentación. Aquí se muestran métodos para hacer la estimaciónde estos modelos de Markov y resolver el problema de segmentación, probandoloscon distintos modelos verificando su desempeño en un entorno sintético para posteriormenteproponer un experimento con mediciones fisioógicas reales.
In the fields of functional connectivity modelling, a problem of interest is to modelbrain connectivity in resting state, where is sought to identify patterns of brain activity, whenthe experimental subject isn’t doing any cognitive task in particular, this identification, isusually performed by an expert who can identify the activity patterns by seeing, for example,an electroencephalogram (EEG), this task is done with a very low temporal resolution, witha labeling of signals per 30 second segments (this process of labeling by segments is calledsegmentation). One way to facilitate the study of brains models is to make this identificationautomatically. Currently, several algorithms and models have been propose in the machinelearning and time series and machine learning literature to solve problems of signalsegmentation, among then, the hidden Markov multivariate Autoregressive (HMM-MAR))has shown in the literature that it effectively solves the segmentation. Here, different methodsto estimate these Markov models and solving the segmentation problem, testing them withdifferent models, verifyng their performance in a synthetic environment to propose anexperiment with real physiological measeurements.
In the fields of functional connectivity modelling, a problem of interest is to modelbrain connectivity in resting state, where is sought to identify patterns of brain activity, whenthe experimental subject isn’t doing any cognitive task in particular, this identification, isusually performed by an expert who can identify the activity patterns by seeing, for example,an electroencephalogram (EEG), this task is done with a very low temporal resolution, witha labeling of signals per 30 second segments (this process of labeling by segments is calledsegmentation). One way to facilitate the study of brains models is to make this identificationautomatically. Currently, several algorithms and models have been propose in the machinelearning and time series and machine learning literature to solve problems of signalsegmentation, among then, the hidden Markov multivariate Autoregressive (HMM-MAR))has shown in the literature that it effectively solves the segmentation. Here, different methodsto estimate these Markov models and solving the segmentation problem, testing them withdifferent models, verifyng their performance in a synthetic environment to propose anexperiment with real physiological measeurements.
Description
Catalogado desde la version PDF de la tesis.
Keywords
MODELO AUTOREGRESIVO , MODELO ESCONDIDO DE MARKOV , SEGMENTACION , SENAL , SERIE DE TIEMPO