Publication: Detección de crisis epilépticas en el dominio temporal a partir del EEG mediante SVMs
Loading...
Identifiers
Publication date
2014-07
Defense date
2014-12-11
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
La aplicación del tratamiento de señales con fines médicos es una constante dentro del diagnóstico e interpretación de enfermedades y otras afecciones de carácter clínico. Pese a ello, todavía hay áreas donde es necesario una mayor investigación que garantice una correcta aplicación práctica.
Se calcula que hay 50 millones de afectados por la epilepsia en todo el mundo, algo que hace que la epilepsia sea tan común como la diabetes. Sin embargo, la epileptología sigue siendo una rama reservada a minorías de neurólogos, entre otras cosas, por el estrecho margen terapéutico en que se mueve el médico al afrontar esta enfermedad (se calcula que el 30% de los pacientes con crisis no pueden ser tratados con medicamentos)
El presente Proyecto Fin de Carrera se centra en la detección de crisis epilépticas en el dominio temporal, a partir de señales electroencefalográficas (EEG), mediante la obtención, y posterior procesado, de características intrínsecas a la naturaleza propia de la señal.
Para ello, se diseñará un clasificador, y se probará su rendimiento, con registros de señales obtenidos a partir de pacientes reales. Este diseño contará con una primera etapa en la cual se extraerán las características de la señal para después, en base a estas características, determinar si se ha producido una crisis epiléptica o no.
Este diseño ha sido implementado en Matlab, mediante el uso de una máquina de vectores soporte (SVM) utilizando una función de base radial (RBF) como Kernel. Además, los atributos seleccionados para la caracterización de la señal serán la Varianza y la Probabilidad Conjunta de la señal.
Una vez realizada la evaluación del clasificador diseñado para cada uno de los pacientes seleccionados, el proyecto termina con las conclusiones obtenidas y algunas indicaciones sobre posibles líneas futuras.
The application of signal processing for medical purposes is something usual in the diagnosis and interpretation of diseases and other injuries of a clinical nature. Nevertheless, there are still areas where further research is needed to ensure a proper practical application. There are 50 million people affected by epilepsy worldwide, which makes the epilepsy as common as diabetes. However, epileptology is a branch reserved for a minority of neurologists, due to the limitations to deal with this disease (around 30 % of patients with seizures cannot be treated with medication). This Master’s Thesis focuses on the detection of epileptic seizures in the time domain on electroencephalographic (EEG) signals, by processing the intrinsic characteristics of the given signal. For this purpose, a classifier has been designed, whose performance will be tested on signals obtained from real patients. The design will have a first stage, in which the characteristics of the signal will be extracted, and then, based on these characteristics, the seizure classification will be made. This design has been implemented in Matlab, using a support vector machine (SVM) with a Radial Basis Function (RBF) Kernel. In addition, the selected characteristics for the analysis will be the Variance and Joint Probability of the signal. Once the evaluation of the classifier designed is performed, the project is closed by the main conclusions obtained and some indications of potential future research lines.
The application of signal processing for medical purposes is something usual in the diagnosis and interpretation of diseases and other injuries of a clinical nature. Nevertheless, there are still areas where further research is needed to ensure a proper practical application. There are 50 million people affected by epilepsy worldwide, which makes the epilepsy as common as diabetes. However, epileptology is a branch reserved for a minority of neurologists, due to the limitations to deal with this disease (around 30 % of patients with seizures cannot be treated with medication). This Master’s Thesis focuses on the detection of epileptic seizures in the time domain on electroencephalographic (EEG) signals, by processing the intrinsic characteristics of the given signal. For this purpose, a classifier has been designed, whose performance will be tested on signals obtained from real patients. The design will have a first stage, in which the characteristics of the signal will be extracted, and then, based on these characteristics, the seizure classification will be made. This design has been implemented in Matlab, using a support vector machine (SVM) with a Radial Basis Function (RBF) Kernel. In addition, the selected characteristics for the analysis will be the Variance and Joint Probability of the signal. Once the evaluation of the classifier designed is performed, the project is closed by the main conclusions obtained and some indications of potential future research lines.
Description
Keywords
MATLAB, SVM, Proceso de señales, Electroencefalogramas, EEG, Diagnóstico médico, Epilepsia, Máquina de vectores soporte