Use this link to cite:
http://hdl.handle.net/2183/39307 Análisis de Características Profundas mediante Transferencia de Aprendizaje para la Detección de Enfermedades Neurodegenerativas a través de Imágenes OCT
Loading...
Identifiers
Publication date
Authors
Rivera Caballero, Diego
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: Las enfermedades neurodegenerativas, como la Enfermedad de Alzheimer, la Enfermedad de Parkinson, la Esclerosis Múltiple y el Temblor Esencial, constituyen un desafío creciente para la salud pública a nivel mundial, especialmente en un contexto de envejecimiento de la población y aumento de la esperanza de vida. Estas cuatro patologías, caracterizadas por un deterioro progresivo del sistema nervioso, afectan profundamente la calidad de vida de los pacientes, causando dificultades motoras, cognitivas y emocionales. En este sentido, la detección temprana de estas enfermedades es crucial para la implementación de tratamientos que puedan ralentizar su progreso y mejorar el pronóstico de los pacientes. La Tomografía de Coherencia Óptica (OCT) se ha consolidado como una herramienta no invasiva fundamental en la oftalmología, y su aplicación en el diagnóstico de enfermedades neurodegenerativas ha ganado creciente interés en la comunidad científica. Estudios recientes sugieren que cambios sutiles en la retina y el nervio óptico pueden ser indicadores tempranos de estas enfermedades, lo que subraya la importancia de desarrollar sistemas automatizados que permitan su detección precoz a partir de imágenes OCT. Este Trabajo de Fin de Grado se centra en la creación de un sistema automatizado que emplea técnicas avanzadas de aprendizaje profundo y transferencia de aprendizaje para la identificación temprana de cuatro enfermedades neurodegenerativas, a partir de imágenes OCT retinianas. Utilizando la arquitectura ResNet-50, se llevó a cabo una extracción de características profundas de las imágenes, seguida de una selección optimizada de dichas características mediante métodos como PCA (Análisis de Componentes Principales) y K-Best. Estos procesos fueron fundamentales para entrenar y evaluar diversos clasificadores, con el objetivo de maximizar la precisión diagnóstica y minimizar la variabilidad interpretativa inherente al análisis humano. El presente estudio no solo representa una contribución significativa al Estado del Arte en el ámbito de la neuro-oftalmología, al ser uno de los primeros en explorar la eficacia de las características profundas combinadas con diferentes profundidades de extractores y selectores de características, sino que también ofrece una base sólida para futuras investigaciones y desarrollos tecnológicos. La implementación de este sistema podría tener un impacto positivo en el diagnóstico temprano y en el manejo clínico de las enfermedades neurodegenerativas, mejorando la calidad de vida de los pacientes y optimizando los recursos médicos.
[Abstract]: Neurodegenerative diseases, such as Alzheimer’s Disease, Parkinson’s Disease, Multiple Sclerosis, and Essential Tremor, pose a growing challenge to public health worldwide, particularly in the context of an aging population and increasing life expectancy. These four pathologies, characterized by the progressive deterioration of the nervous system, profoundly affect patients’ quality of life, causing motor, cognitive, and emotional difficulties. In this sense, early detection of these diseases is crucial for implementing treatments that can slow their progression and improve patients’ prognosis. Optical Coherence Tomography (OCT) has established itself as a fundamental non-invasive tool in ophthalmology, and its application in the diagnosis of neurodegenerative diseases has garnered increasing interest in the scientific community. Recent studies suggest that subtle changes in the retina and optic nerve can be early indicators of these diseases, highlighting the importance of developing automated systems that enable their early detection using OCT images. This Final Degree Project focuses on creating an automated system that employs advanced deep learning and transfer learning techniques for the early identification of four neurodegenerative diseases using retinal OCT images. Utilizing the ResNet-50 architecture, deep features were extracted from the images, followed by optimized feature selection using methods such as PCA (Principal Component Analysis) and K-Best. These processes were crucial for training and evaluating various classifiers, aiming to maximize diagnostic accuracy and minimize the interpretative variability inherent to human analysis. This study not only represents a significant contribution to the state of the art in neuroophthalmology, being one of the first to explore the effectiveness of deep features combined with different depths of feature extractors and selectors, but also provides a solid foundation for future research and technological developments. The implementation of this system could have a positive impact on the early diagnosis and clinical management of neurodegenerative diseases, improving patients’ quality of life and optimizing medical resources.
[Abstract]: Neurodegenerative diseases, such as Alzheimer’s Disease, Parkinson’s Disease, Multiple Sclerosis, and Essential Tremor, pose a growing challenge to public health worldwide, particularly in the context of an aging population and increasing life expectancy. These four pathologies, characterized by the progressive deterioration of the nervous system, profoundly affect patients’ quality of life, causing motor, cognitive, and emotional difficulties. In this sense, early detection of these diseases is crucial for implementing treatments that can slow their progression and improve patients’ prognosis. Optical Coherence Tomography (OCT) has established itself as a fundamental non-invasive tool in ophthalmology, and its application in the diagnosis of neurodegenerative diseases has garnered increasing interest in the scientific community. Recent studies suggest that subtle changes in the retina and optic nerve can be early indicators of these diseases, highlighting the importance of developing automated systems that enable their early detection using OCT images. This Final Degree Project focuses on creating an automated system that employs advanced deep learning and transfer learning techniques for the early identification of four neurodegenerative diseases using retinal OCT images. Utilizing the ResNet-50 architecture, deep features were extracted from the images, followed by optimized feature selection using methods such as PCA (Principal Component Analysis) and K-Best. These processes were crucial for training and evaluating various classifiers, aiming to maximize diagnostic accuracy and minimize the interpretative variability inherent to human analysis. This study not only represents a significant contribution to the state of the art in neuroophthalmology, being one of the first to explore the effectiveness of deep features combined with different depths of feature extractors and selectors, but also provides a solid foundation for future research and technological developments. The implementation of this system could have a positive impact on the early diagnosis and clinical management of neurodegenerative diseases, improving patients’ quality of life and optimizing medical resources.
Description
Keywords
Aprendizaje profundo Neuro-oftalmología Retina Tomografía de coherencia Óptica (OCT) Visión por computador Inteligencia artificial Enfermedad de alzheimer Enfermedad de parkinson Esclerosis múltiple Temblor esencial Deep learning Neuro-ophthalmology Optical coherence tomography (OCT) Computer vision Artificial intelligence Alzheimer’s disease Parkinson’s disease Multiple sclerosis Essential tremor
Editor version
Rights
Atribución-NoComercial-CompartirIgual 3.0 España








