Por favor, use este identificador para citar o enlazar este ítem:
http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/1997| Detection and classification of the nuclear cataract with computational intelligence methods | |
| HANS ISRAEL MORALES LOPEZ | |
| ISRAEL CRUZ VEGA | |
| Acceso Abierto | |
| Atribución-NoComercial-SinDerivadas | |
| NUCLEAR CATARACT MACHINE LEARNING DEEP LEARNING PRINCIPAL COMPONENT ANALYSIS | |
| A nuclear cataract is an eye disease affecting a large amount of the planet’s population at an advanced age. An exhaustive eye examination through a slit lamp by an ophthalmologist detects the presence and the extent of a cataract. In the last decade, some methods using computational intelligence have been developed to solve the automatic cataract classification problem. Image processing techniques do the job of extracting certain features from the image. Subsequently, the classification is done using classical methods of computational intelligence, fed with the previously extracted characteristics, such as multilayer perceptron (MLP) or support vector machines (SVM). The traditional machine learning approach is still prevalent, and many commercial medical image systems rely on this basis. Still, it requires manual or handcrafted feature extraction by an expert. Moreover, the use of high-accuracy and complex ANN-structures, like deep learning, lead to an automatic definition of high- level features. To solve classification problems, systems designed entirely by humans have evolved into systems that automatically learns features for the problem at hand. With the improvement of data processing technology, computing power and deep architectures, the levels of accuracy have been increased due to the possibility of extract high-order features. In this thesis, a novel and specialized framework based on a deep architecture is proposed, from which we can reach high classification accuracy. Also, we compare the proposed deep network with some famous deep network architectures, like AlexNet, GoogLeNet and also with traditional classifiers, such as ANN and SVM. The research framework includes a structural analysis of our proposed network performing the principal component analysis to minimize the number of layers and parameters. The experimental results of this system based on a con- volutional neural network provide promising and competitive results to the existing investigations for this classification problem. Besides, this framework shows us that systems based on deep architectures have a great potential to deal with the problem of classification of medical images. | |
| Instituto Nacional de Astrofísica, Óptica y Electrónica | |
| 2020-12 | |
| Tesis de doctorado | |
| Inglés | |
| Estudiantes Investigadores Público en general | |
| Morales López, H.I., (2020), Detection and classification of the nuclear cataract with computational intelligence methods, Tesis de Doctorado, Instituto Nacional de Astrofísica, Óptica y Electrónica. | |
| ELECTRÓNICA | |
| Versión aceptada | |
| acceptedVersion - Versión aceptada | |
| Aparece en las colecciones: | Doctorado en Electrónica |
Cargar archivos:
| Fichero | Descripción | Tamaño | Formato | |
|---|---|---|---|---|
| Hans-Israel-Tesis.pdf | 41.22 MB | Adobe PDF | Visualizar/Abrir |