Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/122926
Title: Aprendizaje supervisado en detección de osteoartritis de rodilla a partir de imágenes de resonancia magnética
Author: Oyarzo Huichaqueo, Marco Antonio
Tutor: Puig Valls, Domènec Savi
Abstract: Osteoarthritis (OA) is a common disease worldwide, which especially affects older adults, causing physical disability and an impact on quality of life. In the process of diagnosing OA, a specialist doctor uses medical images to diagnose its degree of progress based on the pathophysiological analysis of the disease. Today, this complex diagnostic task is a research field for machine learning (ML) and its form of supervised learning. This Work presents the development of ML classifiers for the classification of knee OA, according to the Kellgren-Lawrence severity scale. For the development of the classifiers, different convolutional neural networks (CNN) were studied: VGG16, InceptionV3, Xception, MobileNet and DenseNet121 based on magnetic resonance images (MRI) of patients, obtained from the Osteoarthritis Initiative (OAI) database. The Work approach combines two transfer learning methods: classification models from pre-trained CNNs; and support vector machine (SVM) classification models based on features extracted from pre-trained CNNs and the principal component analysis (PCA) technique. Furthermore, in order to increase the prediction accuracy of the trained models for OA classification, an assembly method was studied: the majority vote technique. Finally, the results obtained from the experiment show that the assembly method is capable of increasing the prediction accuracy from 55.3% to 60.6% in the OA classification.
Keywords: transfer of learning
machine learning
osteoarthritis diagnosis
Document type: info:eu-repo/semantics/masterThesis
Issue Date: Sep-2020
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
moyarzoTFM0920memoria.pdfMemoria del TFM3,31 MBAdobe PDFThumbnail
View/Open