Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/132968
Title: Detección precoz de covid-19 a partir de imágenes de radiografías de tórax mediante redes neuronales convolucionales
Author: Alfonso López, Blanca
Tutor: Rebrij, Romina  
Others: Perez-Navarro, Antoni  
Abstract: The world is currently facing a global health crisis caused by the Covid-19 disease. Since its appearance in Wuhan (China), the virus has caused around four million deaths worldwide, causing high hospital pressure. The virus stands out for its high propagation capacity. For this reason, it seems essential to have mechanisms for early detection of the disease to increase the speed of treatment and reduce the probability of contagion. Chest X-rays are a very effective test for detecting Covid-19 pneumonia. Deep learning is able to extract characteristics related to clinical results from radiographs. In this Master's Thesis, a convolutional neural network model has been proposed and an attempt has been made to improve it through the use of data augmentation, learning transfer and fine-tuning techniques. The project has been developed in the Python programming language and using the Tensorflow and Keras libraries. For the training of the model, a database of images of chest radiographs from the Kaggle repository was used. These images are classified into four categories: normal, thoracic opacity, viral pneumonia and Covid-19 pneumonia. Finally, a comparison of the created models has been made. The considered best model has been used to show the results through a web application developed with Flask and published on the PythonAnyWhere platform.
Keywords: deep learning
COVID-19
x-rays
Document type: info:eu-repo/semantics/masterThesis
Issue Date: Jun-2021
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Trabajos finales de carrera, trabajos de investigación, etc.

Files in This Item:
File Description SizeFormat 
balfonsolTFM0621memoria.pdfMemoria del TFM1,63 MBAdobe PDFThumbnail
View/Open