Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/147608
Title: Diagnóstico automático de casos de riesgo de melanoma basado en imágenes dermatoscópicas
Author: Rodríguez Bajo, Antonio Carlos
Tutor: Divorra Vallhonrat, Teresa
Others: Merino, David  
Abstract: Melanoma is a skin cancer that can be life-threatening if not properly treated. Patient survival largely depends on early care by medical professionals. The main objective of this Final Project is to demonstrate the applied use of Data Science and Artificial Intelligence to create a support system capable of producing melanoma risk predictions from dermoscopic images. Following the planning of the Project in stages, these phases have been carried out: 1. Study of the state of the art in image processing with convolutional neural networks, selecting EfficientNet. 2. Analysis of open image datasets by ISIC, selecting the data from the challenge of the year 2019. 3. Experimentation and training of models in Google Cloud to obtain the optimal model. 4. Evaluation of performance and fairness metrics. 5. Implementation of the model in AWS cloud, complemented by a web application to perform diagnostics on new images. The outcome of the implementation is considered satisfactory, with the recommendation for its use in adult patients, both women and men, over 30 years of age, with light skin tones. Enhancements in the system derived from the improvement of the quality of the data, the assurance of its interpretability in clinical contexts and the application of MLOps to create and deploy new versions could lead to an implementation in a real-world production environment at the service of the medical community.
Keywords: computer vision
melanoma
machine learning
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: 15-Jan-2023
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 
zardemostolesTFG0123memoria.pdfMemoria del TFG1,95 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

This item is licensed under aCreative Commons License Creative Commons