DEFESA DE TESE DE DOUTORADO Nº 25
Aluno: Sebastião Rogério da Silva Neto
Título: “Clinical decision support for arboviral diseases using machine learning models"
Orientadora: Patricia Takako Endo
Coorientador: Vanderson de Souza Sampaio (UEA/FMT-HVD)
Examinador Externo: Ivanovitch Silva (UFRN)
Examinador Externo: Marcelo Anderson Batista Dos Santos (IF-Sertão)
Examinador Externo: Moacyr Jesus Barreto de Melo Rego (UFPE)
Examinador Interno: Cleyton Mário de Oliveira Rodrigues
Examinador Interno: Wellington Pinehrio dos Santos
Data-hora: 23 de outubro de 2024, às 14:30.
Local: Formato Híbrido (no PPGEC Caruaru e Online).
Resumo:
"The 2030 Agenda is a global plan by the United Nations Organization to achieve a better world for all people and nations by 2030. This Agenda proposed 17 Sustainable Development Goals (SDGs), including SDG 3 "Good Health and Well-being", which aims to end epidemics of AIDS, tuberculosis, malaria, and neglected tropical diseases (NTDs), as well as combat hepatitis, water-borne diseases, and other communicable diseases. Among the NTDs, there are arboviruses, which cause a wide range of diseases, the most common of which are Dengue, Chikungunya, and Zika. These arboviruses are transmitted by mosquitoes, such as Aedes aegypti and Aedes albopictus. Due to factors such as climate change, deforestation, population migration, and precarious sanitary conditions, the arboviruses transmitted by these mosquitoes have become a global health problem. Another critical aspect of these arboviruses is their clinical presentation. Despite being well-established diseases, they are difficult to diagnose. Most infections are asymptomatic, which means that arboviruses can be present in an area without causing outbreaks. Their symptomatic infections are typically clinically indistinguishable. Common symptoms include fever, arthralgia, myalgia, headache, and retro-orbital pain. Early detection of specific arbovirus infections can have a significant impact on the clinical course, treatment, and care decisions. In this regard, scalable and low-cost implementation strategies are required to aid in the differential diagnosis of these diseases. One such strategy is the development of computational models to assist in diagnosis based on clinical data and symptoms. In this thesis proposal, we present an application focused on the development of a model that aids in the classification of these arboviruses using only clinical data from patients. To build the models, machine learning techniques (Random Forest, Adaboost, Gradient Boosting, XGBoost, KNN, Naive Bayes) were employed, along with feature selection and hyperparameter optimization. Explainable AI (XAI) was utilized to enhance the model's interpretability. To address the diagnostic challenges, we continuously train and evaluate the proposed machine learning models using newly collected clinical data from the Sistema de Informação de Agravos de Notificação (SINAN), which is validated by healthcare professionals, in order to improve the model in production regularly."
O Programa
