Titulo Estágio
Prediction for the Obstructive Sleep Apnea Sindrome: a bio-inspired approach
Área Tecnológica
Sistemas Evol. e Comp.
Local do Estágio
CISUC- EcosLab
Enquadramento
The goal of this project is to propose, implement, test and validate a computational framework to help the clinical decision maker in the diagnosis and early assessment of the obstructive sleep apnea (OSA) disease. We aim to propose an approach to the problem of transforming the implicit hidden knowledge, present in the classified data collected from different patients with sleep problems, into an accurate, comprehensible and useful model that support the decision. The approach will rely on the automatic construction of a decision model from a real world clinical database, provided by the Centro de Medicina do Sono (CMS), do Centro Hospitalar Universitário de Coimbra, using bio-inspired data mining algorithms. The approach we propose to develop will be based on Evolutionary Algorithms (EA). EA are stochastic, global search procedures inspired in the principles of natural selection and genetics. They are particular well suited for problems that either do not have an analytical solution or have a huge search space, making traditional approaches not applicable as it is the case of OSA.
Objetivo
Our work will put together the areas of knowledge extraction from data bases, in particular machine learning algorithms for data mining tasks, biological-inspired computation and medical prediction, in the practical context of the obstructive sleep apnea disease. The data to be mined are presented in different, heterogeneous, formats, e.g., including questionnaires, clinical records, biological signals. The first question to be solved will be that of the integration of the data into an unique database involving attributes and the corresponding values. Over the years a vast number of algorithms and techniques were proposed to tackle with the question of knowledge extraction from databases, less so for doing KDD for medical applications, and only rare studies dealt with (some aspects) of OSA. As a consequence of the absence of computer-based classifiers for OSA, with practical application in the day life of a physician, the project aims at providing one. A distinctive trait of our approach is the fact that we will use a data mining learning algorithm that is based on an evolutionary algorithm. This will be the second question to be solved within the master thesis work. Finally, we will try a first validation of the prototype. To summarize, we are proposing a novel approach to the problem of producing a classifier for the prediction of the Obstructive Sleep Apnea (OSA) disease, that can help physicians in their practice.
Plano de Trabalhos - Semestre 1
First Semester:
Task 1 [September - November]: Literature Review (Obstructive Sleep Apnea Disease, Data Mining Techniques, Evolutionary Algorithms for Data Mining)
Task 2 [December - January]: Cleaning and Integration of the Data Base
Task 3 [[September - January]: Document about the literature review, writing the report for the public discussion.
Plano de Trabalhos - Semestre 2
Second Semester
Task 4 [February - May]: Design and implementation of a classifier for OSA
Task 5 [May - June]: Test and validation of the classifier
Task 6 [May - June]: Writing a paper to be submitted to an international conference. Writing of the final Report.
Condições
O estágio não é remunerado. Envolverá visitas regulares ao Centro de Medicina do Sono, nos Covões.
Observações
O aluno deve ter conhecimentos de programação numa linguagem de alto nível (de preferência Python), ter conhecimentos genéricos na área de Inteligência Artificial e, preferencialmente, na área de Computação Evolucionária.
Orientador
Ernesto Costa
ernesto@dei.uc.pt 📩