| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 3.97 MB | Adobe PDF |
Advisor(s)
Abstract(s)
Este documento foi redigido no âmbito da dissertação do Mestrado em Engenharia
Informática na área de Arquiteturas, Sistemas e Redes, do Departamento de Engenharia
Informática, do ISEP, cujo tema é diagnóstico cardíaco a partir de dados acústicos e clínicos.
O objetivo deste trabalho é produzir um método que permita diagnosticar
automaticamente patologias cardíacas utilizando técnicas de classificação de data mining.
Foram utilizados dois tipos de dados: sons cardíacos gravados em ambiente hospitalar e dados
clínicos. Numa primeira fase, exploraram-se os sons cardíacos usando uma abordagem baseada
em motifs. Numa segunda fase, utilizamos os dados clínicos anotados dos pacientes. Numa
terceira fase, avaliamos a combinação das duas abordagens. Na avaliação experimental os
modelos baseados em motifs obtiveram melhores resultados do que os construídos a partir dos
dados clínicos. A combinação das abordagens mostrou poder ser vantajosa em situações
pontuais.
This document was written as part of the Thesis of the MSc in computer science in the area of Architecture, System and Network, Department of Computer Engineering in ISEP. The main theme of this Thesis is to diagnose cardiac diseases, through acoustic and clinical data. The goal of this work is to produce a process for automatically diagnosing heart problems using data mining classification techniques. Two types of data were used: heart sounds recorded in hospitals and clinical data. Initially, we explored the heart sounds using an approach based on motifs. In a second stage, we used the clinical data of the patients. In a third phase, we evaluated the combination of both approaches. Experimental evaluation showed that models based on motifs performed better than those built from clinical data. The combination of approaches has shown to be advantageous in specific situations.
This document was written as part of the Thesis of the MSc in computer science in the area of Architecture, System and Network, Department of Computer Engineering in ISEP. The main theme of this Thesis is to diagnose cardiac diseases, through acoustic and clinical data. The goal of this work is to produce a process for automatically diagnosing heart problems using data mining classification techniques. Two types of data were used: heart sounds recorded in hospitals and clinical data. Initially, we explored the heart sounds using an approach based on motifs. In a second stage, we used the clinical data of the patients. In a third phase, we evaluated the combination of both approaches. Experimental evaluation showed that models based on motifs performed better than those built from clinical data. The combination of approaches has shown to be advantageous in specific situations.
Description
Keywords
Sons cardíacos Motifs Dados clínicos Classificação Diagnóstico Heart Sounds Motifs Clinic data Classification Diagnostic