Detekcija srčanih šumova iz snimki fonokardiograma metodama dubokog učenja

master's thesis

• Title (english): Heart murmur detection from phonocardiogram recordings using deep learning methods
• Author: Marin Matjanec
• Mentor: Alan Jović (mentor)
• Committee member: Alan Jović (predsjednik povjerenstva)
• Committee member: Jan Šnajder (član povjerenstva)
• Committee member: Marko Čupić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-02-19, Croatia

Abstract

Urođene srčane bolesti stvaraju velik teret za zdravstveni sustav i odnose mnoge ljudske živote. Srčani šumovi često mogu ukazivati na neku takvu bolest. Fonokardiogram je zapis zvuka srca snimljen digitalnim stetoskopom. Računalnom analizom fonokardiograma i detekcijom srčanih šumova može se pomoći liječnicima kod dijagnoza srčanih bolesti. U ovom radu su ispitana i uspoređena 4 modela strojnog i 17 modela dubokog učenja na temelju uspješnosti detekcije srčanih šumova na skupu podataka CirCor DigiScope. Naučeni su modeli koji detektiraju srčane šumove s točnosti od 92 %, odzivom od 85 % i F1 mjerom od 78 %.

Abstract (english)

Congenital heart diseases create a great burden on the healthcare system and take many human lives. Heart murmurs can often indicate such a disease. A phonocardiogram is a recording of heart sounds taken with a digital stethoscope. Computer analysis of phonocardiograms and detection of heart murmurs can help doctors diagnose heart diseases. In this paper, 4 machine learning models and 17 deep learning models were tested and compared based on the success of heart murmur detection on the CirCor DigiScope dataset. Models were trained to detect heart murmurs with an accuracy of 92 %, recall of 85 % and F1 score of 78 %.

• Keywords: Fonokardiogram; srčani šum; srčani ciklus; strojno učenje; duboko učenje; Fourierova transformacija; spektrogram; konvolucijska mreža; povratna mreža; transformer
• Keywords (english): Phonocardiogram; heart murmur; cardiac cycle; machine learning; deep learning; Fourier transform; spectrogram; convolutional network; recurrent network; transformer
• Language: croatian
• URN:NBN: urn:nbn:hr:168:508171
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-03-28 07:50:50