Metode za otkrivanje mikrovoltažnih elektrofizioloških potencijala srčanog mišića

master's thesis

• Title (english): Methods for detection of microvoltage electrophysiological potentials of the heart muscle
• Author: Ivan Glogar
• Mentor: Andrej Novak (mentor)
• Committee member: Andrej Novak (predsjednik povjerenstva)
• Committee member: Nenad Pavin (član povjerenstva)
• Committee member: Vedran Đerek (član povjerenstva)
• Committee member: Danko Radić (član povjerenstva)
• Committee member: Matko Glunčić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2024-02-26, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

Kasni ventrikulski potencijali su visokofrekventni, niskoamplitudni signalicija prisutnost u elektrokardiogramu (EKG) ima visoku prediktivnu vrijednost za potencijalno smrtonosne ventrikulske aritmije. U ovom radu implementiramo 3 metode detekcije takvih signala iz 80 EKG zapisa pacijenata porijeklom iz baze podataka PTB Diagnostic ECG Database. Metode uključuju široko prihvaćenu metodu usrednjenja po signalima (SAECG), analize marginalnih komponenti Guaragnella et al. (2019), te modele strojnog učenja temeljene na modelu stabla odluke. QRS komplekse EKG signala u fazi pretprocesiranja detektiramo po vrhu R vala, poravnavamo metodama ”Double-level” i ”Matched filter”, te filtriramo po sličnosti i frekvencijama signala. Koristimo EKG snimke zdravih pacijenata na koje umjetno dodajemo VLP signale na temelju njihovih utvrđenih karakteristika. U korištenju metoda SAECG i metode analize marginalnih komponenti nije ostvarena izrazito visoka točnost detekcije. S druge strane, vrijednost modela strojnog učenja leži u mogućnosti modela za detekciju skrivenih signala na temelju jednog srčanog ciklusa; ovdje, modeli pokazuju značajnu točnost detekcije, gdje najbolji model detektira tražene signale s točnošću od 98%.

Abstract (english)

Ventricular late potentials are high-frequency, low-amplitude signals whose presence in the electrocardiogram (ECG) has high predictive value for potentially fatal ventricular arrhythmias. In this work, we implement three methods to detect such signals from 80 ECG recordings of patients from the PTB Diagnostic ECG Database. The methods include the widely accepted signal-averaged ECG (SAECG) method, the marginal component analysis method by Guaragnella et al. (2019), and machine learning models based od decision tree model. ECG signal beats are detected in the preprocessing phase by the peak of the R wave of the ECG, aligned using ’Double-level’ and ’Matched filter’ methods, and filtered by similarity and frequencies. We use ECG recordings of healthy patients to which we artificially add VLP signals based on their established characteristics. SAECG methods and marginal component analysis did not achieve very high detection accuracy. On the other hand, the value of machine learning models lies in their ability to detect hidden signals based on a single beat; here, the models show significant detection accuracy, with the best model detecting the desired signals with 98% accuracy.

• Keywords: kasni ventrikulski potencijali; Signal Averaged ECG; marginalne komponente; strojno učenje; stablo odluke
• Keywords (english): ventricular late potentials; Signal Averaged ECG; marginal components; machine learning; decision tree
• Language: croatian
• URN:NBN: urn:nbn:hr:217:329335
• Study programme: Title: The University Integrated Undergraduate and Graduate Programme in Physics; Research Physics; specializations in: Research Physics; Course: Research Physics; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: sveučilišni magistar fizike (sveučilišni magistar fizike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2024-04-05 12:01:18