Analiza kardioloških slika korištenjem duboke neuronske mreže

undergraduate thesis

• Title (english): Cardiac Image Analysis Using Deep Neural Network
• Author: Adam Vuković
• Mentor: Sven Lončarić (mentor)
• Committee member: Sven Lončarić (predsjednik povjerenstva)
• Committee member: Donik Vršnak (član povjerenstva)
• Committee member: Franko Šikić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-03, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom radu istražena je primjena dubokih neuronskih mreža za analizu kardioloških slika biopsija transplantiranih srca. Cilj rada bio je razviti model koji može klasificirati stupanj odbacivanja tkiva transplantiranog srca u kategorije 0R, 1R, 2R i 3R, koristeći dvije napredne arhitekture dubokih neuronskih mreža: DenseNet i ResNet.

Na početku rada objašnjene su osnove umjetnih neuronskih mreža. Predstavljena je motivacija za korištenje umjetnih neuronskih mreža, a zatim je objašnjeno na koji način rade i koje su razlike u raznim implementacijama istih. U drugom dijelu objašnjena je implementacija analize slika uz pomoć razvijenih modela umjetnih neuronskih mreža.

Eksperimentalni rezultati pokazali su da oba modela, DenseNet i ResNet, postižu zadovoljavajuće performanse, s blagom prednošću DenseNet modela kod analize slika prepariranih u formalinu, dok je ResNet model pokazao bolje performanse u generalizaciji kod slika koje su dehidrirane. Kroz detaljnu analizu performansi modela, ovaj rad pruža uvid u prednosti i nedostatke korištenih arhitektura te potencijalne smjernice za daljnja istraživanja i optimizaciju modela za analizu medicinskih slika.

Ovaj rad doprinosi razumijevanju primjene dubokih neuronskih mreža u medicinskoj dijagnostici, posebice u kontekstu transplantacija srca, te nudi vrijedne uvide za razvoj automatiziranih sustava za podršku u donošenju kliničkih odluka.

Abstract (english)

The aim of this paper is to examine the application of deep neural networks in analyzing cardiac images of transplanted heart biopsies. The goal was to develop a model that can classify the degree of tissue rejection of the transplant into categories 0R, 1R, 2R and 3R using DenseNet and ResNet, which are two modern architectures for deep neural networks.

The introductory section provides a review of artificial neural network basics. In explaining why artificial neural networks were used, there is an illustration as to how they function and their differences in various implementations. The second part describes how image analysis was done using the developed models of artificial neural networks.

Experiment results indicated both DenseNet and ResNet achieved satisfying performance, with a slight preference in favor of the DenseNet model when it comes to formaline-prepared pictures, while ResNet had better generalization capabilities on dehydrated images. By thoroughly analyzing the models’ performance, this research reveals some strengths and weaknesses inherent in the architectures employed here, as well as possible directions for further studies aimed at model optimization for medical imaging analysis.

This work contributes to the understanding of the application of deep neural networks in medical diagnostics, especially in the context of heart transplants, and offers valuable insights for the development of automated clinical decision support systems.

• Keywords: umjetne neuronske mreže; konvolucijske neuronske mreže; X-PCI; biopsije transplatiranih srca; PyTorch; DenseNet; ResNet
• Keywords (english): artificial neural networks; convolutional neural networks; X-PCI; transplanted hearts biopsies; PyTorch; DenseNet; ResNet
• Language: croatian
• URN:NBN: urn:nbn:hr:168:153879
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 22:29:18