Duboka neuronska mreža za segmentaciju slojeva retine

undergraduate thesis

• Title (english): Deep Neural Network for Retinal Layer Segmentation
• Author: Matej Salković
• Mentor: Sven Lončarić (mentor)
• Committee member: Sven Lončarić (predsjednik povjerenstva)
• Committee member: Ivana Matovinović (član povjerenstva)
• Committee member: Luka Posilović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-07-08, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Optička koherentna tomografija (OCT) je neinvanzivna metoda snimanja bioloških tkiva često korištena u oftalmologiji i medicini općenito. Precizna automatska segmentacija slojeva na OCT snimkama retine je od velike važnosti za oftalmologiju jer omogućava bržu i jednostavniju dijagnostiku. Ovaj rad se bavi implementacijom i analizom kaskadne arhitekture koja koristi mapu relativnih udaljenosti kao dodatan ulaz u mrežu. Arhitektura se sastoji od dvije međusobno povezane duboke neuronske mreže od kojih prva služi za ekstrakciju granica retine, a druga za segmentaciju svih slojeva i tekućina. Arhitektura kombinira potpuno konvolucijsku mrežu i U-Net arhitekturu kako bi poboljšala segmentaciju. U radu se detaljno opisuju arhitektura mreže i postupci treniranja i testiranja. Dobiveni rezultati se uspoređuju s očekivanim rezultatima.

Abstract (english)

Optical coherence tomography (OCT) is a non-invasive method of imaging biological tissues that is often used in ophthalmology and medicine in general. Precise automatic layer segmentation on OCT retinal images is of great importance for ophthalmology as it allows for faster and easier diagnosis. This bachelor thesis deals with the application and analysis of a cascading architecture that uses a relative distance map as an additional network input. The architecture consists of two interconnected deep neural networks of which the first serves to extract the boundaries of the retina, and the second to segment all layers and fluids. The architecture combines a fully convolutional network and U-Net architecture to enhance segmentation. The thesis describes in detail the network architecture and training and testing procedures. The obtained results are compared with the expected results.

• Keywords: retina; segmentacija slojeva retine; optička koherentna tomografija; potpuno konvolucijske mreže; U-Net; LF-UNet; relativna mapa udaljenosti
• Keywords (english): retina; retinal layers segmentation; optical coherence tomography; fully convolutional networks; U-Net; LF-UNet; relative distance map
• Language: croatian
• URN:NBN: urn:nbn:hr:168:431565
• 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: 2023-01-14 09:53:36