Klasifikacija uzoraka HEp-2 stanica

master's thesis

• Title (english): Classification of human epithelial cell's staining patterns
• Author: Sebastijan Dumančić
• Mentor: Damir Seršić (mentor)
• Committee member: Hendrik Blockeel (predsjednik povjerenstva)
• Committee member: Leander Schietgat (član povjerenstva)
• Committee member: Philip Dutre (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2014-06-30, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad predstavlja novi pogled na računalno podržanu dijagnostiku u medicini. Detekcija
autoimunih bolesti uzeta je kao studij slučaja, a zadatak je prepoznati različite uzorke
stanice koji su vezani za specifične autoimune bolesti. Naglasak je stavljen na interpretaciju
modela, umjesto preciznosti. Rješenje se sastoji od 3 koraka - segmentacije stanica, detekciju
razine fluorescentnog intenziteta te indukcije pravila koja opisuju uzorke. Interpretabilnost
modela postignuta je učenje simboličke reprezentacije stanica pomoću dubokih generativnih
metoda strojnog učenja. Performanse četiri različite metode indukcije pravila - FOIL, ALEPH,
RIPPER i stabla odluke - su uspoređena. Performanse predloženog rješenja usporediva
su s do sada poznatim najuspješnijim rješenjima ovog problema, uz dodatak da rješenje može
objasniti odluku, što ga čini vrlo primjenjivim u kontekstu medicinske dijagnostike.

Abstract (english)

In this work we present a new perspective on computer aided assistance in medical domains.
As a study case, a diagnosis of autoimmune diseases is taken, where the system should
recognize different staining patterns related to specific autoimmune diseases. The emphasis is
put on the interpretation of a model, instead of accuracy. The proposed solution is composed of the three consecutive steps - cell segmentation, fluorescence intensity level classification and rule induction. The interpretability of the model is achieved by learning a symbolic representation, with regard to visual appearance, by a Deep belief network. The performance of four commonly used algorithms, FOIL, ALEPH, RIPPER and Decision trees, are compared.
The performance of the proposed solution is comparable with the state-of-the-art solutions reported so far, but also provides the explanation of decision, which makes it very suitable in
computer aided assistance context.

• Keywords: strojno učenje; duboko učenje; analiza slike; dijagnostika autoimunih bolesti
• Keywords (english): machine learning; deep learning; image analysis; autoimmune disease diagnosis; rule mining
• Language: english
• URN:NBN: urn:nbn:hr:168:422827
• 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: 2018-02-06 08:36:43