Treniranje sustava za prepoznavanje slika u domeni zdravstva

master's thesis

• Title (english): Training systems for image recognition in the domain of healthcare
• Author: Andrea Mikulić
• Mentor: Sanja Seljan (mentor)
• Committee member: Ivan Dunđer (predsjednik povjerenstva)
• Committee member: Vjera Lopina (član povjerenstva)
• Committee member: Sanja Seljan (član povjerenstva)
• Granter: University of Zagreb; Faculty of Humanities and Social Sciences; (Department of information and Communication sciences); Zagreb
• Defense date and country: 2024-09-13, Croatia
• Scientific / art field, discipline and subdiscipline: SOCIAL SCIENCES; Information and Communication Sciences; Information and Software Engineering

Abstract

S razvojem strojnog učenja kreirala se i veća ponuda i potražnja za izradom programa koji bi se koristili u medicinske svrhe za poboljšanje i ubrzavanje dijagnostike. Takav program bi se mogao i iskoristiti u dermatologiji za prepoznavanje madeža i melanoma i liječnicima pomoći pri dijagnozi pacijenata.
Cilj ovog istraživanja bio je provesti treniranje i evaluaciju u klasifikaciji slike madeža i melanoma u području medicine te ispitati utjecaj veličine skupa podataka za treniranje na rezultate strojnog učenja. Obradio se postupak prikupljanja podataka za treniranje i testiranje sustava, izrada programa u sustavu Orange, te treniranje programa na različitim veličinama skupa za treniranje i njihovi rezultati.
Rad se sastoji od dva glavna dijela, teorijskog i praktičnog dijela. U teorijskom dijelu obradili su se pojmovi strojnog učenja i podvrste, s naglaskom na nadzirano strojno učenje, algoritmi klasifikacije, te utjecaj veličine skupa na krajnji rezultat programa strojnog učenja.
U praktičnom dijelu rada prikazao se program za prepoznavanje slika madeža i melanomu u sustavu Orange te se prikazalo kako različite veličine skupova za treniranje utječu na rezultat klasifikacije kroz algoritme logističke regresije, neuronskih mreža i naivni Bayes.

Abstract (english)

With the development of machine learning, a greater supply and demand was created for the creation of systems that could be used for medical purposes to improve and speed up diagnostics. Such a program could also be used in dermatology to identify moles and melanomas and help doctors diagnose patients.
The purpose of this study was to conduct training and evaluation in the classification of moles and melanoma images in the medical field and to examine the influence of the size of the training data set on the results of machine learning. It also further examined the process of collecting data for training and testing the system, creating the said program in the Orange application, and training the system on different sizes of the training set and showed their results.
The paper consists of two main parts, the theoretical and the practical part. In the theoretical part, the concepts of machine learning and its subtypes were discussed, with an emphasis on supervised machine learning, classification algorithms, and the influence of the size of the set on the final result of the machine learning program.
In the practical part of the work, the program for recognizing images of moles and melanoma in the Orange system was presented, and it was shown how different sizes of training sets affect the classification result through the logistic regression algorithm, neural networks and naive Bayes.

• Keywords: nadzirano strojno učenje; medicinske slike; madeži; klasifikacija; prepoznavanje slika; algoritmi; dermatologija; umjetna inteligencija
• Keywords (english): supervised machine learning; medical images; moles; classification; image recognition; algorithms; dermatology; artificial intelligence
• Language: croatian
• URN:NBN: urn:nbn:hr:131:271261
• Study programme: Title: Information Sciences; specializations in: Archivistics Study, Library Science, Information Sciences (research), Information Sciences (teaching), Museology and heritage management, Informatology; Course: Information Sciences (research); Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra informacijskih znanosti (magistar/magistra informacijskih znanosti)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2024-09-13 14:14:57