Predviđanje raka dojke temeljem značajki dobivenih iz probirnih mamograma

master's thesis

• Title (english): Prediction of Breast Cancer Using Features Extracted From Screening Mamograms
• Author: Dora Marković
• Mentor: Tomislav Petković (mentor)
• Committee member: Tomislav Petković (predsjednik povjerenstva)
• Committee member: Zvonko Kostanjčar (član povjerenstva)
• Committee member: Marko Subašić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-18, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Mamografija je rendgensko snimanje dojke koje se koristi u probiru i dijagnostici raka dojke. Zbog velikog broja pretraga, interpretacija podataka temeljena na strojnom učenju uvelike olakšava rad i poboljšava rezultate. U ovom radu proučavamo dijagnosticiranje raka dojke uz pomoć dva modela, SVM-a i Bayesovih mreža. Značajke su dobivene iz tri javno dostupne baze, CBIS-DDSM, mini-MIAS te Mammographic Mass Data Set. Značajke su promijenjene iz kategoričkih u numeričke, ako je to bilo potrebno. Za oba model je provedena k-unakrsna provjera te su dobiveni optimalni hiperparametri modela. Uočava se da u većini slučajeva SVM daje malo bolje rezultate od Bayesovih mreža.

Abstract (english)

Mammography is the X-ray of the breast that is used in screening and diagnosis of breast cancer. Due to the large number of searches, interpretation of data based on machine learning greatly facilitates screening and improves the results. In this thesis we investigated breast cancer diagnosis with the help of two models, SVM and the Bayesian network. Features are derived from three publicly accessible databases, CBIS-DDSM, mini-MIAS, and Mammographic Mass Data Set. Features were changed from categorical to numeric, if necessary. For both models, a k-fold cross-validation was performed and the optimal model hyperparameters were obtained. In most cases SVM gives better results than the Bayesian network, but the difference in the obtains scores is small.

• Keywords: mamogram; biopsija; probirni mamogram; SVM; Bayesove mreže; stabla odluke
• Keywords (english): mamogram; biopsy; screening mamography; SVM; Bayesian networks; decision crees
• Language: croatian
• URN:NBN: urn:nbn:hr:168:813386
• Study programme: Title: Information and Communication Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije (magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-01-07 20:04:33