diplomsko delo
Matija Bažec (Author), Žiga Emeršič (Mentor), Tim Oblak (Co-mentor)

Abstract

Rak dojk je velika zdravstvena skrb. Pojav globokega učenja uvaja možnosti za pomoč medicinskemu osebju v boju proti bolezni. V tem delu smo uporabili metode globokega učenja za napovedovanje prisotnosti raka dojke pri bolnicah s tumorskimi lezijami. Razvili smo cevovod, ki vključuje model segmentacije in klasifikacije. Prvi služi za določitev lokacije lezije, drugi pa za ugotavljanje, ali je lezija benigna ali maligna. V sklopu diplomske naloge smo se skušali približati zmogljivosti vodilnih metod na področju in svoj pristop oceniti na lastni podatkovni zbirki mamografskih slik. Kljub dobrim preliminarnim rezultatom klasifikacijskega modela pa na testih celotnega cevovoda nismo dosegli želenih rezultatov. Razlog za to je segmentacijski model, ki mu na vhodni sliki ni uspelo prepoznati večjega števila potencialnih lezij.

Keywords

rak dojk;tumorji;segmentacija;klasifikacija;konvolucijske nevronske mreže;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [M. Bažec]
UDC: 004.85:618.19-006(043.2)
COBISS: 212440323 Link will open in a new window
Views: 92
Downloads: 207
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Prediction of cancer on mammographic images
Secondary abstract: Breast cancer is a major medical concern for people everywhere. The advent of deep learning introduces options to assist medical personnel in combating the disease. In this work we used deep learning methods to predict the presence of breast cancer on patients with tumorous lesions. We develop a pipeline that includes a segmentation and classification model. The first determines the location of the lesion and the second determines weather the lesion is benign or malign. Our goal was to reach the performance of contemporary models in the field and test our approach on a custom dataset of mammographic images. Despite initial success with our classification model, the evaluation of the final pipeline did not achieve the desired results. The reason for this is the segmentation model, which failed to detect several potential lesions in the input image.
Secondary keywords: breast cancer;segmentation;classification;deep learning;convolutional neural networks;computer and information science;diploma;Rak dojke;Globoko učenje (strojno učenje);Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000468
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 1 spletni vir (1 datoteka PDF (36 str.))
ID: 25392602