Institutional Repository of Vilnius University
Plaučių auglio aptikimas neuroniniais tinklais /
Abstract
dc:descriptionIn the bachelor’s thesis, a problem in the field of computer vision, classification of lung nodule candidates in cross-sectional images of computed tomography scans, was analyzed. It is very relevant in today’s society and new effective ways are constantly being sought to solve it. Advancements of deep learning algorithms in the field of medical image analysis are increasingly providing hope that the use of artificial intelligence will help in the early diagnosis of lung cancer and will enable its treatment to begin at the optimal time. In order to successfully identify malignant nodules it is necessary to be able to differentiate between nodules and non-nodules. In the bachelor’s thesis, convolutional neural networks were used for the finding and classification of suspicious regions in cross-sectional images of computed tomography scans. Candidate search was implemented using created segmentation models. Based on the found locations of suspicious lung regions, image fragments with regions of interest were generated with the hope of achieving high results in classification. Although the best classification result was obtained when the inputs were not generated based on predictions of a segmentation model, overall, it can be seen that the performance of a segmentation model positively correlates with classification results. The variant that showed the best metric values, which was based on generated fragments, was not far behind from networks that worked on images of the original size and performed better in terms of parameters.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Šinkūnas, Kristupas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192846293&prefLang=en_US