Identifying morphological features in benign breast disease biopsies to predict breast cancer risk

Simple item page
dc.contributor.advisorMiller, L.E.C.
dc.contributor.advisorLaak, van der, J.A.W.M.
dc.contributor.advisorAswolinskiy, W.
dc.contributor.authorKoelewijn, Rianne-Margot
dc.date.issued2023-11-30
dc.description.abstractIn this project we aimed to identify morphological features in benign breast disease biopsies that have the potential to improve the prediction of the risk estimate for developing breast cancer (BC). Terminal duct lobular units (TDLU) are the structures in breast tissue where the precursors of breast cancer arise. Existing models currently do not use features from biopsies to predict the risk estimate. We expanded and improved the post-processing of a pre-existing pipeline that is able to segment breast tissue biopsies and extract features based on this segmentation. We expanded the pipeline using HoVer-Net and a new calcification segmentation network to extract additional features based on cell count and calcifications. The new pipeline can extract a variety of features from whole-slide images (WSI) on a WSI level and a TDLU level. The exploratory statistical analysis of the extracted features showed that on the WSI level there are no significantly relevant features. On the TDLU level the feature average epithelial thickness was significant (p<0.05). We also found a small correlation with developing BC on the individual level. In the logistic regression model this feature was also significant. Therefore, the average epithelial thickness has potential to improve the prediction of the risk estimate of developing BC as it can be an indicator for proliferation. The features extracted from the calcification network segmentations were not significant. However, the network is not performing optimally yet and thus the current results need to be taken with caution and further research is needed.
dc.identifier.urihttps://theses.ubn.ru.nl/handle/123456789/16412
dc.language.isoen
dc.thesis.facultyFaculteit der Sociale Wetenschappen
dc.thesis.specialisationspecialisations::Faculteit der Sociale Wetenschappen::Artificial Intelligence::Master Artificial Intelligence
dc.thesis.studyprogrammestudyprogrammes::Faculteit der Sociale Wetenschappen::Artificial Intelligence
dc.thesis.typeMaster
dc.titleIdentifying morphological features in benign breast disease biopsies to predict breast cancer risk

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Koelewijn, R. s-4796268-MSc-MKI92-Thesis-2023.pdf
Size:
6.32 MB
Format:
Adobe Portable Document Format
Download

Collections

Faculteit der Sociale Wetenschappen