Abstract
To administer the Norwegian Cervical Cancer Screening program in accordance with national guidelines, the Cancer Registry of Norway (CRN) utilizes routinely collected information on diagnostic exam types and corresponding results for each individual. The availability of this data presents an opportunity to leverage data-driven technology for more personalized strategies in population-level cervical cancer prevention. However, existing machine learning methods for cervical cancer risk assessment rely on more extensive data material, including lifestyle and risk factor information, which is unattainable for the entire screening population. To address this gap, this thesis focuses on developing and validating machine learning methods derived from only the CRN screening data. By introducing methodology designed specifically for the Norwegian cervical cancer screening data, this thesis presents novel approaches to predicting the time-varying risk of cervical cancer development based on individual exam histories. Through numerical experiments, the thesis expands the understanding of the potential applications and limitations of these algorithms in personalized cervical cancer risk estimation.
List of papers
Paper I. Langberg, G.S.R.E., Stapnes, M., Nygård, J.F. et al. “Matrix factorization for the reconstruction of cervical cancer screening histories and prediction of future screening results”. Published in BMC Bioinformatics. (2022), DOI: 10.1186/s12859-022-04949-8. The article is included in the thesis. Also available at: https://doi.org/10.1186/s12859-022-04949-8 |
Paper II. Langberg, G.S.R.E., Nygård, J.F., Gogineni, V.C. et al. “Towards a data-driven system for personalized cervical cancer risk stratification”. Published in Scientific Reports. (2022), DOI: 10.1038/s41598-022-16361-6. The article is included in the thesis. Also available at: https://doi.org/10.1038/s41598-022-16361-6 |
Paper III. Langberg, GSRE et al. “A weighted margin loss for treating imbalanced, overlapping and noisy data in cervical cancer risk prediction”. Submitted to International Journal of Medical Informatics. (2023). To be published. The paper is not available in DUO awaiting publishing. |