Olejnik, Jeremi (2025) The Role of PET in Tumor and Lymph Node Segmentation for Advanced NSCLC Patients Using Deep Learning. Master's Thesis / Essay, Biomedical Engineering.
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
mBME2025OlejnikJ.pdf Restricted to Registered users only until 1 January 2026. Download (10MB) |
|
Text
Embargo en toestemming.pdf Restricted to Registered users only Download (155kB) |
Abstract
This study investigates whether incorporating PET data into a CT_only deep learning model improves the accuracy of automated segmentation of combined primary tumors and pathological lymph nodes in advanced NSCLC. A cohort of 689 advanced NSCLC patients was collected and preprocessed to ensure image consistency. Mask-guided rigid registration was performed between planning and low-dose CT scans using the ITK-Elastix framework, followed by co-registration of PET data. Subsequently, two deep learning models were developed: one trained on CT_only inputs and the other on fused PET-CT volumes. The models were optimized, and finally trained and evaluated on an independent test set. Finally, statistical analyses were conducted to determine the correlations between segmentation performance and clinical or imaging factors. The developed ITK-Elastix model outperformed a clinically approved registration software across all evaluated metrics. Segmentation results demonstrated that the inclusion of PET data substantially improved segmentation accuracy. CT_only model achieved a mean DSC of 0.52 ± 0.03 in cross-validation and 0.34 ± 0.05 on the test set. In comparison, PET-CT variant achieved higher scores of 0.65 ± 0.01 and 0.64 ± 0.01, respectively. Finally, statistical analysis revealed significant associations between PET-CT segmentation performance and various clinical and imaging characteristics.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Sijtsema, N.M. and Ooijen, P.M.A. van |
| Degree programme: | Biomedical Engineering |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 28 Jul 2025 08:20 |
| Last Modified: | 28 Jul 2025 08:20 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/36532 |
Actions (login required)
 |
View Item |