University of Twente Student Theses
Predicting Response of Lung Cancer Patients to Immunotherapy : Based on Routinely Determined Bloodwork and CT Scans
Uum, E.D. van (2025) Predicting Response of Lung Cancer Patients to Immunotherapy : Based on Routinely Determined Bloodwork and CT Scans.
|
PDF
3MB |
| Abstract: | Research aim This research investigates the predictive power of routinely collected blood values and CT scans in forecasting the response to immunotherapy in patients with stage IV non-small cell lung cancer. Method The study involved a retrospective analysis of patient data, including demographic information, clinical characteristics, blood values, and CT images. Various statistical and machine learning methods were applied, including Kaplan-Meier analysis, log-rank tests, multinomial regression, mixed-effects models, and random forest. Results Several blood biomarkers, particularly CRP, emerged as significant predictors of overall survival (OS) at various time points during treatment. In addition, immune cell ratios such as NLR, PLR, and LMR demonstrated notable prognostic value. Blood values obtained after the initiation of therapy showed a stronger association with OS compared to baseline values. Although the nnU-Net achieved the highest Dice scores (0.56) for automated tumor segmentation from CT scans, these scores were insufficiently high to reliably extract radiological features. Mixed-effects models (MEM) and random forest (RF) models that integrated blood values and clinical data demonstrated potential for more accurate prediction of immunotherapy response. However, the low feature importance scores in the RF models indicated that the response to immunotherapy is shaped by a complex interplay of factors rather than by a single dominant feature. Conclusion This study provides initial insights into the relationship between clinical parameters and treatment outcomes in immunotherapy for NSCLC. Although it did not achieve the goal of identifying patients who would benefit the most from treatment, it demonstrated potential for developing models that can potentially identify non-responders. |
| Item Type: | Essay (Master) |
| Faculty: | TNW: Science and Technology |
| Subject: | 44 medicine |
| Programme: | Technical Medicine MSc (60033) |
| Link to this item: | https://purl.utwente.nl/essays/105301 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page