Democritus University of Thrace (DUTH)
Αξιολόγηση απόκρισης της εξωτερικής ακτινοθεραπείας προστάτη με την χρήση προηγμένων απεικονιστικών τεχνικών
Abstract
dc:descriptionProstate cancer is one of the most common malignancies in men, with radiotherapy representing a cornerstone therapeutic option. The assessment of treatment response has traditionally relied on clinical and biochemical markers, such as Prostate Specific Antigen, which, however, show limited reliability, particularly in early stages or following focal therapies. In this context, the present doctoral dissertation investigates the contribution of advanced quantitative Magnetic Resonance Imaging parameters, radiomic features, and machine learning algorithms to the early and objective evaluation of Prostate cancer response to radiotherapy. MRI data were analyzed at three timepoints (before, during, and after the completion of radiotherapy), examining prostate volume, T2 and T2 relaxation times, Apparent Diffusion Coefficient values, as well as radiomic features extracted from T2-weighted and diffusion-weighted imaging. Additionally, the association of these parameters with delivered radiation dose, radiotherapy regimen, and administration of hormone therapy was explored. Machine learning models such as Random Forest, k-Neural Network,Naive Bayes, Logistic Regression and Neural Network, were also developed and evaluated for the discrimination of benign and malignant lesions using imaging, radiomic, and clinical data. The results demonstrated that prostate volume and T2 relaxation exhibited statistically significant temporal changes and correlation with radiation dose, reflecting early radiation-induced alterations. In contrast, Apparent Diffusion Coefficient and T2 values remained relatively stable at the whole-prostate level. Radiomic features, particularly from Diffusion-Weighted Imaging, revealed time-dependent changes in tissue heterogeneity and microstructure. Machine learning models achieved optimal performance when combining radiomic features with clinical variables, such as Prostate Specific Antigen, showing enhanced discriminative and diagnostic capability in lesion characterization. Neural networks and Support Vector Machine models achieved the highest Area Under Curve values, highlighting their accuracy in lesion classification and prediction.In conclusion, this study supports that the combination of quantitative Magnetic Resonance Imaging, radiomic analysis, and machine learning can substantially contribute to the evaluation of radiation-induced effects in the prostate, improving the understanding of tissue alterations and providing reliable tools for clinical decision-making, paving the way for tailored and personalized monitoring and therapeutic strategies.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Bekou, Eleni
- Μπέκου, Ελένη
Subjects
dc:subject × 22- Μαγνητική τομογραφία
- Ακτινοθεραπεία
- Καρκίνος του προστάτη
- Απόκριση
- Απεικονιστικοί Βιοδείκτες
- Μηχανική μάθηση
- Magnetic resonance imaging
- Radiotherapy
- Prostate Cancer (CP)
- Response
- Imaging Biomarkers
- Machine learning
- Επιστήμες Υγείας
- Ιατρική και Επιστήμες Υγείας
- Επαγγέλματα υγείας
- Κλινική Ιατρική
- Ραδιολογία, Πυρηνική ιατρική και Απεικονιστική
- Health Sciences
- Medical and Health Sciences
- Health Professions
- Clinical Medicine
- Radiology, Nuclear Medicine and Imaging
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/61140