Abstract

This thesis develops and uses machine learning tools to predict the prognosis of endometrial cancer. With endometrial cancer being the most common gynaecologic malignancy in low- and middle-income countries, accurate prognostic models are essential for improving patient outcomes. This research aims to develop and validate machine learning-based models to predict CSS (cancer-specific survival) and DFS (disease-free survival) at 3- and 5-year intervals, utilizing patient demographics, preoperative imaging, histopathological characteristics, and treatment data.The study employed multiple machine learning algorithms, including logistic regression, support vector machines, XGBoost, and random forest, to create three predictive models: a preoperative model, a management-based model, and a postoperative model. The XGBoost model demonstrated superior performance across all evaluation metrics, including precision, recall, and F1 score. Compared to the conventional FIGO (Fédération Internationale de Gynécologie et d’Obstétrique) staging system, the machine learning models provided enhanced predictive accuracy and enabled individualized prognostication.A key contribution of this study is the identification of crucial prognostic factors, such as tumour grade, LVSI, and lymph node involvement. In particular, this study also highlights the importance of dynamic variables across different stages of patient management and underscores the value of k-fold cross-validation and external validation for model robustness using a multicentric dataset across 5 European cancer centres.Despite the promising results, some challenges related to model interpretability, and clinical integration persist. Future research should focus on expanding the diversity and size of datasets, integrating real-time patient data from electronic health records. Prospective clinical trials are also necessary to evaluate the real-world utility of these models and their impact on clinical decision-making.By advancing machine learning methodologies and addressing ethical considerations in data governance, this study lays a foundation for integrating artificial intelligence (AI) into clinical practice. The findings demonstrate the potential of machine learning to transform endometrial cancer prognosis, facilitating personalized treatment strategies and contributing to the broader field of precision medicine.

Awarding Institution(s)

University of Plymouth

Supervisor

Shang-Ming Zhou, Jill Shawe, Hossein Ahmadi

Document Type

Thesis

Publication Date

2025

Embargo Period

2025-12-11

Deposit Date

December 2025

Download

Share

COinS