Radiomics in Upper Tract Urothelial Carcinoma: Integrating Machine Learning, CTU Imaging and Clinicopathological Variables for Improved Diagnosis, Prognosis, and Treatment

Abstract

Upper Urinary Tract Urothelial Carcinoma (UTUC) represents a significant challenge in urological oncology, with complex diagnostic and treatment pathways that significantly impact patient outcomes. Accurately predicting tumour characteristics, survival, and recurrence remains crucial for optimal patient management. This thesis explores the intersection of advanced imaging analysis, machine learning, and clinical practice to enhance our understanding and management of UTUC. Throughout this PhD journey, my supervisors allowed me to combine clinical expertise with cutting-edge technological approaches, specifically focusing on radiomics and machine learning applications in medical imaging. Each chapter of this thesis presents novel methodological approaches applied to highquality clinical imaging data to better understand and predict UTUC outcomes in three specific contexts. Firstly, I explored the application of radiomics-based machine learning for predicting UTUC grade and stage. This work demonstrated how advanced image analysis techniques may potentially enhance the accuracy of preoperative diagnosis, addressing a crucial clinical need for more precise, noninvasive diagnostic tools. Secondly, I investigated the prediction of survival in UTUC patients using radiomics features extracted from the computed tomography urogram. This section integrated clinical data with radiomics analysis to develop comprehensive prognostic models, offering new insights into patient risk stratification and outcome prediction. Thirdly, I presented an in-depth analysis of UTUC recurrence prediction using radiomics approaches, demonstrating how image-based features may contribute to identifying patients at higher risk of disease recurrence. Finally, the thesis concluded by exploring how its findings could improve clinical practice and suggesting the next research steps.