Big data and interpretable models for outcome prediction in radiation oncology

This thesis focuses on healthcare (Big) Data and the value of interpretable models for outcome prediction in radiation oncology and highlights the benefits of experts’ involvement in the model development process. For ease of reading, this thesis is partitioned into theoretical and practical sections. The theoretical section discusses healthcare (Big) Data, particularly in radiation therapy. This section talks about the characteristics of (Big) Data, different sources within the confines of radiation oncology, solutions provided for healthcare challenges, with examples where Big Data has improved operational efficiency for clinical excellence, domain applications, barriers, and the future of radiation oncology Big Data. The analytical section discusses the development and validation of three groups of models (Regression, Decision tree, and Bayesian network) to predict various patient outcomes in radiation oncology. This section focuses on models which can be understood, challenged, interpreted, and used by healthcare providers for optimal decision-making.