Funding outcome

Brachytherapy (placement of radiation sources in or near the tumor inside the body) is critical to curing locally advanced cervical cancer, with acceptable side effects. Clinical results and treatment experience have led to an international recommendation that higher radiation dose be delivered via brachytherapy. Safe use of such doses requires better treatment accuracy, especially in how sources are placed using special applicators and how organs are delineated for planning treatment. As applicator insertion can significantly change the anatomy, physicians often attempt to predict this change based on magnetic resonance images (MRI) taken without applicators, using best judgment, to choose an applicator in a way that allows the radiation source to be deployed to a configuration that delivers higher dose where it is needed. Better predictions can be made possible with machine learning, an artificial intelligence technique that identifies patterns within data, learns from them, and then makes predictions (predictive modeling). This project will explore the potential of machine learning in brachytherapy to predict applicator-related anatomic changes and to automatically identify organs on MRI. Based on this prediction, customizable 3-dimensional printed applicators will be designed for individual patients after accounting for anatomy and tumor geometry. Predictive modeling offers promise to improve processes, by personalizing and automating key steps, on the path towards more consistent high-quality cervical brachytherapy. Such improvements can have major implications on patient outcomes, likely reducing morbidity and mortality among cervical cancer survivors.