A Big Win for Big Data
Big data analytics and machine learning hold great potential to transform radiation oncology given the enormous amount of data generated through patient care. ROI Researcher Todd McNutt, PhD, at the Johns Hopkins School of Medicine and Kimmel Cancer Center, saw the tremendous possibility early on and has been building the Oncospace system to compile and analyze data from the treatment planning system and patients’ clinical records for more than a decade. In 2016, the ROI awarded Dr. McNutt a grant to use the data from head and neck cancer patients in Oncospace to build machine learning models that make personalized, evidence-based predictions of treatment toxicities related to weight loss and to begin developing a decision support tool that will assist with treatment planning and clinical interventions based on these predictions.
Dr. McNutt’s ROI-funded research is beginning to yield results, and his team, in collaboration with Harry Quon, MD, recently published two manuscripts demonstrating how the spatial distribution of the radiation dose influences toxicities. The article, “Machine Learning Methods Uncover Radio-Morphologic Dose Patterns in Salivary Glands that Predict Xerostomia in Head and Neck Cancer Patients,” is available in Advances in Radiation Oncology as an accepted manuscript. The second article, “Radio-morphology: Parametric Shape-Based Features in Radiotherapy,” was recently accepted by Medical Physics. The lead authors of the articles are Wei Jiang, PhD, and Pranav Lakshminarayanan, MS, respectively.
These analyses of empirical data from a large number of patients are producing insights on dose patterns that were previously extremely time-consuming or even impossible. The team’s voxel-based dose analysis was able to predict xerostomia in head and neck cancer patients and identify important parotid dose subvolumes associated with predicted risk of severe xerostomia. Large databases such as Oncospace are providing a wealth of information for radiation oncologists, physicists, and biologists to better understand the spatial patterns of influence of radiation dose. Eventually, this type of research will help practitioners improve radiation treatment planning, reduce or prevent toxicities, improve post-treatment recovery, enhance the overall quality of care, and perhaps even improve local control conferred by radiation.