Oncora shares machine learning research at AAPM 2017 in Denver

July 31, 2017

As a company, we seek opportunities to learn from leading scientists and physicians within the field of radiation oncology. Attending conferences is one of our favorite ways to engage with the community and explore new ideas. Oncora is excited to be attending AAPM’s 59th Annual Meeting & Exhibition held in Denver, Colorado on July 30th, 2017. AAPM, the American Association of Physicists in Medicine, is a scientific and professional organization focused on radiation within medicine. Every year, the organization holds a conference for medical physicists and other professionals interested in coming together to discuss innovation and learn about progress in the field.

At this year’s event, Oncora will be presenting our recent research on predictive modeling and clinical outcomes, a collaborative effort between Oncora, the University of Pennsylvania, and the University of Maryland.

The goal of this study was to help guide data extraction efforts by determining which types of clinical variables (e.g., patient demographics, tumor characteristics, radiation treatment information, etc.) contribute the most to a machine learning model’s ability to predict a number of oncology treatment outcomes. In the process, we display the full power of our machine learning infrastructure, using a dataset containing over 16,000 oncology treatment courses with over 230 variables extracted for each course to train a predictive model for each of 76 different outcomes. Based on our analysis, we concluded that tumor characteristics (location, size, and stage) were the most critical to predict most outcomes. It is also worth noting that the model’s trained in this study performed very well: the median test set AUC (across all 76 predictive models) was 0.91.

We look forward to attending the AAPM conference and making connections with leading physicians and medical physicists from around the world.