We develop a machine learning-based medical "fact extractor" capable of extracting structured, actionable medical facts from free-text documents for downstream use in applications designed to reduce chart search and documentation burden, as well as a fact schema capable of capturing the bulk of informational in clinical notes. We demonstrate its feasibility on a public corpus of 100 discharge summaries, and motivate its use via a survey of internal medicine residents at our institution.

Describe the new knowledge and additional skills the participant will gain after attending your presentation.: -Gain a better understanding of the untapped potential of existing free-text medical notes for clinical and research use as well as efficiency improvements
-Understand the potential of modern machine-learning methods to quickly extract complete information from such documents with no pre-processing or external dictionaries
-Understand neural network architectures for structured information extraction


Jackson Steinkamp (Presenter)
Boston University School of Medicine

Wasif Bala, Boston University School of Medicine
Abhinav Sharma, Boston University School of Medicine
Jake Kantrowitz, Boston University School of Medicine

Presentation Materials: