Extracting what, when, why, and how from Radiology Reports in Real World Data acquisition (2021)

Presented by: Rajesh Jacob - Senior Technical Product Manager at GE Healthcare and Hasham Ul Haq - Data Scientist at John Snow Labs
Pharma companies typically face issues with data that limit the quality of patient outcomes and profitability as a result of delayed clinical trials, limited channel access to quality data partners to deploy/validate innovative algorithms at scale or manage data from 3 rd parties to mitigate compliance risk. The EDISON platform from GE Healthcare is a customizable and
modular software platform that enables deployment of HIPAA compliant healthcare solutions which can aggregate, correlate, analyze, process, display and share data from connected medical devices, sensors and health information systems. EDISON NLP is one such service that aims to deliver insights from unstructured clinical data to drive the development of pharma-oriented solutions. GE Healthcare and John Snow Labs collaborated
to build an end-to-end solution to process radiology reports capable of extracting and linking all the concepts presented in the report. The concepts include Dates, Imaging Tests, Test Technique, Risk Factors, Body Parts, Measurements, and other general findings. We then link them all together to identify the type and technique of the procedure, the body part where it was performed, when it was performed, and any conclusions as a result of the
procedure. Also, since the reports contain tabular data, we built a system to extract tables from the unstructured text.