CogStack speaking at NVIDIA GTC – Organizing, Structuring, and Utilizing the Vast Amount of Clinical Information Available in Electronic Health Records


Learn how CogStack-MedCAT, an open-source natural language processing platform, is used to unlock knowledge available in electronic health records (EHRs) and provide the groundwork for a multitude of downstream medical applications. In-hospital patient data are rarely structured and easily accessible. To facilitate R&D we present CogStack, a data-ingestion and harmonization pipeline used to organize, link, and index multi-source hospital data. As ~80% of EHR data are noisy, ambiguous free text, we present MedCAT, a named-entity recognition and linking tool able to extract and contextualize biomedical concepts from free text. Finally, we’ll present research and applications enabled by CogStack-MedCAT — (1) Multi-hospital COVID-19 projects, and (2) initial results of two novel health-care applications: (a) a self-supervised approach based on language models for temporal modeling of patients and diseases, and (b) a prototype summarization model generating phenotypically accurate, coherent clinical text.

Read more on NVIDIA GTC.

Photo credit: NVIDIA GTC.

Leave a comment
© 2020 CogStack - PhiDataLab | Made by Suara