New AI-powered Truveta Language Model to help improve medical research

New AI-powered Truveta Language Model to help improve medical research

iStock/MarcoVDM

The Azure-based solution transforms electronic health records into normalised data points to facilitate research into drugs, diseases and more

Amber Hickman |


Healthcare data solution provider Truveta has introduced the Truveta Language Model (TLM), which is built on Microsoft Azure and transforms electronic health record data into data points that can be used for research into drugs, diseases or medical devices.

The artificial intelligence-powered TLM is trained to extract patient diagnoses, medications, lab results and other unstructured data from sources such as clinician’s notes and insurance claims, which are often filled with misspellings and differing jargon or abbreviated terms.

Truveta has combined pre-trained open large language models with additional training on a complete and representative clinical data set from 28 health system partners. This enables TLM to normalise medical data with more than 90 per cent accuracy on diagnoses, medications, lab results, clinical observations and more.

 The model is also trained without commercial bias to ensure that research is conducted solely on clinical outcomes as opposed to billing.

“Other industries are benefiting from the advancement of AI, but the private, fragmented and unstructured nature of healthcare data has made applying AI to patient data extremely challenging to this point,” said Jay Nanduri, chief technology officer at Truveta. “Accurate AI requires the most advanced technology matched with an incredible volume of data trained by the best expertise. By using clinical expert-led AI to unlock the power of rich healthcare data, researchers can now ask and answer complex medical questions of a real-time, fully transparent view of US health.”

The TLM is also able to identify concepts within clinician notes that are often overlooked by standard structured data sets, and map relationships between them.

“Initially, Covid-19 symptoms and details weren’t put in the structured data, but often captured in the clinician notes, making it hard to detect the virus’s patterns quickly,” said Michael Lucas, principal machine learning engineer at Truveta. “Clinician notes capture missed medications, diagnoses, surgeries, symptoms, adverse events, and so much more. By adopting state-of-the art AI, Truveta makes critical insights available for research to find cures faster.”

Subscribe to the Technology Record newsletter


  • ©2024 Tudor Rose. All Rights Reserved. Technology Record is published by Tudor Rose with the support and guidance of Microsoft.