MIT Spin-Out Applies AI to Chart Reviews of Unstructured Data

Nov. 16, 2023
Layer Health is backed by $4 million in funding from GV (Google Ventures), General Catalyst, and Inception Health

Layer Health, a company spun out of MIT, says its first product, Distill, uses artificial intelligence to quickly perform any clinical, administrative, or research task that requires chart review from unstructured data. Use cases include registry submissions, quality measurement, curation of real-world evidence, clinical document improvement, and revenue cycle management. 

The startup is backed by $4 million in funding from GV (Google Ventures), General Catalyst, and Inception Health, and among its beta customers is Froedtert & the Medical College of Wisconsin health network, which is using Distill to support quality improvement efforts. The platform is being used by the organization’s nurse abstraction team during chart review, helping them more effectively find and submit data to clinical registries. 

Layer Health says that Distill integrates into existing products and workflows, ingesting clinical notes and analyzing them at scale. “Behind the scenes, Layer Health’s machine learning (ML) algorithms leverage the power of large language models (LLMs) to deliver accurate results without the need for labeled data, reducing development time from months to as little as a day. Distill also learns and adapts from customer interactions, creating highly efficient customer-specific models fine-tuned for specific use cases,” the company says. 

Another user the company cites is xCures, a health technology company that is using Distill to organize and structure health data for more precise cancer treatment recommendations and efficient clinical trial matching. Using Layer’s language models trained and validated on xCures' unique data, the company can more accurately extract intricate and nuanced details from patient medical records.

Layer Health’s CEO is David Sontag, Ph.D., an MIT professor with more than 100 published papers in AI and machine learning.