Dyania Health is using artificial intelligence to automate manual patient chart review, boosting providers' ability to comb through medical records and speeding up medical research.
The female-founded tech company developed Synapsis AI, a proprietary LLM-based end-to-end platform that tackles an efficient and time-consuming process. The medically specialized AI system automates manual chart review and abstraction to pre-screen patients for clinical trial enrollment and observational studies.
Dyania, founded in 2020, banked $10 million in series A funding round led by HealthX Ventures. Tech Square Ventures and Cleveland Clinic Ventures also backed the round, along with significant participation from existing investors, the company announced Thursday. The startup picked up a $5.3 million seed round in 2022.
The company will use the fresh capital to expand its AI capabilities and build partnerships with more leading organizations to drive improved outcomes in clinical research and patient care, executives said. Dyania also plans to scale up its engineering team.
Dyania’s platform is initially being used at large healthcare systems to enable enterprise-wide clinical trial pre-screening, run automated chart review for observational studies, and automate complex registry reporting to reduce manual workloads and ensure more accurate and timely patient care.
“Effectively, what the technology does is automate what is otherwise a very human effort to read and draw conclusions from medical records. What would take a human an hour and a half to two hours, we do in half a second," Eirini Schlosser, CEO and founder of Dyania Health, told Fierce Healthcare on the sidelines of the HLTH 2024 conference in Las Vegas this week.
"The use cases for that are quite vast. It’s kind of like a faucet on a sink, you can water a plant and boil of an egg. Once it’s set up within the healthcare system's firewall, we're not doing any data mining. No data leaves the firewall. The AI is running in the background to read and ask questions about the patient population for either clinical trial qualification for individual patients or to run observational studies and automating registries,” she said.
Synapsis AI can read and deduce clinical conclusions from across the vast database of electronic medical records within a health system's firewall. What takes a human 30 minutes takes Synapsis AI less than five seconds with greater than 94% accuracy compared with a human medical professional and more than 15,000 times faster, according to company executives.
Clinical records contain vast amounts of unstructured medical data such as physician notes, pathology reports and imaging notes. It's estimated that 80% of EMR data is unstructured and it's difficult for clinicians to connect the dots fast enough to benefits patients who could be eligible for clinical trials. It can take healthcare professionals hours to read through all this information, then multiply that by millions of patients treated at academic medical centers.
It would a medical professional 12.5 weeks to read 1,000 EMRs, while Dyania Health's technology can accomplish that in 8.3 minutes.
Dyania's Synapsis AI platform uses its proprietary LLM along with advanced data engineering to glean insights from both large volumes of unstructured medical records and structured data such as codified information or numerical lab results. Using the technology, physicians can interrogate patient data with complex clinical questions.
The Synapsis AI platform also is trained to provide a justification pinpointing the exact part of the medical record that led the AI to give the result, Schlosser said.
And, as Schlosser noted, Dyania Health's technology works within the health system’s firewall, with AI processing done locally, so data never leaves the healthcare system’s computing environment.
Schlosser, a serial entrepreneur who has worked in natural language processing for the past decade, built a team of full-physicians working hand-in-hand with applied AI researchers. The team combines mathematical acumen with deep medical expertise to build large language models for medical grade performance in clinical research, she said.
Dyania Health's team spent the COVID-19 pandemic working together to annotate mass data sets and utilize those annotated EMRs to train their LLM and build their proprietary Synapsis AI platform.
Launching the company at the beginning of the COVID-19 pandemic was actually advantageous, Schlosser noted. The company picked up pre-seed funding in early 2020 to focus on building the technology.
“We combined a few different components into what was a really unique offering, effectively having full-time physicians work hand in hand with AI researchers to apply mathematics that had come out of big tech and had direct experience training large language models before the world knew about ChatGPT. We spent the pandemic training and annotating datasets, and had the luxury to do that because of the pandemic," she said.
At the end of 2022, when ChatGPT launched, Dyania Health was already starting to work with Cleveland Clinic, she noted.
“We went elephant hunting out of the gate, which was the right approach in hindsight, because we wanted to work with institutions that had the computing resources to be able to support AI deployments on premise," Schlosser said.
Dyania Health has formed strategic partnerships and commercial relationships with several healthcare systems and life sciences companies. The company focuses on working with academic medical centers that have, along with computational infrastructure, significant clinical research operations and high volumes of diverse patient populations.
During its development phase, Dyania Health took a first-mover decision to acquire GPU for computing resources and then focused on continuous innovation and training its models, which now consistently outperform human experts in medical chart review, according to Schlosser.
Researchers at Mass General Brigham found that artificial intelligence software like ChatGPT could help speed up the screening process to find patients eligible for clinical trials, as Fierce Medtech reported.
Schlosser grew up in a family of physicians and was initially a pre-med student in college. But, she switched out of medicine and spent the last decade leading multidisciplinary teams to focus exclusively on building technology platforms with various applications of natural language processing. Through this work, she said she recognized the untapped value of data that was buried in free text, especially in healthcare.
“I have been running from healthcare all my life, but it was just too big of a problem to not try and tackle. We just want to be able to impact what is otherwise taking up an immense amount of time and prohibiting patients from getting the care they need,” she said. “When you're looking for a patient who can have a very thin window of time where they would qualify for a certain clinical trial, or during that window of time, they would have an optimal therapy that would actually be able to help them, if they're not found during that period of time, they just are lost. That was the original motivation for me to be able to automate what would otherwise just have left healthcare decades behind the rest of the world in terms of technology.”
Cleveland Clinic is now planning an enterprise-wide roll-out of Synapsis AI after a year of working with Dyania Health in the cancer and cardiovascular disease areas.
“Artificial intelligence has the potential to enhance patient care and increase the speed of medical research. Our efforts with Dyania Health combine deep computational and medical expertise,” said Rohit Chandra, Ph.D., Cleveland Clinic’s Chief Digital Officer said in a statement.
Cleveland Clinic’s Heart, Vascular and Thoracic Institute and Cancer Institute is seeing early success with the technology in streamlining and accelerating the clinical trial recruitment process, Chandra noted.
"The system’s medically trained large language model can screen electronic medical records to identify patients that meet the eligibility criteria for clinical trials. Removing these bottlenecks will give patients easy access to clinical trials, as well as accelerate scientific progress," Chandra said.