Federated imaging network startup Avandra has raised $17.5 million in series A funding to give researchers access to larger volumes of imaging data to train AI models or develop novel therapies.
The round was co-led by Aegis Ventures and Spring Rock Health. Two major health systems, Memorial Hermann and Northwell Health, also backed the company in the round. Many other investors were involved, such as Greater Ventures, Compass Ventures and independent investor Scott Gaines.
Murray Brozinsky of Aegis Ventures and Kirsten Morbeck of SpringRock Ventures will join Avandra’s board as part of the deal.
Avandra is building out a federated imaging network complete with thousands of CT scans, MRIs, PET scans and echocardiograms. While the field of radiology has quickly advanced to use AI, other fields like neurology, cardiology and oncology have lagged because of the lack of large sets of imaging data.
The imaging network is poised to help researchers, drug developers and providers understand more about complex conditions. It can also be used to train the “last mile” of AI, Avandra’s co-founder Ryan Tarzy said in an interview with Fierce Healthcare.
“I became aware of the challenges of imaging, specifically in the role that the lack of wide availability of research-quality imaging datasets has on our ability to move healthcare forward and to bring the future forward,” Tarzy explained.
The hurdles in obtaining large amounts of imaging data are steep, Tarzy said. For one, storing imaging data can be costly because they quickly take up large volumes of space in the cloud. Moreover, it can be difficult for researchers to find and collect imaging data because they are dispersed across systems.
While it has become commonplace to use large data sets of payer data and electronic medical record data, imaging—outside of radiology—has not yet had an accessibility revolution.
Conversely, researchers need a wealth of imaging data to be able to identify patterns and make new discoveries. “The reason you need large quantities is increasingly these diagnoses, these AI models, are based upon really narrow criteria, where you need extremely highly-curated specific data,” Tarzy said.
He continued: “[Academic medical centers] can only do so much with the data within their own walls, right? They cannot build a data set that is diverse enough. They need more data to solve these, you know, rare use cases.”
With Avandra’s imaging network, imaging data stay securely at health institutions but are indexed and cataloged in its network. The company spent three years working in stealth before its public launch at the HLTH conference in October, mainly because it wanted to ensure it could keep patient data safe and ensure the technology could handle billions of large imaging files, executives said.
By the end of 2025, Tarzy said the company plans to have 200 million patient studies and 100 billion images, based upon the systems contracted and in the network right now.
The health systems can also make a small return when other researchers access their images, though Tarzy noted that the participating institutions join for the mission rather than the compensation.
With the nearly $18 million in funding, Avandra has been hiring out the team and continuing to accelerate the technology it uses to index and de-identify data.