Artificial Intelligence, Diagnostics

Rad AI, developer of workflow automation tool for radiology, raises $4M in seed round

Most AI-powered startups in radiology tackle an image problem, but Rad AI have been focused on a workflow problem by automatically generating the impressions section of a radiologist’s report, which is a summary of the report’s other sections.

machine learning AI

Most AI-powered startups in radiology tackle an image problem: they strive to scan millions of patient images to detect the earliest signs of disease.

They include Arterys and Aidoc, which raised $27 million in April 2019. Tech company Nvidia, meanwhile, has been working to ensure radiologists have access to AI tools and a role in developing them.

But Dr. Jeffrey Chang and the team at Rad AI have been focused on a workflow problem: the 100 or more reports that radiologists dictate every day.

In late November, the Berkeley, California startup’s efforts attracted $4 million in seed funding led by Gradient Ventures, Google’s AI-focused venture fund. Other seed-round investors include UP2398, Precursor Ventures, GMO Venture Partners, Array Ventures, Hike Ventures, Fifty Years VC and various angels.

The funding will help Rad AI hire engineers and attract more customers for its product, also called Rad AI. The product automatically generates the impressions section of a radiologist’s report, which is a summary of the report’s other sections and considered the most important piece, said Chang, a radiologist and co-founder of Rad AI. Founded in 2018, the company employs 13 people.  He anticipates the staff will grow to 30 by the end of 2020

The automation saves radiologists about an hour a day, Chang said, adding that efficiency is increasingly vital as a growing workload falls on a static pool of radiologists.

“You have to figure out what the problem is first before you try to build the solution,” Chang said in a phone interview. “For us, this seemed to make the biggest difference for radiologists right now, given what is possible in machine learning.”

Designed to fit in with existing workflows, Rad AI’s product fills out the impressions section using each radiologist’s preferred wording for how they describe what they see. The product “learns” by scanning millions of existing reports submitted by its radiologist customers, and it is compatible with the voice-recognition tools used by most radiologists: PowerScribe 360, PowerScribe One and Fluency for Imaging.

The company’s existing customers include Greensboro Radiology in North Carolina, Medford Radiology in Massachusetts, Einstein Healthcare Network in Pennsylvania and Bay Imaging Consultants in Northern California, one of the largest private radiology groups in the U.S. [Chang is an ER radiologist at Greensboro Radiology.]

The product currently handles X-rays and CT scans, but is expected to cover ultrasounds and MRIs within the next two to four months, Chang said. Future products could automate other portions of the radiologist report and integrate the actual imaging. Other sections of the report include the indications, or the reason for the report, and the findings, which is a list of the radiologist’s observations of what was scanned.

The potential evolution of Rad AI’s product is one of the attractions for investors.

“The team at Rad AI is uniquely suited to apply innovative technology to this field, with strong radiology and AI experts and firsthand knowledge of this market,” Zachary Bratun-Glennon, a partner at Gradient Ventures, said in a statement. “It’s exciting to see the quantitative benefits and positive feedback from their radiology customers, and we’re looking forward to the impact of their future products.”

 

 

Shares0
Shares0