Good ultrasound images of the heart can be especially elusive; the heart is a perpetually moving target, tucked behind the thick walls of the chest. San Francisco-based startup Caption Health is aiming to make heart sonograms more accurate and accessible, using cutting-edge artificial intelligence.

Caption’s software aids the user throughout the ultrasound process. It offers guidance about the proper positioning of the ultrasound probe, and automatically snaps the right image at just the right moment.

Austin Walters, managing partner of SpringTide Capital Management, a health tech VC firm focused on AI, explained that Caption’s kind of functionality is “similar to automatically snapping a photo of a check via mobile deposit. AI can help acquire the right image, a non-trivial task in ultrasound.” After capture, the software then helps identify potential pathologies in the image, flagging them for closer analysis.

Caption, formerly known as Bay Labs, was founded in 2013 by President Charles Cadieu and CTO Kilian Koepsell. Andy Page, former 23andMe President, subsequently joined as CEO, having been introduced by one of Caption’s investors, billionaire venture capitalist Vinod Khosla. Page told Stat News he thinks Caption’s software is a “a co-pilot” for practitioners, assisting doctors who might struggle to get a particular scan, or who are less experienced ultrasound users.

Caption’s field is promising and growing fast. Other firms using AI to optimize ultrasound include Butterfly Network, Ultromics and DiA. 

Walters said the most significant aspect of this tech is how it “can democratize the use of ultrasound, given the decreasing price and footprint of the hardware.”

The AI-based solutions also lower the skill threshold required, broadening the pool of potential ultrasound users. Caregivers who might not be proficient users of traditional ultrasound systems could find these solutions easy to use well. 

  • Caption recently completed a study involving eight nurses, with no previous experience in cardiac ultrasound, performing four kinds of scans on 240 patients. The results were promising: an independent panel of five cardiologists found that for the left ventricle, 98.8% of Caption’s images were of sufficient quality. The Caption-enabled images of the right ventricle, which is typically harder to see, rendered 92.5% useable images.
  • Caption has raised $18 million from investors like Data Collective and Khosla Ventures.
  • Caption’s founders previously were part of the team that started IQ Engines, a company that uses deep learning to identify images. IQ Engines was acquired by Yahoo in 2013.
  • Caption doesn’t make its own ultrasound hardware, so the company has partnered with an ultrasound manufacturer called Terason.
  • Caption has received breakthrough device status from the Food and Drug Administration, a good sign for its regulatory future.