fbpx

Caption Health Disrupts Medical Imaging with AI Ultrasound

August 30. 2020. 5 mins read

Almost every day we read about how artificial intelligence is outcompeting humans. The latest involves an algorithm that outmaneuvered a fighter pilot in a simulated dog fight. While no one (well, there’s probably someone) is ready to hand over the flight controls to a machine, we’re not too far from when AI algorithms will help co-pilot a jet. We already let machines pick stocks, push paperwork around, and even diagnose kidney disease. Healthcare is certainly one area where we see widespread adoption of machine learning and other AI learning systems, and possibly the most advanced use cases involve medical imaging. One of the hottest startups in that space is Caption Health, which has developed an AI ultrasound platform for scanning hearts that even a heartless MBA could operate.

AI and Medical Imaging

Advances in areas like computer vision have enabled a revolution in healthcare, where machines help capture, enhance, and interpret medical images. Dozens of startups are now developing medical imaging AI algorithms to do everything from measuring breast density to preparing surgeons for surgery. Probably the most well-known AI ultrasound company up until now is Butterfly Network, which developed a handheld ultrasound device that uses deep learning to guide users in its operation. While ultrasound technology can also be used to treat disease, like breaking up blood clots that cause strokes, the main application is still in medical imaging and analysis. 

A Medical Imaging Startup

Click for company

Founded in 2013, San Francisco-based Caption Health has raised more than $60 million in disclosed funding, with the lion’s share of dollars invested last month in a $53 million Series B. The roster of investors numbers nearly 20, with the most high-profile name on the list being VC firm Khosla Ventures. Another notable name is Edwards Lifesciences (EW), a $50 billion medical device and services company that specializes in cardiovascular health. The startup was founded the same year that Caption Health co-founders Charles Cadieu and Kilian Koepsell, along with their other partners, sold an image-recognition startup called IQ Engines to Yahoo for an undisclosed sum. The duo reportedly first met 15 years ago at the University of California Berkeley where they were both working in the field of artificial intelligence.

It’s been a pretty good year for Caption Health – even “the rona” has provided an opportunity to showcase its tech, which we’ll talk about a bit later. The company made the cut this year as one of the 150 most promising digital health startups in the world out of nearly 8,000 reviewed, according to the big brains at data research firm CB Insights. More importantly, it has received not one but three different approvals from the U.S. Food and Drug Administration (FDA) in 2020 for its AI ultrasound technology that helps guide and interpret medical imagery for patient heart scans. 

The Caption Health AI Ultrasound Platform

The company claims its Caption AI ultrasound platform can turn any amateur into an expert sonographer. The Caption Guidance software provides more than 90 types of real-time instructions and feedback. Visual prompts help users make specific transducer movements to optimize and capture a diagnostic-quality image. At the same time, the machine continuously assesses and records the images, providing real-time feedback on image quality.  

A study to determine if Caption Guidance could indeed do the deed – make an expert out of a novice – assessed the outcomes of 240 ultrasound exams by eight registered nurses with no prior ultrasound experience. A panel of expert cardiologists judged the exam a success if it met a set of specific qualitative visual assessments. Almost all the scores ranged in the high 90s, with the lowest at 92.5%.

Caption Health clinical evidence.
Credit: Caption Health

Then there’s Caption Interpretation, which automates a measurement of cardiac function called ejection fraction (EF) from the imagery right at the point of care. EF is a key measurement to assess cardiac health across a spectrum of cardiovascular conditions. Caption Interpretation uses deep learning – trained on millions of images – to automatically select the best clips from ultrasound exams, incorporating three views into its fully automated EF calculation. While there are other commercially available EF measurement software solutions out there, Caption Health says its solution is the only one capable of analyzing every pixel and frame in a given clip to produce highly accurate EF measurements.

FDA Clearances for AI Ultrasound

The success from the study involving the nurse-led ultrasound exams prompted the FDA to clear Caption Guidance as a novel technology back in February under the agency’s De Novo regulatory pathway. Caption Health claims this is the first medical software authorized by the FDA that provides real-time AI guidance for medical imaging acquisition. 

The Caption AI Guidance platform.
Credit: Caption Health

In May, the agency granted expedited clearance to an update of the Caption Guidance software following a brief 25-day review in order to get the AI-guided medical imaging acquisition system in the hands of more frontline healthcare workers due to COVID-19. An increasing body of research suggests that the novel coronavirus poses even more risk to the heart than the world’s biggest hamburger. A recent study in the journal JAMA Cardiology, for example, reported COVID-19 patient mortality to be 15 times higher among hospitalized patients with cardiac injury. 

Finally, Caption Health got the green light from the FDA in July for updates to Caption Interpretation. This approval also appears to be driven by the need to get more diagnostic tools to healthcare workers battling COVID-19. A global survey published in a separate journal reported that cardiac abnormalities were observed in half of all COVID-19 patients undergoing ultrasound of the heart. Those findings based on the images directly led to how patients were managed in a third of all cases. That makes Caption AI a game changer, especially at facilities that lack enough trained personnel to conduct ultrasounds.

Conclusion

There are about 130,000 such qualified individuals employed as diagnostic medical sonographers and cardiovascular technologists and technicians, according to the Bureau of Labor Statistics (BLS). They are also well paid, with a median salary of almost $70,000. The BLS predicts 14% job growth in the coming years, so it doesn’t look like anyone is in danger of losing a job to Caption AI yet. But as we’re learning: Automation is a job killer.

You know what else is a killer? Heart disease, which claims nearly 700,000 lives per year, according to the U.S. Centers for Disease Control and Prevention. And that was before COVID-19. As of last month, Caption AI was in use at 11 medical centers in the United States. Based on the pressing need for quick and easy diagnostics, we would expect that number to grow fast in the next few years, especially with the new infusion of cash that Caption Health has at its command for marketing. Who knows? Maybe an MBA will be doing your next AI ultrasound and then billing you in real-time. 

Share

Leave a Reply

Your email address will not be published.