HeartFlow’s AI coronary test outperforms conventional alternatives

Coronary artery disease is notoriously tricky to diagnose, which is one of the reasons it accounts for 801,000 deaths per year in the U.S. alone, according to the American Heart Association. (That’s more than all forms of cancer and Chronic Lower Respiratory Disease combined.) It’s also expensive — the cost of treating cardiovascular diseases is projected to pass $1.04 trillion globally by 2030, up from $863 billion in 2010.

That’s why HeartFlow, a Redwood City, California-based health startup founded by Stanford University alums Charles Taylor and Christopher Zarins, has spent the better part of 11 years commercializing an alternative to traditional screening. Their solution, the U.S. Food and Drug Administration-cleared FFRct (fractional flow reserve) Analysis, uses deep learning algorithms to build a personalized, three-dimensional model of the heart and highlight potential problem areas for clinicians. Today, at the EuroPCR Conference in Paris, HeartFlow announced data from a clinical study showing that its solution outperformed other commonly used (and in some cases more expensive) tests.

The first trial (PACIFIC), which included 208 patients who underwent HeartFlow Analysis, CTA, SPECT, and PET scans and a three-vessel invasive fractional flow reserve (a procedure that measures heart circulation using a catheter and pressure-sensitive guidewire) shows that HealthFlow’s solution obtained routinely superior results on a per-vessel basis. The FFRct Analysis achieved a diagnostic performance score of 0.94, compared to the PET (0.87), coronary CTA (0.83), the SPECT (0.70). (Higher is better.)

In the second of the two trials (SYNTAX III), two teams of medical professionals, each with an interventional cardiologist, radiologist, and cardiac surgeon, made treatment recommendations for more than 200 patients based on either data from a coronary CTA and HeartFlow Analysis or a coronary angiogram (another invasive procedure that uses a catheter to access the heart). Results showed the two teams were in almost “perfect” agreement.

In the U.S., an angiogram can cost $4,000 or more, while Medicare reimburses $1,450 for HeartFlow Analysis screenings.

“The new data […] shows that the HeartFlow Analysis can hold great clinical value in real-world practice for the non-invasive evaluation of coronary artery disease, as it provides both anatomic and functional evaluation of coronary lesions,” said Campbell Rogers, chief medical officer at HeartFlow. “A HeartFlow-guided pathway provides physicians with functional information that was previously only available via invasive measures and can reduce the need for patients to undergo additional tests.”

Rogers said that the HeartFlow’s ability to show both the severity of coronary artery disease (CAD) and its impact on blood flow is unmatched among rival non-invasive tests. It works like this: A scan of the patient’s heart (a non-invasive coronary computed tomography angiogram, or CCTA) is uploaded from a hospital’s electronic records system to HeartFlow’s AWS-powered cloud platform. The HeartFlow Analysis then applies algorithms to simulate the blood flow and assess the impact of blockages, and the final result is made available to physicians through an online portal.

HeartFlow, which announced a $240 million funding round in February from Wellington Management, Baillie Gifford & Company, and other existing investors, says that HeartFlow Analysis has been used by clinicians at over 80 medical institutions across the U.S., Japan, Europe, and elsewhere to diagnose more than 15,000 patients. The company counts GE subsidiary GE Healthcare, Siemens, and Philips among its collaborators.