FDA clears Koios Medical's ultrasound-reading AI that spots breast, thyroid cancer in 2 seconds

Cancer cells
Koios DS uses artificial intelligence algorithms to analyze ultrasound images in a matter of seconds, comparing them to a compendium of hundreds of thousands of previous diagnoses to detect breast and thyroid cancer. (koto_feja)

In the grand tradition of the inquisitive Greek titan that it’s named for, Koios Medical is asking big questions of current medical practices—then developing the advanced artificial intelligence technology to answer them.

Its flagship product is Koios DS, an AI-powered software platform that analyzes ultrasound images to help diagnose both breast and thyroid cancer. Combined, the two cancers comprise 375,000 diagnoses in the U.S. each year, with more than 2.2 million biopsies of breast and thyroid tissue performed annually.

The software is newly cleared by the FDA, only a few months after it received the agency’s breakthrough device designation in April.

The Koios DS AI algorithms compare ultrasounds to a compendium of hundreds of thousands of previous images with confirmed diagnoses, collected from 48 sites around the world. The resulting analysis takes approximately two seconds to complete, ensuring that clinicians can make an accurate diagnosis and begin cancer treatment as quickly as possible.

“The loss of life globally to cancers found too late, or misdiagnosed, is tragic. When combined with the millions wasted on avoidable procedures, we are compelled to relentlessly innovate, building powerful AI models directly into easy-to-use software,” said CEO Chad McClennan. “This new FDA clearance means physicians are now poised to save tens of thousands of lives while freeing up precious time and resources that can be used more effectively and elevate quality of life.”

RELATED: Ultrasound-analyzing AI improves breast cancer diagnoses by 37%, NYU study finds

In studies of the Thyroid DS software, physicians were able to detect thyroid cancer at a rate about 14% higher than when using other methods, and the AI also allowed them to cut back on unnecessary biopsy orders by more than 35%. Meanwhile, they were able to slash the amount of time spent analyzing each case by nearly 25%.

Breast DS, meanwhile, has been validated in a host of previous studies. It was originally cleared by the FDA in 2016, then received a follow-up regulatory OK in 2019 for an updated version of the breast cancer-spotting software. It began rolling out in Europe, the Middle East and South America earlier this year.

The breast cancer-specific algorithms were shown in a 2017 study to spot 100% of actual cancer cases. With that huge boost in accuracy, clinicians could reduce the number of benign lesions unnecessarily sent for biopsies by almost 70%.

“The typical decision-making paradigm relies on tradeoffs; trading sensitivity for specificity, efficiency for thoroughness, but the only thing enforcing this paradigm is the inability to shift off of these tradeoff curves in place of shifting along them. This novel software demonstrates that using AI for decision support, physicians can make clinically meaningful shifts in performance improving interpretation efficacy and diagnostic performance, improving sensitivity and reducing false positives,” said Lev Barinov, Ph.D., Koios’ clinical vice president.

RELATED: An instant 2nd opinion: Google's DeepMind AI bests doctors at breast cancer screening

The Koios DS software integrates into a hospital’s existing picture archiving and communication systems to analyze imaging data. It can also be installed directly into GE Healthcare’s Logiq E10 ultrasound scanner.