Mayo Clinic struck two joint ventures to develop and commercialize AI diagnostic tools. The idea was to develop algorithms to support clinicians’ decisions, starting with a suite of cardiology tools.
The news comes a little over a month after Mayo Clinic announced it plans to create an “AI factory” as it expands its partnership with Google.
Mayo Clinic is launching the two companies, Lucem Health and Anumana, in conjunction with two startups.
“Why did we create two portfolio companies, one to do the data ingestion, orchestration and connectivity, and one to do algorithms? In many ways, they’re complementary problems,” Mayo Clinic Platform President Dr. John Halamka said in a Tuesday press briefing.
Although the current solutions are focused on cardiology, “you also have to imagine there will be neurology, behavioral health, radiation oncology, there will be many other algorithm companies that will be generated over time,” he said.
A joint venture for data
The first of the two joint ventures, Lucem Health, was created to pull in data from remote devices and normalize it so it can be used to create algorithms and new diagnostics tools. Mayo Clinic is pouring $6 million into the startup with Commure, a General Catalyst-backed startup that is based in San Francisco and Boston.
The idea is to pull in data in from all kinds of devices — “Everything from a remote ECG to a connected pulse oximeter to a connected stethoscope,” Lucem CEO Sean Cassidy said in a Wednesday press briefing.
It will also support algorithms and help integrate them into clinicians’ workflow.
A joint venture for algorithms
The second joint venture, called Anumana, was created to develop and commercialize algorithms for the early detection of disease. Mayo Clinic and Cambridge, Mass.-based Nference are jointly launching Anumana. They’re launching the joint venture with a $25.7 million series A investment.
It will build on an existing partnership between Mayo Clinic and Nference. Mayo Clinic was a lead investor in two of the startup’s funding rounds, including a recent $61 million series C round.
Last year, they struck a partnership to derive more information from anonymized medical records. As part of that, the goal was for Nference to digitize 25 million pathology slides in the next two-and-a-half years.
To start, Anumana will be tasked with developing neural network algorithms using ECG signals. It will also commercialize an algorithm developed under Mayo Clinic’s chair of cardiovascular medicine, Dr. Paul Friedman, for early detection of heart disease. The goal is to submit it to the FDA this year.
“We have a rich pipeline of ECG-based AI algorithms and we are actively pursuing FDA approval,” said David McMullin, chief business officer of Anumana and Nference. “We’re very excited about the potential of the initial pipeline that we have and the pipeline we will develop.”
Mayo Clinic doesn’t just plan to use the algorithms it developed internally. It also plans to make them available for other healthcare facilities.
“These are meant for the entire healthcare industry, not just Mayo Clinic patients,” Halamka said.
But getting there is a process. It requires FDA clearance, and as a recent paper in Nature pointed out, testing algorithms prospectively and across multiple sites is a best practice for ensuring their accuracy.
Patients’ privacy also must be protected in collaborations with outside companies.
Anumana does not plan to run a prospective study of the algorithm it’s submitting to the FDA for clearance, but it does plan to test the algorithm across multiple sites, McMullin said.
Both partnerships use de-identified patient records. An institutional review board had to oversee the process before data could be used, and [patients give consent for their data to be used when they come to Mayo Clinic.
Halamka said patients have the ability to deny the use of their health data for research. But going forward, he acknowledged the use of data will become increasingly complicated as companies work with a combination of health record data, patient-submitted data and genomic data.
Beyond the first FDA submission, the companies plan to develop more algorithms in the future.
“We publish scientifically what we do and Anumana will release a roadmap for when they will come out,” Friedman said. “There are a number that are already developed and we’re testing others that are in the earlier stages of the pipeline.”
Photo: from2015, Getty Images