Four healthcare startups with artificial intelligence technologies are bringing their offerings to the market with help from the Mayo Clinic.
The companies are part of the first cohort of Mayo Clinic Platform_Accelerate, an inaugural program that has been working with these startups since March. On Wednesday, the companies — cliexa, Quadrant Health, ScienceIO and Seer Medical — had the opportunity to showcase their technologies at the Mayo Clinic Platform Conference.
In order to apply for the program, startups need an AI model either on the market or close to market, according to Accelerate’s website. Each company accepted to the 20-week program receives $200,000. Mayo’s experts work with the companies to understand AI model requirements, check for fairness and bias in their AI models and help them understand the Food and Drug Administration clearance process. Accelerate also provides the participants with access to Mayo Clinic’s deidentified patient data, which helps them validate their models and plan clinical validation studies.
“These companies were able to take already exceptional products and services and accelerate them with access to rich deidentified data sets and expert advice,” said Jamie Sundsbak, partner relationship manager of the program.
For these companies, the program has proven substantial benefits, according to Eric Harnisch, vice president of partner programs with Mayo Clinic Platform. Despite a tough job market at the beginning of March, one of the companies was able to fill a few roles that had been open for a long time because of Accelerate. All of the companies were contacted by investors, and some were contacted by providers who wanted to work with them, Harnisch said.
This is just the start of the Accelerate program. A second cohort of seven companies is beginning this week, Harnisch added. Accelerate will continue twice a year.
The four companies in the first cohort of Mayo Clinic Platform_Accelerate:
Melbourne, Australia-based Seer Medical’s focus is on helping those with epilepsy. Currently, there is limited access to quality testing, and physicians have difficulty diagnosing and managing the disease, said CEO Dean Freestone. Seer Medical provides physicians with data that makes it easier to diagnose and manage conditions. It is already established in Australia and is using the Accelerate program to help expand its presence in the U.S.
“Currently, it takes on average about five years for a patient to get to a good treatment,” Freestone said. “We want to reduce that down to a matter of a few weeks.”
Denver-based cliexa provides a suite of support tools and software that helps providers deliver care faster. The platform spans patient onboarding, remote patient monitoring, provider documentation, automated scheduling, revenue cycle management and telehealth, according to its website.
“We sell solutions to providers to help fill in areas of their care that they may be doing repetitive work on, where we can essentially provide them tools that give them a quicker way to provide the same level of care,” said Arin Seidlitz, director of product design.
Boston-based ScienceIO creates machine learning models that take unstructured healthcare information — such as documents, physician notes and internal databases — and make them more structured for easier analysis. Ultimately the technology reduces the administrative process for both payers and providers.
“Realistically, what this means is if you’re a hospital, you’re spitting out thousands and thousands of pages of just notes about your patients,” Founder and CEO Will Manidis said. “It’s really hard for you to find out what’s happening in those notes. Our API [application programming interface] takes those notes in addition to other documents, and turns it into a thing you can actually use for data science or machine learning or provide better care.”
New York City-based Quadrant Health aims to use AI to analyze electronic health records and patient data to predict diseases before they occur, said CEO Anin Sayana. He claimed that Quadrant’s model can predict diseases 42 hours before clinicians suspect it, and provides 40% fewer false positives than typical models.
“When clinicians use our model in a clinical setting, they are much more likely to trust that the models are accurate, will deliver useful predictions and give them sufficient time to actually make a difference in the patient,” Sayana said.
Photo: metamorworks, Getty Images