Last year, investors rode some record highs. The market charged upward, despite The Rona. More investors spazzed out on SPACs, to the tune of nearly 250 IPOs that raised about $80 billion. That’s about as much funding ($80.6 billion) that was funneled into global healthcare startups in 2020. That’s a new record, according to the big brains at CB Insights. AI healthcare startups alone raised $6.6 billion – another record high.
The old adage is that 90% of startups fail within five years, which means most of that money could have been put to better use, like a lost weekend in Cabo with your favorite drug dealer. It’s often difficult to gauge the success of private companies because they generally don’t let the public nose around in the books. Instead, they release meaningless press releases that claim a 100% increase in revenue with no baseline for comparison.
One metric to judge how well an AI healthcare startup is delivering the goods – especially in diagnostics or therapeutics – could rely upon whether its platform or medical device has approval or clearance from the U.S. Food and Drug Administration (FDA). The federal agency in charge of regulating everything from hemorrhoidal creams to herpes vaccines, FDA is finally recognizing that AI and machine learning require some different criteria to evaluate safety and effectiveness. We expect that some of the changes underway will accelerate FDA approval of AI algorithms to diagnose and treat disease. There are currently about 60 FDA-approved AI-based algorithms and medical devices, according to a database compiled by The Medical Futurist.
The technology first began to gain real traction in 2018, with only three approvals prior to then. That was the same year we compiled our own list of FDA-approved algorithms. In all, we have written about at least 18 companies whose machine-learning platforms or AI-powered medical devices have gotten the green light from FDA for different kinds of applications.
At least a couple of them – Viz for spotting blood clots and Digital Diagnostics for identifying diabetic retinopathy – have been approved for Medicare and Medicaid reimbursement. That generally denotes mainstream adoption, so expect your local insurer to be overcharging the government soon enough.
AI Healthcare Startups with FDA-Approved Algorithms
That still leaves a lot of ground to cover. In this article, we stick to AI healthcare startups not previously profiled by Nanalyze that have mainly developed hardware-agnostic software solutions.
FDA-Approved Algorithms for Radiology
Radiology is probably one of the most common use cases for AI algorithms today thanks to the maturation of computer vision technologies for scanning through images faster than a teenager with his dad’s Playboy collection. Founded five years ago this month, Tel Aviv-based Aidoc has raised $61.5 million, mainly from hometown investors and a VC firm out of Australia. Started by three graduates of an elite science program of the Israeli Defense Forces, Aidoc’s most recent round was a $20 million Series B last September. The Israeli startup has developed not one but six algorithms that have been cleared by both the FDA and its European Union equivalent:
The algorithms from Aidoc specialize in triaging critical, unexpected conditions that exist outside of the original diagnostic goals of the test. For example, the most recent FDA-approved algorithm flags and communicates incidental pulmonary embolism, a problem that can occur in cancer patients. The company’s algorithms for triaging stroke are now covered by Medicare and Medicaid.
About one in six deaths from cardiovascular disease are caused by strokes, a scary statistic that AI healthcare startups like RapidAI are trying to improve. The 10-year-old Silicon Valley company raised $25 million last September in its only disclosed round. The Rapid platform uses AI to create high-quality, advanced images from various CT and MRI scans. Its first FDA-approved algorithm was cleared for use in 2018. Its most recent clearance came through last August for an algorithm that detects suspected large vessel occlusions in as few as three minutes, with a sensitivity of 97% and a specificity of 96%.
RapidAI claims its platform has performed a million scans from more than 1,600 hospitals in over 50 countries.
FDA-Approved Algorithms for Diabetes Management
While a cure for diabetes remains elusive, it is possible to manage the disease and lead a life almost as fulfilling as an MBA’s, especially if you can enlist AI to tell you how to do it. Founded in 2014, DreaMed is another AI healthcare startup out of Israel. It has raised about $10.3 million, including a $5 million grant last September. One of its earliest investors was medical device company Medtronic (MDT), which eventually licensed DreaMed’s artificial pancreas algorithm. The company is now developing an AI platform called Advisor Pro for personalized management of diabetes:
The algorithm ingests data from an insulin pump and glucose monitoring device (self administered or continuous). The adaptive learning algorithms reputedly “emulate the way expert endocrinologists evaluate their patients, progressively refining their understanding of each case using accumulated information including collating, cross-referencing and analyzing all that critical, patient-specific information – both real-time and archived.” DreaMed is marketing Advisor Pro as a telehealth solution.
For those who can’t manage their insulin, one complication involves diabetic retinopathy, which could lead to blindness. Like Digital Diagnostics, Los Angeles-based startup Eyenuk has developed an FDA-approved algorithm called EyeArt for detecting both early and severe forms of the eye disease. Founded in 2010, the company has raised $16.7 million across nearly a dozen rounds. Validated on more than 100,000 patient visits, EyeArt autonomously analyzes retinal images from a fundus camera, returning results in less than 60 seconds. A new solution, EyeMark, is in the works to automatically track how the disease is progressing from visit to visit.
Eyenuk is also expanding its eye-detection analytics into other disease conditions, including age-related macular degeneration and glaucoma.
FDA-Approved Algorithms for Cardiovascular Disease
Heart disease is the No. 1 killer in the United States, so there are a ton of companies looking at ways to use AI to monitor and diagnose cardiac conditions. We recently covered a couple of FDA-approved algorithms. AliveCor uses machine-learning algorithms to process and interpret heart rhythms, while Eko Device has created an AI-powered stethoscope. Now we have Ultromics, a company spun out of the University of Oxford in 2017. It has raised $26 million in funding to develop algorithms that can analyze echocardiograms. So far, it has released two cloud-based solutions – EchoGo Core and EchoGo Pro – that return results in just minutes.
The former automatically calculates cardiac function, while the latter assists physicians identify heart disease risk. Both have been cleared by the FDA, with EchoGo Pro receiving clearance just last month, thanks to a diagnostic performance of more than 90% that has been validated in 6,000 patients and counting.
FDA-Approved Algorithms for Sleep Disorders
Sleep disorders like insomnia and sleep apnea have been on the rise for years. A company like Inspire Medical Systems (INSP), which offers a novel electroceutical therapy for sleep apnea, has seen its value skyrocket by +313% compared to only +88% for the NASDAQ over the last two years. That would seem to bode well for EnsoData, a Madison, Wisconsin-based startup that uses AI for sleep analysis. Founded in 2015, the company has raised $11.1 million, including a $9 million Series A in June 2020 that included VC firms with names like Dreamit and SleepScore. The company’s platform analyzes what’s called waveform data (the squiggly and curvy lines from echocardiograms and similar sensors) and other sources like electronic health records to automate the workflow for sleep technicians.
Crunchbase reported that EnsoData has amassed a dataset from more than 400,000 users, which the company claims is 50 times bigger than any public dataset. Its software is used by more than 300 sleep clinics in the United States and recently expanded to Latin America. The startup is also developing its Waveform AI platform for additional applications in neurology, cardiology, and ICU settings.
The proliferation of FDA-approved algorithms is making it something of a nightmare to keep track of them all. For instance, Nightwear out of Minneapolis hasn’t shown up on The Medical Futurist database, even though it got FDA clearance last November for an Apple Watch and iPhone app designed to improve sleep quality for those suffering nightmares related to PTSD. The company has raised only $305,000 to date. The algorithm works by learning the wearer’s sleep patterns by monitoring heart rate and movement, using similar technology developed for fitness wearables. After about 10 days of observations, the watch will vibrate enough to interrupt the nightmare, without waking the wearer. The 2.0 version will even check under the bed for monsters.
FDA-Approved Algorithms for Remote Patient Monitoring
The use of AI to monitor patients remotely is certainly a boon at a time when no one really wants to spend more time than necessary in the hospital. Founded in 2013, Chicago-based PhysIQ has raised $26.4 million to continue cranking out algorithms designed to make better sense of wearable sensor data. It uses a deep neural network to provide cloud-based analytics using a range of physiological data, including heart rate, heart rate variability, atrial fibrillation detection, and respiration rate. Here’s the platform, pinpointIQ, in a format that even an MBA can understand:
In one peer-reviewed research paper, PhysIQ scientists demonstrated that the company’s AI can predict impending heart failure from sensor data as well as an implanted device, which means we can monitor patients without invasive surgery and at much lower costs.
There are now thousands of startups that claim to be developing AI solutions across a dizzying range of applications. It’s often hard to figure out the real deal from the hype. However, when it comes to relying on a piece of software for life and death matters, you want to be sure that it does what it says it does. Algorithms that have FDA approval or clearance are most likely going to be the real deal. That provides the companies themselves a way to stand above the crowd in a much-hyped and competitive space. It’s also one metric retail investors should keep in mind as more opportunities arise to invest in AI healthcare companies.