Healthcare, according to CBInsights, is the hottest area of investment within the very hot artificial intelligence space. “Investors have poured in $1.5B across nearly 190 deals into healthcare-focused AI startups in the last 5 years,” it writes.
Devices that can teach themselves how to do things, aka “smart machines” are blowing up all over the place. According to Gartner, smart machine revenues will reach $29 billion by 2021 and 30 percent of large companies will be using them.
Gartner analyst Tom Austin believes we are finally in the “Smart Machine Age.” Gartner also predicts that“the most disruptive class of technologies” over the next decade is tech involving smart machines.
Experts at Gartner and elsewhere tend to agree about the three prime movers of recent machine learning advances. They are:
- Better hardware
- More data
- Improved deep neural networks
In almost no other field has the sheer amount of data increased as fast as it has in healthcare.
Here are five healthcare AI startups to watch:
In 2016, two new AI-focused unicorns emerged: iCarbonX and Flatiron Health.
- China Bridge Capital
- Tencent Holdings
Category: Personalized medicine
A person’s response to lifestyle changes, medication, surgery, and other interventions depends in part on their genomics, lifestyle, and other factors. Personalized medicine connects those factors to risks and outcomes in order to provide treatments and preventative measures tailored to each person.
The challenge is threefold. First, you have to collect enough data to gain useful insights. Second, you have to clean and store that data in connected databases. Third, you need efficient, effective machine learning algorithms to actually make the connections.
He recently left BGI to found iCarbonX. His new company is combining machine learning with genomics to crack the nut of accurately predicting health outcomes and creating personalized healthcare plans to reduce health risks.
Wang claims his company can collect the same data as his competitors for less money. Specifically, Wang predicts he’ll have samples and data from one million people within five years. And with more funding than any of its competitors, plus its impressive partnerships, iCarbonX is poised to succeed.
It’s currently developing an app called Meum (Latin for “my”) where clients can enter data and receive personalized advice on questions like what to eat or how much sleep they need.
- Zone Startups
Category: Diagnostic clinical imaging
Healthcare-focused AI funding has clustered around imaging and diagnostic startups, with a third of all healthcare AI startups that raised after January 2015 working in this area. In addition to Advenio TecnoSys, companies like Proscia and Imagen Technologies also recently received their first fundings.
Advenio TecnoSys founder Mausumi Acharyya holds multiple diagnostic clinical imaging-related patents. She is also one of 15 women selected to participate in empoWer, Zone Startups India’s six-week accelerator program backed by the likes of Vodafone India and Google.
Advenio TecnoSys is developing hardware-neutral CADx plug-ins that will harness machine learning to help improve and automate clinical imaging. In 2015, the company submitted a handheld portable device to StartHealth Competition India that could diagnose retinal disorders in eyes that haven’t been dilated.
- Bain Capital Ventures
- First Analysis
- SSM Partners
- Undisclosed Angel Investors
- Undisclosed Investors
Category: Risk analytics
Insights and risk analytics topped CBInsights’ list of aggregate healthcare AI deals since 2015.
As discussed in the context of iCarbonX, using machine learning to make accurate predictions about treatments, outcomes, and health risks requires not just tons of data, but structured data. Unfortunately, 80 percent of patient health data is unstructured, like your physician’s handwritten notes.
Apixio raised over $19M in Series D in 2016 and is working to structure patient data and then use natural language processing to correctly assess a patient’s risk of certain illnesses based on their medical history and demography.
“If we want to learn how to better care for individuals and understand more about the health of the population as a whole, we need to be able to mine unstructured data for insights,” CEO Darren Schulte said.
Medical coders are experts trained in reading medical charts, finding information about diseases and treatment, and coding the information. Apixio is robotic process automation for the task, helping coders code two or three times as many charts per hour with up to 20 percent fewer errors than without software.
Gaps in patient documentation happen when there’s no recent assessment or plan for a patient’s chronic disease. Apixio is better than humans at identifying those gaps. For instance, Apixio coded a population of 25,000 patients and found 5,000 instances of diseases that were not documented clearly and appropriately over a nine-month period.
Where iCarbonX is targeting end users, Apixio is looking to provide this information to insurance providers as well as healthcare delivery networks (HDOs) including hospitals and clinics.
Its first product is called HCC Profiler. HCC is a risk scoring method used by the Center for Medicare Services to adjust payments for plans.
- Andreessen Horowitz
- Cardinal Health
- Data Collective
- Founders Fund
- Shana FisherRead
Category: Liquid biopsy diagnosis
At least 80 percent of cancers are fully curable when caught in time. Which means that the cure for cancer effectively already exists. The challenge, then, is detection. It’s for this reason that TechCrunch calls pinpointing the exact direction and severity of cancer “the biggest challenge in medical testing today.”
The cure for cancer “will be our ability to predict that someone has cancer far before traditional means,” says Vijay Pande. He’s a longtime Stanford professor and is leading Andreessen Horowitz’s months-old, $200 million Bio Fund, which has invested in Freenome.
Freenome diagnoses cancers and recommends treatment based on their composition using AI to detect their particular mutations. Cancer or not cancer doesn’t tell you what stage the cancer is in, what direction it’s growing in, what type of cancer it is, or other indicators relevant to prognosis and treatment.
- Deep Fork Capital
- Dolby Family Ventures
- Intermountain Healthcare
- Khosla Ventures
- Marc Benioff
Category: Medical imaging insights
Wired reports that Zebra Medical claims that its AI beat human radiologists at detecting cancerous cells. While Zebra reached 91 percent accuracy, radiologists only reach 88 percent on average. And Zebra got fewer false positives too.
It plans to submit its breast cancer detection algorithm to the Food and Drug Administration toward the end of 2017, after US clinical trials in hospitals.
“In five or seven years, radiologists won’t be doing the same job they’re doing today,” said Zebra Medical founder Elad Benjamin. “They’re going to have analytics engines or bots like ours that will be doing 60, 70, 80 percent of their work.”
Below is a heatmap from CBInsights of the hottest areas of investment for healthcare AI startups.
Watch these five healthcare AI startups, because they’re changing what healthcare looks like. In five years, jobs like medical coder and radiologist will be utterly transformed by AI.
To learn more about AI check out our artificial intelligence software directory.
Looking for Medical Practice Management software? Check out Capterra's list of the best Medical Practice Management software solutions.