Machine learning healthcare startups you should definitely know about

Machine Learning is emerging as a viable solution for healthcare startups

Healthcare startups are integrating machine learning in their products

A few days ago, I wrote a piece on how some startups engaged in big data analytics are transforming drug discovery landscape (here) and also about some unique big data healthcare startups (here). Thankfully for everyone, big data juggernaut is not limited to these applications. Innovators have gone a step further and are integrating machine learning algorithms to their data analytics platform to build next generation predictive tools. This has already led to common talks on the potential of machines replacing most of the manual work in the healthcare service segment.

I too am widely asked when will we see functions such as medical diagnosis completely taken over by machine learning. My answer is not in the foreseeable future. Why? Because one has to realize that the healthcare industry has several stakeholders, some critical, some influential and some both. A practitioner’s agenda is slightly different from a hospital’s agenda, which is slightly different from a lawmaker’s agenda. You talk to a practitioner and he will tell you why a machine will never be able to exclusively tell you what he can. You talk to a hospital and it will give you its own reasons why it can’t rely on a machine for the most critical part of its operations that will make it vulnerable to errors and lawsuits. These arguments are not entirely false. But has this really hurt the innovation process? Not really. There are always believers everywhere. If we look around, we’ll see that most ML start-ups do end up getting support from some hospitals, venture capitalists, healthcare facilities and academia, however far and few these stakeholders might be. Evidently, some technological interventions such as EHR management and imaging analytics are gaining faster grounds compared to other innovations. But today, we’ll have a look at the other segment. We’ll look at some of the start-ups that are rewriting the rules of employability of machine learning algorithms in the healthcare segment, particularly in diagnosis and patient care.

Diagnosis and Treatment Plan

Lumiata is a 2013 Silicon Valley based healthcare Graph Analysis startup. It employs machine learning to generate multi-dimensional probability distribution graph containing tens of thousands of nodes (e.g., symptoms, diseases and patient data points) and their weighted connections strengths in real-time. Its results can help identify patients who need urgent hospitalization, follow-up visits, medication plan, etc. It customer base includes hospitals and healthcare facilities. The startup refrains from calling its product a diagnosis tool and instead suggests that since it focuses on specific patient data, and factors in for location and time of the patient, it analysis opens up for more exploratory analysis for the doctor to probe on than exact diagnosis. Last year, Lumiata raised $10 million venture capital funding.

Founded in 2014, San Francisco based startup Enlitic stands out from the competition for its stated strategy of employing deep learning to assist doctors in the diagnosis and prognosis of diseases. It compares patient data (radiology, pathology, genomic, EHR, etc.) with similar data from millions of other patients to build its decision case. It raised $2 million of funding late last year and topped that with another $3 million investment earlier this year.

Oncora Medical
Philadelphia based Oncora Medical raised $20,000 in seed funding earlier this year. Founded by PhD students, the startup uses past cases data to help oncologists to create personalized treatment plans for patients based on what worked best in the past. Its machine learning algorithms also helps medics to see predictions on how the patient will respond to different treatment plans.

Founded in 2012, Israel headquartered MedAware aims to address the issue of prescription errors. Apparently, such prescription error can result in excessive healthcare costs and even mortality, in some cases. MedAware’s patent-pending technology uses big data analytics and machine learning algorithms to analyze large scale data of EHR to learn automatically how physicians treat patients in real life scenarios. Its software platform then detects prescription errors before they happen, employing a robust machine learning predictive analytics technique. While some EHR systems have in-built tools to detect excessive dosage and drug interactions, detection of a wrong drug will be a new valuable addition. It raised Series A funding of $1 million in last October. Brigham and Women’s Hospital in Boston is currently tested the product on its system.


Patient Care

Founded in 2010, New York based AiCure is a medication adherence startup. The company employs smartphone camera for observing and logging the patient taking her medication. A machine learning system loaded with facial recognition and motion sensing algorithms then verifies if the correct patient took the medication. This is considered to be especially useful in the cases of patients suffering from mental illnesses. It received a grant of $1 million from National Institute on Drug Abuse (NIDA) for running clinical trials with Cincinnati Addiction Research Center (CinARC). That was followed by another grant of $3.4 million from National Institute of Health (NIH). AiCure now wants to expand its portfolio and is focusing on developing its next product for monitoring and intervention for patients seeking therapy for opioid addiction.

Naperville based PhysIQ calls itself the first personalized physiology data analytics platform. Its platform can work with wearable health device or a physiological sensor. It made big news recently for receiving an FDA 510(k) clearance for its personalized physiology analytics system. What exactly is it? Its product is a proactive health monitoring system, categorized as a Class II healthcare device that calculates summary indices based on vital health inputs signs and acts as an early warning detection system. How is it different? PhysIQ product builds a personalized machine learning model for each patient that focuses on comparing patient data (heart rate, respiration rate, oximetry, and blood pressure, etc.) with her own baseline indicators and not with the population mean. The startup argues that this unique approach helps it to detect anomalies earlier. PhysIQ platform is for companies that are collecting vital health data from sensors and devices. The company has been noted for its partnerships with Samsung (SAMI) and Scripps Translational Science Institute (Ebola patient monitoring). It raised a funding of $4.6 million last year.

Healint is a Singapore based startup that launched its first product JustShakeIt in Dec 2013. JustShakeIt is an emergency triggering system that can be alerted by shaking a smartphone with a hand. The trigger system sends an alert to caregiver(s). This can be vital for patients with neurological conditions such as strokes, epilepsy and migraines, and who need urgent attention. The startup is working on developing predictive analytics tools and algorithms on smartphone sensors. A lot of the focus of the machine learning algorithms is on distinguishing an actual emergency shake of the phone from everyday jiggling. Healint promises to come up with more such products in the future. It received early funding from a startup accelerator in Asia and some angel investors. In March this year, it raised seed funding round of $1m. It has also partnered with UK-based Migraine Action.

It is refreshing to see how innovators of today are quickly integrating machine learning to address the myriad issues in healthcare. These are exciting times to be in!

Anubhav, a data scientist, writes about new developments and future trends in the machine learning and data analytics domain.
He can be reached at
Follow him on Twitter at:

Leave a Reply