And the AI Healthcare Oscar goes to…

You’ve probably heard that healthcare is running out of power. Growing demand, plummeting manpower and increasing documentation requirements - all of that sounds overwhelming. But does it have to be that way? Definitely not, because just around the corner is the technology which is ready to help. Startups have been trying to address these problems for many years. Let’s take a look at what happened in 2018. What has changed? What challenges do companies face? Finally, let’s find out who deserves the “Healthcare Oscar” (I invented this award for this article. No offence, Academy!).

Taking over for medical specialists is probably not the main purpose of bringing AI to life - or is it? In May 2018, a study revealed that AI was better than dermatologists at detecting skin cancer. In 95% vs 86.6% of cases the computer was more accurate than specialists with many years of training. But does this mean that we no longer need doctors? For me, the answer is definitely “no”, and to demonstrate this, we have the Google x-ray project, which shows that AI won’t replace doctors any time soon, but rather will assist them in making better judgments.

Groundbreaking news came from the Food & Drug Administration (FDA) in the U.S. when two AI solutions were approved for use. In February, 2008, the FDA permitted marketing of clinical decision support software alerting providers when a potential stroke is detected. The software is called Viz.AI Contact, and is designed to analyze Computer Tomography images of the brain. When it detects a potential blockage, it sends a text notification to a neurovascular specialist, while at the same time the first-line specialist conducts a standard review of the images.

Later, in April 2018, the FDA approved another AI-driven program, IDx–DR, which looks for diabetic retinopathy that is worse than mild. This example is special, because IDx–DR was the first to be used without the need for confirmation by a specialist! With astonishing 87.4% detection accuracy, it is one of the best examples of how powerful AI can be.

The latest FDA approvals seem to be a clear message to all healthcare companies that Software as Medical Device (SaMD) is very welcome. Are we going to see more SaMD permitted for public use in 2019? I think so!

Big Pharma is also in on the search for innovations. In May 2018, Pfizer began a partnership with the AI startup XtalPi to predict pharmaceutical properties of small molecules in order to develop new drugs.

One of the biggest M&A deals in the AI market was made by another pharmaceutical company – Roche, which acquired Flatiron Health for $1.9 billion in 2018. Flatiron uses machine learning (an important part of all AI) to mine patient data. Roche hopes to use this ability to support its oncology pipeline. Personally, I hope that these initiatives will bring new drugs to the market much faster and that those drugs will be more effective in fighting diseases. Can you believe that in today's world, drug development takes around 12 years?! There is definitely room for improvement.

The next case worth noting is a company called Cardiogram. A recent study conducted by the company, in which 14,011 participants accessed the Cardiogram app through Apple Watch or Android SmartWear, suggests that diabetes can be detected using heart rate monitors in wearables. AI analyzes the heart rate variability path and detects undiagnosed diabetes with 85% accuracy. That’s an impressive result, so does this mean we can say goodbye to needles in diagnosing diabetes? Not yet. Remember that 85% is still not 100%. So why do we need companies like Cardiogram?

“By continuously monitoring using non-invasive wearables, we can identify when you’re at high risk of undiagnosed pre-diabetes or diabetes, offer you a free A1c test, and then guide you to clinically appropriate treatments,” says Cardiogram CEO Brandon Ballinger. The purpose of this novel approach is to reduce the number of people who are unaware that they have diabetes or may be at risk of developing it. This concept is the same as that suggested by Apple Inc. with their Heart Study, which is aimed at reducing the number of people unaware of irregular heart rhythms such as atrial fibrillation. In the end, the proper diagnosis is still in the hands of healthcare professionals.

What was the well-established annual tradition at the Oscars? Nominating Leonardo DiCaprio and not giving him the award at the end, of course. So here he is, our favorite Leo, a.k.a. Apple, which did a great job and is disrupting healthcare like never before. For a few years Apple has been on a mission to give back control over health data to patients themselves. At the beginning of 2018, it was announced that iPhone users would have access to all their electronic health records (EHR) from participating hospitals and clinics through the Health app.

Later, in June, Apple deployed the Health Records API for developers. This might not sound like a big deal in the context of this summary, but it opens up vast possibilities for early diagnosis, disease management, or clinical research studies.

In an age where information is money, nobody wants to share their medical records with third-party companies without a good reason. However, according to a recent survey by Accenture, 88% of consumers are willing to share the data from their wearable devices, like smartwatches, with health care professionals. Even more surprisingly, the same study revealed that 72% of respondents are fine with their health insurers having the same information as health professionals! It looks as though in the case of healthcare data, patients and consumers do not see a problem. Nevertheless, companies should secure the data as well as possible in order not to end up like Facebook with Cambridge Analytica, which scared some people off.

In 2017, the FDA, together with healthcare giants such as Apple Inc., Samsung and Roche, introduced the Pre-Cert Pilot Program. This program is intended to be an answer to the continual changes in digital health software, so that the certification process will not slow down development. Companies themselves, rather than software, are certified as meeting the agency’s safety and effectiveness standards. Is this a good approach? Some say that it only favors big players, while small startups are put in a losing position. In my opinion, this is probably not the best proposal, and I have to agree with those expressing concerns. But for now, it is good to hear that FDA is trying to adapt to the medical software environment, which on the one hand needs to feel as free as possible to develop, but on the other hand must be regulated - after all, this is healthcare we are talking about, with real people as the end users.

Across the ocean, the European Union Parliament and Council prepared a new set of regulations called the Medical Devices Regulation (MDR), which is going to replace the former Medical Device Directive (MDD) in 2020. While new regulations are a response to the fast-changing world of medical devices, they are going to impede the process of obtaining the required permissions and certifications.

Let’s get back to our nominees. This may be anticlimactic, but the answer is… everyone! And here’s why. The 2018 AI healthcare stage provided me with strong evidence that it had met and probably exceeded all expectations. With so many great projects in diagnosing diseases, helping develop new drugs, or supporting patients and doctors, AI in healthcare was probably one of the hottest topics of the previous year. But let the numbers speak for themselves. In the second quarter of 2018, AI in healthcare funding had its best result in history and is still growing. Since 2013, healthcare AI startups have raised around $4.3 billion, overtaking other industries taking part in AI deal activity. According to an article by Accenture with the self-explanatory title “AI: Healthcare’s new nervous system”, the market is expected to grow from $600 million in 2014 to a tremendous $6.6 billion in 2021, while at the same time bringing down the cost of healthcare.

2018 was not only special for AI as a whole. It was a breakthrough year for us as well, but that’s a story for another article.

As you can see, these and many more exciting discoveries (not included in this summary) went public in 2018 alone,  which encourages me to look forward to the future. What more could happen?

2019, bring it on!