Inside the Briefcase

Webcast: HOW TO SCALE A DATA LITERACY PROGRAM AT YOUR ORGANIZATION

Webcast: HOW TO SCALE A DATA LITERACY PROGRAM AT YOUR ORGANIZATION

Join data & analytics leaders from Starbucks, Cardinal Health,...

How EverQuote Democratized Data Through Self-Service Analytics

How EverQuote Democratized Data Through Self-Service Analytics

During our recent webinar on scaling self-service analytics, AtScale...

How Analytics Governance Empowers Self-Service BI

How Analytics Governance Empowers Self-Service BI

The benefit of implementing analytics policies at the semantic...

How To Create A Powerful SMS Marketing Strategy

How To Create A Powerful SMS Marketing Strategy

Many small businesses are looking for more ways to...

Emerging Frameworks & Technologies that Combat the Rising Threat of Cyber Attacks

Emerging Frameworks & Technologies that Combat the Rising Threat of Cyber Attacks

The creation of the first computer virus in 1971...

How Machine Learning is Beginning to Influence Medical Device Technology: An Example with AED Units

April 17, 2020 No Comments

Featured article by Micah Bongberg, a medical device industry executive

While it may seem like everything is being automated, and artificial intelligence and machine learning are taking over the world, that’s not the case in the medical device industry. Since medical devices have to go through pre-market approvals prior to sales in the United States, sometimes important, even, life-saving equipment doesn’t use the latest technology.

As consumer products, like cell phones and tablets, have been powered by AI for years, it’s a new trend for AI to begin its transformation into healthcare. So, while you may have thought (or “hoped”) to have the “best” tech to solve the most severe threats we face to our health, that’s not always the case.

Many times, after a lot of development time and resources, get poured into a project, when time is money, manufacturers have to get across the FDA finish line in order to begin recouping their investment. This can create a desire to build products that will get approved sooner than later, such as a product that uses a predicate device that has already been approved.

Inventors, and the investors who fund them, may be less likely to try to push the limits and add AI and new machine-learning capabilities to their products because they worry that it’ll slow down their timeline and pathway to approvals. The FDA may require additional test data, which adds time and money, to get comfortable enough to offer approval.

Take, for example, one important medical device, an automated external defibrillator (or “AED unit”). AED units treat people who succumb to sudden cardiac arrest. Sudden cardiac arrest is an electrical condition that creates an abnormal heart rhythm. It’s different from a “heart attack” which is a “plumbing” condition.

Since it’s an electrical condition, it takes the lives of all genders, races, and people of any age. In fact, 7,000-10,000 children go into sudden cardiac arrest and die every year.

When people go into cardiac arrest, their heart stops pumping oxygen-rich blood to the brain and other vital organs. Lacking oxygen, victims collapse and are clinically dead (no heart rhythm), though there’s a chance that their heart can be “shocked” and brought back into a normal heart rhythm before they biologically die, and can’t be brought back to life.

Contrary to popular knowledge, sudden cardiac arrest is the leading cause of death in the United States, killing 350,000 people per year. By way of comparison, fires only kill 4,000 people per year.

With early defibrillation, AEDs can save the majority of people who go into cardiac arrest. They are easy to use and safe. AEDs use “smart technology” to read a patient’s heart rhythm, and, if they’re in a rhythm that needs a shock, the AED will charge and allow the responder to deliver the shock (semi-automatic versions) or deliver the shock automatically (fully-automatic versions). Since they use this technology, AEDs are difficult to misuse – you can’t shock someone who doesn’t need it – thus the medical therapy can’t hurt someone.

So, you have the leading killer in the US. A therapy that can save the majority of victims. The therapy is safe and effective; it can’t be misused to hurt someone (like a drug that can be misappropriated, misdiagnosed, or misused by someone other than the patient).

Yet, when we look at AED units currently marketed, they’re using twenty-year-old (+!) technology.

We don’t use a 20-year old computer. What was your cell phone like 20 years ago? Did you even have one?

A promising new development in the defibrillator world is a recent publication sponsored by Avive Solutions out of San Francisco. Avive produced an algorithm that uses machine learning (used to determine whether or not a patient should or shouldn’t receive a “shock”), and, the results published in the Journal of the American College of Cardiology are astounding.

chizzle 234x300 How Machine Learning is Beginning to Influence Medical Device Technology: An Example with AED Units(source: http://www.onlinejacc.org/content/75/11_Supplement_1/3468)

“Overall, the convolution neural network demonstrated greater than 99% diagnostic accuracy for shockable and non-shockable rhythms. The direct application of this innovation embedded in a new AED device is unique, and offers significant implications to improve sudden cardiac arrest process of care,” states Dr. Sanjeev Bhavnani, Division of Cardiology and Healthcare Innovation at Scripps Clinic, the study’s principal investigator.

Not only did the Avive machine learning algorithm exceed the American Heart Association’s performance criteria, but it did so by using a sample set of data that far exceeded the recommendations. This means that the data is more accurate and less likely to be influenced by error.

The product is still under FDA review and it’s not yet available for purchase.

With a little more time, when we think of AI and the role machine learning can play in healthcare, it will soon fall under the “game-changing” category of innovation.

While AED units are one important example of the opportunities that exist with machine learning, the theory of using massive amounts of data to make clinical decisions is exciting. An approach that can be applied to many healthcare and medical products and software services to support them.

About the Author

Micah Bongberg is a medical device industry executive, currently working on bringing a smaller, lighter, more affordable AED unit to the market.

HEALTH IT

Sorry, the comment form is closed at this time.

ADVERTISEMENT

Gartner