fbpx

AI Detects Heart Failure From One Beat

Virtual image of human heart with cardiogram

Heart failure has been in the news a lot lately. It used to be the disease of older people, but younger ones (40-50) are gaining. The negative trend for younger people is mostly related to lifestyle. For us older (and wiser?) people, nearly 10 percent of those over 65 are afflicted with some form of congestive heart failure (CHF). As a side note, my mother died from CHF. Inspired by memories of what she went through, I did some research on using home telemonitoring to prevent readmissions to the hospital, a phenomenon which, unfortunately, accompanies CHF. My doctoral dissertation about the research is here.

More than 25 million people around the world have CHF. It is a costly chronic disease. CHF has a variety of causes but it is usually the result of the heart being unable to pump blood effectively through the body. Detecting the disease early and efficiently can have a large impact on the total cost of healthcare. A lot of the cost is associated with tests. In the long run, I believe the solution will revolve around AI and data.

The good news is there is a growing amount of data. The introduction of Apple Watch and other mobile/wearable health devices are collecting continuous streams of data from millions of people. If the data is anonymized, meaning stripped of any personally identifiable information, and accumulated in publicly available databases, great progress can be made.

Medical research in Europe gives a clue as to the potential. If you read Robot Attitude: How Robots and Artificial Intelligence Will Make Our Lives Better, you know about artificial intelligence (AI). A subset of AI is machine learning (ML), which I explained in the book. One of the algorithms used in the field of ML is called convolutional neural networks (CNN). What the researchers have demonstrated with CNN is mind boggling. They have been able to identify CHF almost instantly by applying the algorithm to just one heartbeat from ECG data. The accuracy of the detection was measured across a very large database of known CHF patients and those without. The accuracy was 100%.

New Atlas, a nearly 20 year old technology website, reported,

Even more interesting is the possibility of wearable health monitoring devices being able to help doctors identify at-risk patients without having to examine them in clinical contexts. Using short ECG recordings to detect CHF, could pave the way for health wearables that constantly monitor patients in real-world conditions.

The research I have described was published in the journal Biomedical Signal Processing and Control and reported in New Atlas as “100% accurate AI detects heart failure from single heartbeat“.

I mentioned at the beginning of the article detecting CHF early and efficiently can have a large impact on the total cost of healthcare. In my opinion, AI tools are going to have a huge impact. Some studies have suggested the cost of unnecessary tests and procedures in American healthcare is as much as $1.5 trillion. Visit the waiting room at a Florida cardiology practice or a radiology imaging center and you will see rooms full of seniors waiting to get tests.

Consider the impact. If there are 50 million people without healthcare insurance or who are under-insured, and if the cost per year of their care would be $10,000 (the average for a Medicare patient), the cost to give them healthcare, would be $500 billion or just one third of the unnecessary cost. This does not include the cost of fraud, the billions spent on TV advertising which add no benefits to our health, the over-charging of medications, and other inefficiencies.

I am not suggesting we give free healthcare. I believe everyone should pay some fair portion of the cost, if possible. I am simply making the point the real problem with our healthcare system is not the insurance, who is in and who is out, etc. The #1 problem is the excessive cost of our healthcare compared to other developed countries. AI is coming to the rescue.