AI researcher merged engineering and medicine to bring meaning to his work

Dr. Antonis Armoundas was studying electrical engineering at the National Technical University of Athens in his native Greece when he decided the subject was a bit too dry for his taste.

"I couldn't find any meaning to what I was doing," he said. So, he started looking into biomedical engineering, which applies engineering's problem-solving tools to biology and medicine to improve human health. His interest piqued, he dug into a senior thesis on the topic, then pursued a master's degree in biomedical engineering from Boston University and a doctorate in nuclear engineering from the Massachusetts Institute of Technology.

Along the way, Armoundas' passion for medicine grew.

"I felt like I was serving a purpose that was more meaningful," he said. "Medicine became my main passion, but using engineering principles in medicine provided that meaning."

Armoundas now studies the ways artificial intelligence, or AI, can improve health care. He is principal investigator at the Cardiovascular Research Center at Massachusetts General Hospital in Boston and an associate professor of medicine at Harvard Medical School. He talked about his AI work with American Heart Association News as part of "The Experts Say," a series in which specialists explain how they apply what they've learned to their own lives. The following interview has been edited.

How is AI being used in the field of cardiology?

We know that AI is used in every aspect of cardiology, including in the diagnosis, classification and treatment of cardiovascular diseases. I chaired the writing committee for an American Heart Association scientific statement earlier this year where we addressed the main areas in which it is applied. Some examples of how it's used include interpreting cardiac imaging and matching genetic variants to health-related traits to better understand genetic risks for disease.

We know that AI has even exceeded the performance of medical experts in some cases when interpreting cardiac imaging. It can detect subtle, non-visually detected signatures in data to make a more accurate diagnosis.

How do you use AI in your work at Massachusetts General Hospital?

My focus is to try to improve in-hospital and out-of-hospital patient monitoring.

Regarding in-hospital monitoring, we want to integrate all patient-related information available in order to make the best possible decisions with these patients. That means being able to predict and prevent adverse events.

We also aim to use AI to help determine the optimal time to discharge a patient. We don't want to keep patients longer than they should be staying or discharge them faster than they should. AI can help make those decisions. It will integrate information obtained from different sources that is dynamically changing during a patient's hospital stay, such as electrocardiograms, blood pressure and electronic health records.

Since we can't have a doctor next to the patient all the time, AI can collect all this data and integrate it, see if certain parameters are being exceeded. Then it can make a recommendation, perhaps that patient X should stay one extra day.

With respect to out-of-hospital monitoring, we know many, many people use wearable devices such as smartwatches that collect data, such as whether someone is having an irregular heart rhythm. Many devices are not medical grade, so we can't fully trust these data yet, but we're heading toward better technologies.

We want to use the data these devices collect to make decisions about when a patient should be coming to the hospital. We don't want to bring someone to the hospital for a checkup without a reason, but we also don't want to bring them too late when their condition has worsened. AI can help us make better decisions about when they should come in, which could not only improve care but also reduce costs, which is another significant benefit.

How does AI account for non-measurable data, such as how a person feels?

We do need to recognize and integrate how the patient is feeling – their symptoms – into this process. We need more than a person's vital signs and how those may be fluctuating.

But there are apps that can record this by asking questions. How do you feel during or after exercise? How is your mental state? Do you have an appetite? It can be a questionnaire and the person can log into the app every day to report how they are feeling. All these states of being can be integrated with the other data we collect.

Do you use AI outside of work?

AI has entered every aspect of our lives, whether we know it or not.

It is used in navigation systems in the car to help us assess traffic and choose the optimal route to get somewhere. It is used to determine whether an email is spam. It is used in social media to direct our interests to certain topics. We may not pay attention to it or understand how it's being used, but it has entered our lives to stay.

The foundational principles behind AI are not new. They have been taught for many years back, but they were not called AI or machine learning. The AI research and development today is enabled by an explosion in the volume of data we collect, as well as the vastly increased computational power that is available to process these data.

When a new technology appears, I may not use it directly in my work, but I will experiment with it. For example, I'm taking a paper I've written and I want to see how well a new application of AI can understand it, whether it can generate an accurate summary of what I wrote.

I've also had discussions with my son, Alkinoos, who is a 22-year-old, second-year medical student, about a project he's doing to compare a physician's opinion in making a diagnosis to how well AI could make a correct diagnosis using the patient's symptoms, physical exam and diagnostic and lab tests. Even more importantly, he wants to know whether AI could guide the process of decision-making for a physician. Could AI help improve the prognosis or management of a case?

How might AI be used in medicine in the future?

The future is the integration of all this data. We want to take all these disparate aspects of clinical knowledge, genomics, data taken from wearable devices and electronic health records and our medical history and integrate it into systems that can make optimal predictions about a person's health.

I'm fascinated by the integration of data for multi-organ systems research – that is, how medicine is starting to be addressed holistically, not focusing on a single organ problem but rather the body as a whole. For example, how a person's mental state is linked to cardiovascular disease.

Each one of us is different and we need that personalized care to improve outcomes. This is the hope of AI in medicine, the transition one day to providing equitable care and improving outcomes on an individualized basis.