Advancing the Science

Mayo Clinic Medical Science Blog – an eclectic collection of research- and research education-related stories: feature stories, mini news bites, learning opportunities, profiles and more from Mayo Clinic.

August 13, 2019

Inspired to innovate: Detecting heart disease earlier with the help of artificial intelligence

By Advancing the Science contributor

By Nicole Sisk

Paul Friedman, M.D., explains some of the AI work in to colleagues in the Department of Cardiovascular Medicine.

Artificial intelligence will play a key role in Mayo Clinic's future and that of health care worldwide. The Department of Cardiovascular Medicine is combining artificial intelligence with standard tests to improve patient care.

Each year at Mayo Clinic, 250,000 patients have an electrocardiogram, or ECG. It's a common, inexpensive test designed to detect heart rhythm and other abnormalities.

But what if that common, inexpensive test could reveal more? Could the humble ECG, developed more than 100 years ago and readily available nearly everywhere, uncover deeper insights about a patient's heart and health?

Physicians and scientists in the Department of Cardiovascular Medicine wanted to find out. They believed that the ECG and other data-rich exams held the potential to provide a wealth of information that could open up new ways of diagnosing and treating heart conditions. And those insights, the team suspected, were hiding in plain sight.

It would just take a new way of seeing to find them.

Innovation through collaboration

That new view began coming into focus when Itzhak Zachi Attia joined the Department of Cardiovascular Medicine. Attia isn't a cardiologist. But with degrees in electrical and electronics engineering, and expertise in machine learning, he had the perfect pedigree for a department eager to explore the potential of data science and artificial intelligence.

Itzhak Attia

"We have a long history at Mayo Clinic of sharing ideas across specialties," says Paul Friedman, M.D., chair of the Department of Cardiovascular Medicine in Rochester. "Adding a new group of specialists to the mix — AI engineers — is building on that tradition. They see the work we do with fresh eyes and ask questions that help us reexamine how and why we do things."

Attia and his colleagues — there are now five data scientists working in Cardiovascular Medicine — attend rounds, observe procedures and have office space alongside clinicians. "We have the ability to share ideas on a daily basis," Attia says. "We hear medical conversations and share new things that are happening in tech."

That's led to some fruitful cross-pollination. Attia, Dr. Friedman and others in the department have been exploring the possibilities created by merging their unique perspectives.

"The ideas started from sitting and brainstorming how we could use the huge amount of data that we have to help patients get better care and a faster diagnosis," Attia says. "I think being embedded in Cardiology is one of the key elements of our success."

Beyond the basic ECG

The close collaboration between clinicians and data scientists has revealed that there is much more information that can be gleaned from an ECG than once believed. Applying artificial intelligence to the test has given the team the ability to gauge high potassium levels and detect previously "invisible" long QT syndrome, a heart rhythm condition that can potentially cause fast, chaotic heartbeats that may lead to fainting, seizures or sudden death.

"The old paradigm has been that you wait until you feel sick before you see a doctor and have diagnostic tests," Dr. Friedman says. "But sometimes that's too late. In cardiovascular medicine, the first sign that something is wrong may be a heart attack or stroke. But the conditions leading up to that may be going on for decades."

The old paradigm has been that you wait until you feel sick before you see a doctor and have diagnostic tests. But sometimes that's too late.

Paul Friedman, M.D.

"Our bodies are giving off invisible signals all the time. We're finding ways to use technology to pick up those signals," he says. "The goal is to detect problems early so we can intervene sooner and prevent bad things from happening."

Perhaps the most promising finding to emerge from the group's work so far is a method for detecting a weak heart pump long before a patient experiences any symptoms. By applying artificial intelligence to ECG data, the team can pick up signals indicating a patient has asymptomatic left ventricular dysfunction, or a weak heart pump. Left untreated, the condition can progress to heart failure. Results of the study were published in Nature Medicine earlier this year. (Read related news release.)

Putting artificial intelligence into practice

The discovery could have a significant impact on patient care. Asymptomatic left ventricular dysfunction affects more than 7 million Americans — 2% of the population and up to 9% of people 60 and older. Early diagnosis can lead to effective treatment that reduces the likelihood of heart failure. But until now, there hasn't been an affordable, noninvasive screening test for the condition. The team's findings could change that.

Peter Noseworthy, M.D.

"We hope this will be in clinical use by the end of the year," says Peter Noseworthy, M.D., a cardiac electrophysiologist who has been involved with the study.

Before that can happen, Dr. Noseworthy and his colleagues will review the results of a randomized clinical trial set to launch soon throughout Mayo Clinic's community practices in Minnesota and Wisconsin. The algorithm to detect asymptomatic left ventricular dysfunction has been embedded in Mayo's ECG analysis, and half of the eligible patients who screen positive for the condition will have a report generated in their health record. The study will examine what clinicians do with the information provided.

"We're now going to see how humans and machines come together," Dr. Noseworthy says.

It's a new question — and not one Dr. Noseworthy anticipated asking as recently as two years ago. "I didn't envision AI being part of my work," he says. The new focus has been a welcome — and exciting — surprise.

Ensuring smart, equal data

That excitement is understandable. Artificial intelligence has the potential to transform health care in countless ways. But its potential depends on the human intelligence that informs it.

"You have to pick the right question and make sure you have the data to support it," Dr. Noseworthy says. "You want to avoid the problem of garbage in, garbage out by doing the work beforehand to make sure you're setting your study up right."

Sharonne Hayes, M.D.

That includes determining whether artificial intelligence algorithms apply equally to all people.

"AI is not biased, but we could potentially create biases as humans who input bad data," says Sharonne Hayes, M.D., a preventive cardiologist and director of the Office of Diversity and Inclusion. "As Mayo is moving forward and working to be a leader in AI, we want to make sure that we're not perpetuating stereotypes or delivering unequal care through our work."

Dr. Hayes notes that medical research has a long history of focusing solely on the majority, often white men. That lack of diversity has contributed to health care disparities. Artificial intelligence, she says, "has the potential to make things better — if we are smart about using it."

The great thing about this study is that people with a diversity of expertise were brought in to be part of the conversation. It's how science should work.

Sharonne Hayes, M.D.
LaPrincess Brewer, M.D.

Dr. Hayes and LaPrincess Brewer, M.D., another preventive cardiologist with expertise in health disparities research, were invited to review the artificial intelligence algorithm used in the ECG study. They determined that the algorithm worked equally well for women and men, and across racial groups.

"In other AI algorithms, we might find one that doesn't work for different groups or works differently for different types of patients," Dr. Hayes says. "The great thing about this study is that people with a diversity of expertise were brought in to be part of the conversation. It's how science should work."

Dr. Hayes says she can't imagine a study like this happening anywhere other than Mayo Clinic.

"In addition to brilliant colleagues doing the work, the fact that we have access to Mayo's data resources is significant," she says. "Other places are doing AI, but no one else has access to patient populations and records that we do."

Attia agrees that Mayo Clinic has a unique advantage in the field of artificial intelligence.

"In industry, the product is the main thing. In academia, research is the main thing," he says. "But here at Mayo, we are finding the middle ground and using solid research to create products. We're eager to take things to practice and translate research to actual tools for helping patients. I think the fact that we all have the same goal — the needs of the patients come first — really helps focus our effects."


Related resources

Meet our researchers and check out the research and education that after 150 years, continue to provide the underpinning for the first and largest integrated, not-for-profit medical group practice.

Mayo Clinic podcasts and other discussions of research and practical patient information can be found online.

Information about many of the clinical trials available across Mayo Clinic is online as well.

Much of our content is available in Spanish, and we also have news and patient resources in PortugueseMandarin Chinese and Arabic.

Tags: artificial intelligence, cardiology, cardiovascular medicine, Center for the Science of Health Care Delivery, diversity, health disparities, heart disease, Innovations, Itzhak Zachi Attia, LaPrincess Brewer, News, Paul Friedman, Peter Noseworthy, Sharonne Hayes

Please sign in or register to post a reply.
Contact Us · Privacy Policy