When someone suffers chest pain and calls 9-1-1 for a possible heart attack, medics do an electrocardiogram—EKG for short—on the spot to examine the heart’s rhythms and determine its condition.
However, diagnosing a minor heart attack from the EKG with the human eye is difficult; it’s main meant for major cardiac events. More subtle heart attacks, which account for nearly 70% of all cases, require more testing and prolonged observation, which usually requires hospital admission for observation. Some cases of heart attacks take several hours to determine.
“Patients go through tons of testing, blood draws, scans and stress tests. It’s frustrating for the patients, who may have to stay in the hospital for about two days,” said Salah Al-Zaiti, associate professor of nursing at the University of Pittsburgh, focusing on computational modeling of ECG signaling in heart disease. “Sometimes, doctors may not find anything, or on the other side, the patients will have to go to the cath lab (cardiac catheterization lab) after those two days of testing.”
New work from his team, published today in Nature Communications, aims to help emergency personnel identify heart attacks earlier and without so much testing.
Al-Zaiti and other members of a multidisciplinary team at Pitt recently developed an artificial intelligence system based exclusively on EKG data from previous medical events that can help clinicians identify 37% more heart attacks during initial screening in real time.
“This is considered a gigantic increase in accuracy,” Al-Zaiti said. “Once built, this system would speed up treatment for those who need it and people won’t have to undergo unnecessary tests or scans. This in turn would save on resources for hospitals and costs for patients.”
For now, the data used to power the AI is stored on a computer system. But in the future, Al-Zaiti said the research team will look at devices to make the data easy for emergency personnel to use.
“Our research aims to use computer models to identify heart attacks that are not traditionally identified using an EKG,” said Christian Martin-Gill, associate professor of emergency medicine in Pitt’s School of Medicine, who specializes in prehospital care and is part of the research team. “We may be able to alter the path in which we take care of patients with heart attacks and avoid hospitalization or prolonged treatment in others.”
Al-Zaiti said the next phase of the research will be building the technology for workflow purposes. Since joining Pitt in 2013, he has served as a key member of scientific investigative teams on 10 research projects contributing to the development and implementation of research protocols, processing EKG data streams and applying advanced machine-learning techniques to address unmet clinical needs.
Other researchers on the project include Clifton Callaway, executive vice-chair and professor of emergency medicine; Samir Saba, chief of Pitt’s Division of Cardiology and co-director of the UPMC Heart and Vascular Institute; and Ervin Sejdic, associate professor of electrical and computer engineering, intelligent systems and bioengineering at Pitt’s Swanson School of Engineering.
The project is funded by a federal grant through the National Heart, Lung, and Blood Institute. It has received a patent through US Patent and Trademark Office.