AI stroke care
There isn't a minute to waste when it comes to treating strokes, and while good treatments exist, they often come too late, leaving with lifelong disabilities or sometimes dead.
Stroke is a leading cause of serious long-term disability and the fifth leading cause of death in the United States, killing about 142,000 people a year, according to the American Heart Association. But research suggests those grim statistics can improve if more patients receive proper diagnosis and intervention sooner - two new technologies developed at Erlanger Health System in Chattanooga promise to do just that.
"This has a lot of ramifications for the city, not only for the health care of people in our region, but from a business development standpoint," said Dr. Tom Devlin, director of Erlanger's Neuroscience Institute and chairman of neurology at the University of Tennessee College of Medicine Chattanooga.
Strokes occur when either a blood clot or hemorrhage interrupts the brain's blood supply. Unlike muscles, which are able to store energy, the brain relies second to second on blood flow carrying oxygen and sugar to keep cells alive.
Clots cause more than 80% of strokes, which result in about 1.9 million brain cells lost every minute the clot isn't removed, Devlin said.
Smaller blood clots can be removed using clot-busting drugs. But the most dangerous strokes caused by large blood clots in the head or neck require a special procedure - called a thrombectomy - to mechanically remove the clot.
Patients needing a thrombectomy must be taken to a comprehensive stroke center where specially-trained physicians capable of performing the procedure are on call 24-7. For the Chattanooga region, that's the Erlanger main campus.
Devlin said what often happens is patients wait for hours at an outside hospital before being diagnosed and their brain is too damaged by the time they arrive.
"Better diagnostic tools are needed to rapidly identify stroke patients, and these new technologies that we're developing right here at Erlanger are going to be very impactful at fixing this problem," he said.
A TESTING HUB
Two California-based tech companies are working with Erlanger to harness the power of artificial intelligence and speed up the stroke treatment process.
One company - Viz.ai Inc. of San Francisco - created a phone application that alerts the stroke team when a stroke patient's head scan reveals a blood clot. The other company - Neural Analytics Inc. of Los Angeles - uses ultrasound from a robotic headset placed outside the skull to non-invasively locate clots.
Stroke prevalence is highest in the Southeast, and Erlanger's stroke center reputation along with its location in what's called the "stroke belt" made it a prime location to test these products, said Henry Hewes, global vice president of commercial at Neural Analytics.
"Their experience with this disease, as well as their access to patients who need the advanced care that's required for stroke management, is obviously a key driver for their success in both evaluating and partnering with companies developing these technologies," Hewes said.
To develop their software, the companies rely on Erlanger's vast database of brain scans, which security and compliance teams protect and de-identify so no patient information is shared.
Neural Analytics taught its robotic headset to recognize blood flow characteristics after Erlanger "painstakingly" collected, analyzed and determined what characteristics of a waveform reveal where a clot is in the brain, Devlin said. It can be used in a hospital setting or by paramedics in an ambulance, eliminating the need for a trained ultrasound technologist.
The Viz.ai software uses a form of artificial intelligence called deep learning, where a mathematical algorithm learned to recognize strokes after being "fed" thousands of images.
"We say, 'It's a stroke and there's a clot in the brain,' but we don't tell it anything about what features in the picture actually make it a stroke," Devlin said. "The more information we give the computer, the smarter it gets at figuring it out."
Now, when a patient enters any Erlanger hospital with stroke symptoms - trouble walking, speaking, understanding, and paralysis or numbness of the face, arm or leg - one or both of the technologies may be used to help determine if there's a clot and whether or not they should be transferred for treatment.
Both technologies are now approved by the federal Food and Drug Administration and being sold commercially across the U.S.
The Viz app has "revolutionized the entire workflow" around stroke care at Erlanger, according to Devlin.
Dr. Justin Calvert, Erlanger's chief of radiology and one of four Erlanger doctors who perform thrombectomies, said the app reviews the brain scan of every suspected stroke patient who enters an Erlanger hospital. It then alerts the stroke team whenever it suspects a clot, allowing them to review images right there and make an immediate clinical decision.
"That's the beauty of Viz," Calvert said. "Now, we have the capability to actually look at the images done at these outside smaller, regional hospitals to determine if the patient is a good candidate for intervention."
This means that patients needing a thrombectomy are fast-tracked to Erlanger, and those who don't aren't necessarily transferred.
It's also easy to navigate and communicate with the nurses, physicians and techs on a stroke team.
"The whole team can communicate with each other quickly and in one place," he said. "It really helps on patient handoffs, making sure that our team is called in and we're talking to each other. Communication is everything when you're taking care of critically ill patients with a time clock in place."
The software doesn't replace clinicians but works as a second set of eyes to get patients the right diagnosis and treatment ASAP, Calvert said, adding that the technology does sometimes "overcall."
"At the end of the day, you rather it call the diagnosis, and you can determine if it's true or not," he said.
Erlanger is working to partner with other regional hospitals so when a hospital takes the scan of the head and neck, those images enter a database that uses artificial intelligence to do a preliminary interpretation of the images looking for a large vessel occlusion, a clot that's lodged in one of the bigger vessels of the head and neck.
Hamilton Medical Center in Dalton, Georgia, will be the first non-Erlanger hospital to roll out the software. The hospital is currently working through the initial steps of implementation, said Hamilton Vice President Ernie Elemento.
"We're always looking at new technologies and ways to improve care for our patients," Elemento said. "Hamilton has had a partnership with Erlanger for high level stroke care for a number of years. This software and arrangement just enhances that partnership and provides faster care for stroke patients who need a high level of specialized care."
Contact staff writer Elizabeth Fite at [email protected] or 423-757-6673.