It was a typical Saturday morning.
Angela Watschke was shuffling some of her kids around four years ago while the others were at home. Her then-two-year-old daughter Abrielle was lying on the bed when she started slipping off as her eyes rolled back. The toddler was experiencing cardiac arrest.
One of Watschke’s daughters texted her asking for help. The color faded from Abrielle as her father performed CPR and called 911. When Watschke arrived home, she found her street filled with first ambulances and paramedics. She and her husband clung to each other until finally someone said Abrielle was breathing on her own.
The local hospital wasn’t sure what to do with the toddler. They sent Abrielle to another emergency room nearby, where doctors informed Watschke her daughter had a heart condition called Long QT Syndrome.
“I had no idea a normal, healthy 2-year-old in literally one moment to the next could go from perfectly fine to heart stop,” said Watschke, who lives near Minneapolis, Minnesota.
Named for the abnormal electrical wave patterns that characterize it, LQTS causes dangerously fast irregular heartbeats in response to stress and exercise, according to the National Institutes of Health. Many people don’t know they have it until it causes them to faint, have seizures or even die.
Artificial intelligence could change that.
AliveCor, a medical device startup, and Mayo Clinic used artificial intelligence to identify LQTS in patients whose EKG results appear normal. Their findings from a study, published in an abstract today at the Heart Rhythm Scientific Sessions conference, showed the technology accurately diagnoses the genetic condition 79 percent of the time.
It could one day help doctors diagnose the condition earlier and more accurately than they currently can. It could also help consumers access tests more easily than they can now.
“I will submit that when the QT interval is caught in our patients early, this will be a life-saving modifier that we will have come upon,” said Dr. Michael Ackerman, Abrielle’s doctor and director of Mayo Clinic’s Genetic Heart Rhythm Clinic and the Windland Smith Rice Sudden Death Genomics Laboratory.
Signs of LQTS can appear in an electrocardiogram, but they’re not always apparent and physicians don’t always recognize it.
About 1 in 7,000 people are estimated to have LQTS, but no one knows for sure because it usually goes undiagnosed, according to the NIH. A paper published in 2009 says it could be even more common at 1 in 2,000 live births. The condition causes about 3,000 to 4,000 sudden deaths in children and young adults in the U.S. every year, according to the NIH.
LTQS can either be inherited or acquired. Neither of Abrielle’s parents carried the condition. One of her genes morphed when she was developing, called a de novo mutation, causing her to be born with it.
AliveCor built a deep neural network using EKG results from more than 1,000 patients with congenital LQTS and more than 1,000 patients without it. The system identified relevant features and continued to learn from the data.
It detected the condition in people where the length of electrical waves measured in an EKG were indistinguishable from normal ones — picking up on signals doctors hadn’t seen. This could help diagnose people earlier and prevent sudden deaths, Ackerman said.
To conduct the study, researchers used a traditional EKG machine but used only one lead, or sensor, instead of the usual 12. That way, they could see whether this type of testing could be added to AliveCor’s EKG devices that use one lead.
LTQS is often the culprit in stories of teen athletes suddenly collapsing and dying during a game. AliveCor CEO Vic Gundotra said he dreams that one day every coach in every high school would have a device that could perform a 30-second EKG on student athletes and screen them for LQTS.
He said it could also be used in hospitals to try and prevent some cases of sudden infant death syndrome. Pharmacists could also one day use it to avoid giving people drugs that may cause them to develop the acquired form of the condition.
“There are so many opportunities if the science is right and we go through the correct regulatory pathways, the applicability of this technology in schools, in homes, in hospitals, in pharmacies, is kind of extraordinary,” said Gundotra, a former executive at Google and Microsoft.
Abrielle’s life would have been different if she had been diagnosed with LQTS before she went into cardiac arrest. She could’ve been treated with medicine instead of having her ribs cracked open and undergoing open heart surgery to have a defibrillator implanted in her.
“It’s hard to even put into words how traumatic that was and all that she experienced in (three weeks) in the hospital,” Watschke said. “Everything from the repeated shock, the needles, the pokes, the prodding, the barrage of medications. Boy, it’s hard to even think about when thinking back and it’s hard to talk about her being in the hospital.”
Mayo Clinic invested an undisclosed amount in AliveCor and partnered with the startup in 2016 to identify hidden health signals displayed in EKGs. The pair started with looking for abnormal potassium levels and announced last summer they would work together to detect LQTS.
The findings shared Wednesday are early signals that AI could help diagnose LQTS, Gundotra said, but it could take years before the technology is introduced in the market. More studies will be required, and any new technology would need to receive approval from the Food and Drug Administration.
Nearly four years since her cardiac arrest, Abrielle remains completely stable. She celebrated her sixth birthday earlier this month.
Watschke reads stories about other lives that are suddenly lost because they weren’t aware they had LQTS. The 41-year-old mother of six, soon to be seven, knows she’s fortunate Abrielle survived and is now thriving.
She also knows she could have prevented the incident if she had known what LQTS was and that her daughter had it.
“People say it’s so cliche to say knowledge is power,” she said. “But knowledge really is power”