.svg){kind=small}
Imagine a frightening at-home medical situation: You wake up with chest pain, heart palpitations and shortness of breath. You can barely get out of bed, and worry something is terribly wrong — but the world is in the midst of the COVID-19 pandemic and you’re not keen on going to an urgent care facility for risk of virus exposure. You look at your Apple Watch and see that your heart is racing.
Here’s where machine learning comes in. Apple Watch’s ECG (electrocardiograph) app can guide you to see if your heart is exhibiting symptoms consistent with atrial fibrillation (Afib). Starting in June, your doctor can now use the ECG app as a remote detection device. Instead of going to urgent care, you perform your own ECG at home and send a PDF with the results to your doctor.
Artificial intelligence (AI) and machine learning have not only entered the healthcare space, these technologies are as close to a patient’s side as a nurse or doctor. Let’s not suggest robots are going to learn bedside manner anytime soon (it will take time for people to get used to that). However, just as machine learning complements human skills in other industries, it has the potential to release healthcare workers from mountains of paperwork so they can spend more time empathizing, listening, and healing.
When you hear about AI and machine learning in healthcare, you may recall algorithms that can identify malignancy in skin cancer, or those that match likely drug solutions to disease. These are fascinating and hopeful technologies that can help patients take action long before they may become sick. The promise of applying this technology to patient care is also promising as devices and intelligent services make disease detection devices available to the patient, for the doctor to review remotely, or in the context of a wider scope of patient medical history.
This can have far reaching effects for doctors caring for people in remote countries where the closest X-ray machine is hundreds of miles away, and for patients quarantined in any country today. Machine learning’s application to healthcare is exposing more people to the benefits of disease detection in much more accessible ways.
The pandemic’s push behind these technologies is undeniable, as seen with the Apple Watch ECG app. And that could bode well for the companies that have been recently creating devices and technologies that use machine learning for more personalized care.
The FDA green lighted the Apple Watch to stand in for in-clinic electrocardiographs (ECGs) during public health emergencies, such as the COVID-19 pandemic. Clinicians can now use the watch as a remote monitoring ECG device so wearers can get a preliminary ECG diagnosis, which can be one of the first tests done to see if someone is having, or has had a heart attack, or is at risk of a stroke.
Fitbit is currently conducting an Afib study of their own to evaluate an algorithm that analyzes data from users’ devices. Like the Apple Watch, if heart activity suggestive of atrial fibrillation and the algorithm match, Fitbit’s app will notify the wearer and recommend a Telehealth consultation for a full evaluation. And, much like Apple Watch wearers, they would receive an ECG patch in the mail to wear and send back to a lab for further evaluation.
If the wearer believes they are having a medical emergency, they are urged to seek immediate medical attention, and the watch is only using a single “lead,” instead of the 12 leads (creating more closed electrical loops) that a doctor would use. But it’s a step in the right direction when it comes to detecting risk factors in some patients.
"It's like night and day how much more information I get and how I'm able to manage their atrial fibrillation without bringing them into my office or an emergency room or putting expensive monitors on them," said Dr. Anthony Pearson, a Missouri-based board-certified cardiologist. "It's dramatic how improved my care is with these devices."
His patients use Kardia Mobile, another single-lead device that connects to an iPhone. Combined with Kardia Pro, a cloud-based monitoring platform, he’s eliminating short- and long-term cardiac monitors.
Wearable devices aren’t just informing the health-conscious about calories burned and mid-workout heart rates. They’re standing in for diagnostic procedures that, not long ago, could only be performed by trained medical professionals.
“Doctors don’t have the time to review all of the data about a patient — sometimes it’s not available, sometimes it’s located in another health system,” said Eric Topol, a cardiologist, geneticist, and digital medicine researcher is author of Deep Medicine: How Artificial Intelligence can Make Healthcare Human Again. He spoke at a recent discussion at the Aspen Institute. He calls this modern symptom of data-entrenched healthcare workers “shallow medicine.” “Practitioners have almost become data clerks,” he said.
Artificial intelligence could save time and allow medical professionals to spend more of their days fulfilling their healing roles rather than doing paperwork. Doctors could let the machines do the data sifting and synthesis and allow humans to listen, express empathy and return to the caring time spent with patients. Machine learning in healthcare has the potential to restore face to face contact, Topol said.
Managing diabetes can be challenging on a good day, and life-threatening if something goes wrong. It’s not just monitoring blood sugar levels, which require consistent and careful attention. People living with diabetes must also eat healthy, exercise, reduce risks, totaling hundreds of decisions each day. Medtronic, maker of diabetes management devices, teamed up with the AI team at IBM Watson to create a platform that could relieve patients from making so many decisions, which Medtronic estimates could comprise up to 95% of a diabetes patient’s annual care.
The app, called Sugar.IQ, is still under development, but is designed to identify trends that affect glucose levels so patients can be warned when it appears they are heading for a blood sugar crash, or other situation that they could manage on their own. And when patients manage their blood sugar on their own more consistently and for longer, that can make a big difference when considering their risk of other diseases associated with diabetes, such as heart or kidney disease.
“Any big problem in healthcare is an amalgam of smaller, addressable problems. By solving the day-to-day problems, we can make positive improvements in patient outcomes while advancing toward our ultimate goals,” said Vice President of Customer Solutions at Medtronic Diabetes, Paul Acito.
Personalized care means the world to people as they age and want to stay in their homes, even if their family members worry about them living alone. GeckoSystems’ CareBot is a response to this concern. Powered by up to 12 AI-powered engines, the CareBot, a cross between a loving, apron-wearing caregiver and, well, a robot, can detect sound, heat, movement and colors. It’s designed to check blood pressure and monitor vitals, as well as remind someone to turn off the TV, take medications or walk the dog. Family members can also chat through CareBot’s video conference interface — all in an effort to reduce elderly isolation.
For every hour a doctor spends with a patient, they are estimated to spend two hours on paperwork. Machine learning in healthcare has massive potential to refocus the attention of clinicians and doctors back to patients. Machine learning can help with interpreting the mountains of data that accompany patient records. Natural language processing, if leveraged in the doctor’s office, could allow more of the stories of symptoms to shed light on a patient’s health. AI-powered sensors that use images, or listen to voices could detect depressions or disease — especially helpful if a patient is not as forthcoming with symptoms or cannot speak.
Bringing humans back to healthcare is an inspiring ideal, and one that can be made possible by combining the empathy and incredible skills possessed by humans, with the massive data processing capabilities of machine learning.
