How AI — and ASU — will advance the health care sector

Editor's note: This feature article is part of our “AI is everywhere ... now what?” special project exploring the potential (and potential pitfalls) of artificial intelligence in our lives. Explore more topics and takes on the project page.

A patient walks into a hospital complaining of stomach pain.

After a scan is performed, the doctor confirms the diagnosis of appendicitis, surgery is scheduled and the patient leaves the hospital the following day, his appendix out and his pain gone.

It’s a situation that happens every day, in hospitals across the world. But what if, using an artificial intelligence learning model, that scan — along with past medical history, lifestyle choices and other relevant data — also showed the patient is at risk for a heart attack or stroke?

How many lives might be saved?

“AI can see things that humans cannot,” said Dr. Bhavik Patel, the chief AI officer at Mayo Clinic in Arizona. “AI can take that scan and tell us what your heart risk and stroke factors are even though we didn’t image the heart or brain. AI can help us impact things to where we’re making a difference in people’s lives.”

It would be a misnomer to say AI is the next frontier in health care. AI has been part of medical care for more than 20 years, used to analyze medical imaging data such as X-rays or MRIs, transcribe medical documents and streamline administrative tasks.

But as AI’s technology improves, it has the potential to process information faster, thus accelerating diagnoses and enabling earlier and possibly life-saving treatment for patients.

Imagine:

• Algorithms that spot malignant tumors. 

• A smartphone app that can alert a caregiver before someone falls.

• Learning models that recognize changes in speech patterns, which could indicate neurological conditions.

The possibilities, medical professionals say, are endless.

“I think AI represents the fourth industrial revolution, and health care is just one of the many verticals it’s revolutionizing,” Patel said. “You shouldn’t say health care anymore without saying AI. It should be an integral part of it.”

Sherine Gabriel, executive vice president of ASU Health, stressed that AI will not replace health care professionals or dramatically alter the doctor-patient relationship.

Instead, Gabriel said, AI is a tool that will ultimately result in better patient care.

“It’s not going to make the diagnosis for you, but it’s going to make the diagnosis easier, simpler and way quicker,” Gabriel said. “It’s going to identify diagnostic targets more quickly than we have been able to in the past. The thinking isn’t really that different. It’s just putting everything on overdrive.”

One example:

Bradley Greger, an associate professor in ASU’s School of Biological and Health Systems Engineering, part of the Ira A. Fulton Schools of Engineering, is using AI and machine learning to read and analyze signals from the brain.

The technology could have multiple applications: Enabling paralyzed people to control a robotic limb; helping blind people to see with the aid of a camera connected to the visual processing areas of the brain; aiding patients with seizure disorder.

“There’s unhealthy electrical activity in the brain that causes seizures. It’s very complicated datasets,” Greger said. “We can use AI to look for patterns, and then we can tell people when they’re going to have a seizure or they need to have more medication or something like that.

“AI is really good at searching through very complex datasets and looking for patterns that correlate or give you some knowledge about what’s happening with the pathology itself. A human can do it, but it just takes a human a long time.”

AI can also help with administrative tasks, thus allowing doctors to have more face time with their patients.

DeepScribe, an AI-powered medical scribe, uses machine learning and language processing to extract medical information from a phone conversation between a provider and patient and almost immediately produce a finished medical note that goes into a patient’s record. Typically, experts say, doctors spend more than three hours a day documenting conversations with their patients.

“AI is a really great tool for doctors to have at their disposal,” Greger said.

AI will be a central tenet of ASU Health, which includes a new medical school called the School of Medicine and Advanced Medical Engineering, the School of Technology for Public Health, the Health Observatory at ASU and the Medical Master’s Institute — as well as the existing College of Health Solutions and Edson College of Nursing and Health Innovation.

“AI will be used in almost every conceivable way imaginable,” Gabriel said. “It’s going to be interwoven into everything we teach. We want to be certain that the next generation of providers have all of those skills at hand in order to optimally improve health care.”

Gabriel said ASU Health will approach AI with a humanistic perspective, thus the creation of the AI + Center of Patient Stories. ASU Health will use reporters and students in the Walter Cronkite School of Journalism and Mass Communication to engage with patients and write their stories.

The center will then use AI applications like virtual reality, augmented reality or real-time mixed reality to digitally enhance the stories.

“We’re trying to create not just a video for them to watch, but an immersive educational experience for students to help build empathy, help build understanding and help them build the connection I think they need to really be able to care for those patients and populations,” Gabriel said. “We really want to use technology to enhance the humanistic aspects of health care.”

ASU Health also will create an AI + Medical Suite of the Future. Gabriel described it as a patient care setting where AI will be used to create information that is packaged in a way that patients or their loved ones can access it and “understand what’s going on in a much deeper level.” The suite, Gabriel said, could include an “AI coach” that can answer questions.

“There are lots and lots of studies that show a patient will have a conversation in a doctor’s office, but even if that conversation is taped, the patient will only retain a small part of it, either because the doctor didn’t explain it well or it’s an emotionally charged situation,” Gabriel said. “Or they’ll go home, and a loved one will say, ‘So, what did the doctor say?’

“That happens every day of the week. But these (AI) tools give us the opportunity to change that structure. It’s about bringing all of the curated information together from that patient encounter in a way that optimizes their health outcome.”

Across ASU, faculty is working with AI to improve patient care.

Thurmon Lockhart, a professor in the School of Biological and Health Systems Engineering, has developed a wearable device that goes across a patient’s sternum and measures body posture as well as arm and leg movements in real time.

“Traditional fall-risk assessments for seniors don’t always target specific types of risk, like muscle weakness or gait stability,” Lockhart said in a previously published ASU Thrive story.

When the risk of falling is deemed high, a smartphone app called the Lockhart Monitor can alert the user or caregiver.

Visar Berisha, the associate dean of research and commercialization in the Ira A. Fulton Schools of Engineering, is working with Julie Liss, an associate dean and professor in the College of Health Solutions, to develop AI models that analyze a person’s words and speech patterns.

Those patterns, Berisha said, can help determine whether a patient may suffer from neurological conditions like Alzheimer’s, Parkinson’s or amyotrophic lateral sclerosis, also known as Lou Gehrig's disease.

“A lot of that analysis has been done manually in the sense that patients would come to a lab, record samples, and then clinicians would listen for different types of changes in the speed signal that were indicative of the underlying condition,” Berisha said. “You needed to be an expert, have lots of training and a lot of time.”

Berisha said he and Liss have been working for more than a decade on an AI model that would automate the process by having patients download a mobile app on their personal devices and then provide speech samples for analysis.

“It scales it in a way that wasn’t possible before,” he said.

In addition, Berisha said, AI might be able to notice subtle changes in a person’s speech pattern that a clinician might not spot.

Patel said the AI “revolution” in health care isn’t something that will happen in five, 10 or even 20 years.

“Just because of the way the technology has improved, I think we’re very near to getting those exponential benefits,” he said.

Greger predicted AI will be omnipresent in the medical field.

“It’s going to be everywhere,” he said. “It’s just going to be this thing around that everybody uses. And it will have very specific applications. The radiologists will use it for image processing. The neurologists will use it to process the signals that come out of the brain and help them identify diseases. It’ll be a tool that’s very widely used. Just like the computer or car is today.”

Not all health care experts are as certain of AI’s immediate impact, though. 

“I don’t know that the rate of change is going to be very fast,” Berisha said. “There are some structural complexities in the American health care system that make it really difficult to introduce new tools in everyday care.”

A 2019 paper in the National Library of Medicine amplified those difficulties. The paper noted that health care decisions have been made almost exclusively by humans, and the use of smart machines could raise issues of accountability, permission, transparency and privacy.

The paper, titled “The Potential for Artificial Intelligence in Healthcare,” also questioned whether health care providers will be able to adequately explain to a patient how deep learning algorithms led to, say, a cancer diagnosis. In turn, that could impact the doctor-patient relationship.

There’s also a question of accuracy within AI diagnoses, according to the paper. Machine learning systems could be subject to algorithmic bias and predict a greater likelihood of disease on the basis of gender or race when those aren’t causal factors.

Berisha said health care professionals may be reluctant to turn to AI for simpler reasons: time and money. He said overburdened care providers who only have “seven minutes to spend” with a patient may decide they don’t have the time to evaluate information they’re not certain will help with diagnosis or treatment.

In addition, he said, the cost of AI, plus the hundreds of applications that will be available, may dissuade providers.

“I agree it’s going to be transformational,” he said. “I just think it’ll take a while.”