Health Tech

Google exec at Mayo Clinic conference: AI, ML ‘is math, not magic’

Michael Howell, chief clinical officer of Google, discussed the ways AI and ML are serving healthcare, as well as some lessons he learned since he started working at Google.

Healthcare is extremely complex, but artificial intelligence and machine learning are advancing in ways that can help consumers, caregivers and the community as a whole, according to Michael Howell, chief clinical officer at Google.

Speaking Wednesday at the Mayo Clinic Platform Conference, Howell discussed the ways AI and ML are serving healthcare, as well as some lessons he learned since he started working at Google. A majority of his experience has been as a pulmonary and critical care physician and health system leader, but he joined Google about five years ago.

“I’ve learned some things about AI and ML since working at Google that I didn’t know before,” he said.

His first piece of advice to attendees of the conference is that AI and ML is “math, not magic.” It requires skepticism about how it’s going to impact healthcare, and it still has a ways to go, he said. Like other forms of medical care, it requires publishing peer-reviewed research.

“The math is harder and we want to get it right and we think peer review helps … We want to show our work, and we want to show our work over time because this is a really important tool,” Howell said. “It’s part of how we understand how technologies like AI and ML enter healthcare.”

Another important lesson is that ML learns from examples, unlike traditional programming that follows explicit rules, such as “if-then” statements. An example Howell gave is email spam. A traditional program would have code that states specifically what kind of emails need to go into the spam file. An ML program finds examples of what emails are not being opened and constantly adjusts to reduce errors.

“That changes the kinds of things that we can do,” he said.

How AI is helping today

Howell identified three population buckets that AI and ML can help in healthcare today: consumers, caregivers and the community.

When it comes to consumers, the goal is to help on a day-to-day basis. An example he gave was Google Fit, which allows people to check their heart rate and respiratory rate using their phone’s camera. 

To measure their respiratory rate using the app, consumers just need to put their head and upper torso in view of their front-facing camera and breathe normally. To measure their heart rate, consumers need to place their finger on the rear-facing lens.

“We want to help people’s daily lives,” he said. “Not just in the doctor’s office, not only in the hospital, but in their daily lives.”

When patients are in a healthcare setting, AI and ML can help reduce medical burden as well, whether it be physicians, nurses, payers or anyone else in the field, he said.  

For example, Google is working with Mayo Clinic to streamline the process for radiotherapy treatment. When planning radiotherapy, clinicians go through a step called “contouring,” where they identify the areas of cancer and the nearby healthy tissues that can be damaged by radiation. This process can take up to seven hours per patient.

Google’s partnership with Mayo Clinic includes researching ways to use AI to reduce treatment time and assist contouring.

Lastly, Google is working to use AI and ML to improve community health, and introduced several features amid the Covid-19 pandemic that can help with infectious diseases.

For Covid-19, Google introduced several features to improve research, including a dataset of search trends on disease symptoms to help researchers understand the spread of the virus and how it affected communities. Google also introduced tools to help with vaccination access. Its Covid-19 Vaccine Access Dataset helps find locations where it is difficult for people to access vaccination sites and where pop-up clinics or new sites are needed.

“We’ve really seen a great uptick in research and the public health community in using these kinds of aggregated anonymized datasets to further the research efforts around Covid-19,” he said.

Howell also provided insight on some areas Google is currently working on that may be discussed at the next Mayo Clinic Platform Conference. One of these is an AI model called MUM, designed to identify people in crisis and connect them with the best resources when they are searching the internet for help.

“I think a year from now … we’ll be talking more about these new kinds of technologies that are just coming around the corner,” he said.

Photo: ipopba, Getty Images