AI in Healthcare
The use of AI in healthcare conjures images for many of opportunity but also uncertainty? How is data being collected? How is it being used? Did I give permission? And, maybe the biggest of all, will AI replace my provider?
Three experts gathered to present and answer attendee questions at this month’s Healthcare.mn event, April 10, hosted by Fredrikson & Byron.
Where Will Deep Learning Lead?
Optum Fellow Kerrie Hollie presented first discussing the degrees of AI, weak and strong. Within the Optum Machine Learning Center of Excellence, researchers are leveraging algorithms and deep learning to create micro services. According to Kerrie, algorithms are allowing researchers to attack the big data problem in ways they haven’t before.
Some of this work has been inspired by activity in the human brain. Kerrie noted facial recognition represents an enormous area of opportunity.
Data sets are large and will only continue to grow. Current studies have demonstrated that data garnered from stacking nonlinear processing units and their subsequent extractions (deep learning) often yields better results than machine learning, its larger category.
Replacement or Diagnostic Aid?
John Fraser, Co-Founder and CEO of Treatment.com, answered the question top-of-mind for many. Will AI replace providers? For John, the answer is, “No.” Instead, the AI will serve as a diagnostic tool. In the current model of medicine, John discussed the limitations presented by quick visits — read: five to ten minutes with a patient. Doctors are unable to absorb all of the historical and contextual information captured in records. John sees AI supporting healthcare practitioners by enabling the medical industry to build a new, more personalized diagnostic norm that takes further advantage of contextual patient information provided by past and present data.
John echoed Kerrie’s assertion that the data generated by care consumers is rapidly advancing. Care consumers are actively generating data with wearables and fitness, sleep and biometric trackers. For John, all this data adds up to a very specific picture of each patient’s situation. With the help of AI, practitioners can take steps to mobilize that data and transform it from a curiosity to a critical factor in care decisions.
The Merlin (AI) engine built by the Treatment.com team will help providers access the maybe 15 pages of medical reporting each patient brings with them to an appoint, while humanizing the interactions. If providers are less focused on scanning for the minute details AI can provide from records, they can focus on improving the quality of interactions with patients and reading between data to establish complex diagnoses.
For John, Melin won’t replace doctors, but it will bring some of the background, diagnostic support needed forward to augment not antagonize, the already successful practices of providers.
Can We Use a Smartphone to Predict Patient Behavior?
According to Dr. Julian Wolfson, Assistant Professor at the University of Minnesota Division of Biostatistics, the answer is yes. He hopes to lead the charge to predict human behavior and identify patterns from the qualitative data users are already collecting via smart phones. Via the app Daynamica, the team is working on, Julian hopes to be able to assemble a “picture of someone’s day.”
Julian echoed the first two speakers, noting an explosion of sensor information will help make this possible. The app works, in part, by using the smartphone’s accelerometer to detect changes in trips, activities or mode of transportation.
The hope, on day, is that Daynamica will be able to help characterize activity patterns associated with health outcomes and play a role in delivering behavioral interruptions that can encourage users to maintain solid health habits.
The key to it all? Collect all this information about the user’s “quantified self” without draining the phone’s battery. The patent-pending, Android app is currently deployed in several ongoing research studies.
Myths About AI
For John, the myth of the singularity, where AI advances to become smarter and control humans, is one myth that will remain rooted in science fiction. Julian dispelled the idea that AI algorithms are applicable and ready for use “off-the-shelf.” Rather, scientists must carry out much sanitization and organization of the data.