Q&A: Unlocking the ‘black box’ of AI technology

DARIEN, Ill. – As recently as four or five years ago, sleep medicine professionals were skeptical about the possibilities of artificial intelligence, but as they become more educated on the technology, they are seeing the possibilities it offers, says Dr. Anuja Bandyopadhyay. 

“Once you start seeing beyond the black box and once people started realizing, ‘OK, this can be beneficial, it’s helping me save time, helping me save resources, improving access to care,’ than it became a natural solution to quite a few of our problems,” said Bandyopadhyay, chair of the Artificial Intelligence in Sleep Medicine Committee for the American Academy of Sleep Medicine and an assistant professor of pediatrics at Indiana University. 

Bandyopadhyay is also the lead author of a new research paper on the use of AI-enabled technology in sleep medicine. 

HME News: Is sleep medicine a good fit for AI technology? 

Dr. Anuja Bandyopadhyay: In sleep medicine, I think we are perfectly positioned because we have access to all of these different body signals – we are getting your brain waves, your oxygen levels – all of that in one page. Sleep is something we do very naturally, but it’s also closely linked to mental health, to your cardiovascular health and it's so easily influenced by your external input, your environment and your habits. There's a huge branch of science which is looking at the psychosocial aspect of it and how we can change your behavior. There’s a lot of hidden treasure within all those data points. 

HME: What is an example of a problem within sleep medicine that AI technology could help solve? 

Bandyopadhyay: I would say the three key health care domains are clinical care, lifestyle management and population health. (For example), if you think about lifestyle management, more people are aware now of their sleep habits (thanks to sleep trackers). It’s not just the number of hours you sleep, but what happened right before that to get a nice clear pattern. Patients are now empowered because they have access to all this data themselves. 

HME: How can AI help with patient compliance? 

Bandyopadhyay: Compliance is a huge problem. In the past 10 years, we’ve had all these patient engagement tools, which have helped. There are studies that show that just looking at your PAP usage history, they can predict which patient is going to stop using it in the next week or so. If I have the technology that can help me handpick the patients who really need me then I’ll get you in for a visit sooner than later.  

HME: The Food and Drug Administration recently approved the Samsung Galaxy Watch for use to detect signs of obstructive sleep apnea. Where do you see wearables fitting into the overall picture? 

Bandyopadhyay: There’s always going to be a difference between consumer grade and clinical grade devices, but for the consumer grade device, it is a great screening tool. It can pick up some subtle difference or change or some non-specific abnormality. So, it’s really alerting you that (maybe) there’s something wrong.