AI and Wearables
Wearables are everywhere and like many technical terms the early entrants have become synonymous and part of the vocabulary
Are you Fitbit enough?
But in the corridors of healthcare systems and doctors’ offices, there’s less excitement about the penetration of these devices and a healthy skepticism for the value of tracking the number of steps you complete each day.
While many of the early devices suffered from a drop off in use with as many as 50% of users losing interest within 6 months the tide appears to be turning thanks to improved features, functions, and form that is creating a surge of data that can be a struggle to process
The early pioneers worked to capture and integrate data from a wide variety of wearables and Chris Dancy was an early pioneer who funneled thousands of color-coded daily data points into his google calendar
Image Courtesy of Chris Dancy
As he described
By the end of 2012 I had enough data and routines in place to start shaving pounds off my body, slow my mind, and stop smoking. I understood that binge watching TV was changing the way I ate food, and that posting on social media after exercising produced more “likes”.
Image Courtesy of Wearable Insider
How do we Tame the Data
But even if we had the time and resources to carry out the data integration and analysis the overwhelming nature of data collection rapidly outpaces the capacity of the human mind to process, associated and find patterns and knowledge.
As the space develops its inevitable to find Artificial Intelligence making inroads – Sensoria Fitness were early to the space of AI-driven app coaches. They produce an extensive line of gear designed to monitor basic biometrics as well as advanced parameters fed from wearables, textile sensors, and conductive fibers. The early use case involved improving running routines using performance analytics
But has expanded to include healthcare specific capture and use of wearable data to help mitigate and treat diabetic foot ulcers, fall detection and gait monitoring in chronic diseases and rehabilitation tools to improve recovery time and track response to therapy.
Kardia (previously known as AliveCor) had early success in creating cost-effective EKG wearables that recorded a single-lead EKG. With a large user base, widespread adoption across clinicians and patients they quickly captured a large volume of EKG data that provided the training data sets and proof points necessary to develop automated AI algorithms for abnormality detection fast-tracking to FDA approval. The Kardia service now offers one of the first FDA cleared algorithms for the detection of atrial fibrillation
Future Data Opportunities Abound
This is only just the beginning as we explore the potential to integrate larger swathes of data as Propeller Health that used machine learning to analyze patient medication data and environmental conditions such as pollen counts and air quality to forecast potential asthma attacks.
Using knowledge and experience learned in other industries that have been applying machine learning to our world we have seen success with Siri, Alexa and Google Now all vying for our attention as these agents learn from our data and behaviors and apply the results to customize our experience and tailor everything for our personal tastes. This year alone, according to some estimates, the wearables data generated by Android Wear, Fitbit, and Apple Watch will likely generate two trillion health measurements and its set to get larger. One thing for sure, the human brain and our clinical professionals will never be able to review all of this data but generating actionable intelligence will be essential if we are to turn the healthcare tide to better, widely accessible and more cost-effective care. Taking insights developed in the search engine a group of engineers have set up Cardiogram that is using AI to develop insights and knowledge from the increasing streams of personal wearable data and create a continuous automated health monitor that not only helps you track your activities but will offer guidance that won’t just identify disease as it occurs but predict it before it impacts you.
The single data point of blood pressure measured in the highly stressful environment of a hospital or doctor’s office will seem quaint as we look back with the benefit of hindsight. New continuous data streams of patient data will add to the trillions of data points being captured and flow directly into deep learning systems that consume this raw data with the goal of developing correlations and ultimately causations of disease linked to our interaction with our surroundings. But to achieve this lofty goal will require human-directed machine learning in the early stages that is essential to triage erroneous correlations quickly from the system and key to an efficient process will be the presentation of the data in representations that allow the human brain to visualize these new relationships.
So perhaps, in this new age of AI processing of data, the early leaders will come from the data presentation and business analytics world as they hold the key to accessing the knowledge locked up in the data – integrating the data in unique and unexpected ways
And for those of you wondering what the binge-watching TV revelation was – Chris discovered that when he watched any television show sequentially his health and food behavior was worse. So rather than watching Breaking Bad Season 1, followed by Episode 2 which he found to be associated with more frequent visits to the snack cupboard and fridge a simple change to watch another series in between each episode improved his behavior. Watch Breaking Bad Season 1, Episode 1 but follow it with Game of Thrones Season 1 Episode 1.
Wearable data will become just another data stream of input to the deep learning systems processing large amounts of data that individually we cannot process or reconcile to create new insights and predictive models that will bring precision and customized medicine choices to each and every one of us.
This post previously appeared on AI Med
