By: Tyra Redmond
Multimedia By Author
From wearable devices to sleep
trackers, data-driven technology is advancing and there’s more to
explore.
Smart
technology combined with AI algorithms is increasingly popular for recording
health patterns with personalized features that monitor well-being.
Professor
Jasleen Kaur is a computer scientist and engineer who studies health
informatics at Waterloo University. Kaur is also the developer manager for
the Ubiquitous
Health Technology Lab. The lab relies on subjective questionnaires and variable devices for sleep monitoring, behavioral patterns, and physical activity.
Their recent testing has shown motion sensing data to be the most accurate in
sleep quality metrics that generate cues of sleep disturbance, durations, and
sleep quality.
“Starting
with the collection of data we begin with a pilot study, then recruit 10-15
participants and spend months collecting information based on individual
activity, from that evaluation we do an analysis,” Kaur said.
Kaur
uses smart thermostats as a data source for evaluating sleep quality and health
monitoring. She
aims to find sleep improvement strategies for sleep technologists and
healthcare policymakers.
She
believes wearable technology is effective in tracking the accuracy of sleep and
health with a minor setback.
“The
main limitation is the batteries in variable devices,” Kaur said. “When the
smartwatch, phone, or ring dies, that data cannot be recorded, energy
efficiency is crucial.”
Professor.
Kaur dedicates the first few months of a new study to selecting efficient
technology to support viable findings.
“Majority
of engineers' research is related to the embedded process,” Kaur said. “But
sensory analysis research is most accurate when paired with existing
technology. When we conduct studies we
create predictions based on individual physical activity impacting
sleep.”
Biomedical
Sensory Monitoring
Pulkit
Grover is a Biomedical Engineer at Carnegie Mellon University. He depends on
information theory to assess bias in existing biomedical systems, including EEG
to design neural sensing and stimulation interfaces.
Grover
uses AI to design stimulation parameters for neurostimulation, e.g. for
cell-type specific stimulation.
“I
am not directly working on energy efficiency at this point, but overall, this
would be about designing better algorithms that do not require high power and
integrating the hardware with power,” Grover said.
Therefore,
he finds wireless, wearable neurostimulation devices very exciting.
For
conducting research, Grover does animal and phantom experiments to ensure systems are reliable for unique testing.
Thomas
Murphy is a professor at Georgia Southern’s Armstrong campus. Unlike Grover, he
designs sensor circuits and digital filters of health data for the Biodynamics
and Human Performance Center.
Most
of Murphy’s work stems from noninvasive techniques involved with human
subjects. However, since COVID-19 the lab has lost its ability to have physical
subjects.
Understanding
innovative advancements is essential for learning sensors and their
limitations.
“Luckily,
there's a standard procedure for designing technology and most of it isn't
changing that rapidly in the market right now. If anything, it's improving
rather than differentiating: such as camera sensors and revolution,” Murphy
said. “In today's technology, we have more computing power that is faster
with sophisticated analysis.”
Even
though research indicates smart devices offer a reasonable degree of accuracy,
they do not substitute the standards of a clinical setting or a sleep
laboratory.
“Anything
where you’re trying to make a judgement based on data requires some
kind of algorithm for your program or application to come from,” Murphy
said. “They
play a large role from the designers' standpoint, whoever creates the app on
wearable devices develops a claim based on their sensor measurements.”
Murphy
believes the' fit' is the biggest inaccuracy of wearable devices when tracking health. He
made an analogy to the reliability of ECG readings on skin that needs to be properly prepped with shaving or gel application.
“This
has always been an issue when tracking any biomedical sensors doing surface
measurements,” Murphy said.
Even
more, Apple and Fitbit products mention the proper fit of their device is very
important to receive precise information.
“Sensors
have an inherent limitation on what they can detect. If it’s a poor fit, that
contributes to error,” Murphy said.