Bringing real-time stress detection to scale: Development of a biosensor driven, stress detection classifier for smartwatches - PROJECT SUMMARY/ABSTRACT
The goals of this mentored, patient-oriented, research career development award are two-fold: 1) Characterize
the autonomic nervous system correlates of stress-reactivity, both in laboratory and ambulatory contexts in
order to inform the development of a biosensor driven, stress-detection classifier algorithm that can run on
commercially available smartwatches, and 2) establish the principal investigator as an independent researcher
at Massachusetts General Hospital - Harvard Medical School. The specific aims of this research will be
accomplished through an innovative study leveraging the strengths of traditional laboratory-based,
psychophysiological research, and cutting-edge, in natura monitoring of stress and stress’ autonomic
correlates using a combination of ecological momentary assessment of affect, and ambulatory
electrocardiogram monitoring. This research will inform the development of a stress-detection algorithm that
will run on commercially available smartwatches. The clinical and health applications for real-time stress
detection are numerous, but this technology holds particular promise for individuals in early recovery from
alcohol use disorder for whom unchecked stress heightens risk for alcohol use and engagement in other
maladaptive coping behaviors. The smartwatch-embedded stress detection algorithm developed in this
research will ultimately be linked to existing smartphone-based relapse prevention apps that will prompt
patients with real-time coaching to mitigate alcohol use risk. Aims of the principal investigator’s career
development and training plan include, 1) learning fundamental principles of machine learning with an
emphasis on biosensor technologies, 2) gaining facility with Matlab programming, with an emphasis on signal
analysis and psychophysiological model building, 3) broadening expertise in cardiovascular waveform and
interval analysis with particular emphasis on artefact management, and 4) acquiring skills in the development
and application of mHealth-based clinical interventions. These goals will be achieved through a training plan
comprised of mentorship, formal coursework, seminars, conferences, and manuscript preparation. Knowledge
gained via the training plan will be augmented by the research undertaken. Drs. John Kelly, Paolo Bonato, Gari
Clifford, and Bettina Hoeppner will serve as mentors on this award, and will provide targeted expertise in
machine learning approaches, Matlab programing, artefact management, and mHealth treatment development.
Massachusetts General Hospital - Harvard Medical School provides an exceptional environment in which to
conduct this training and research. By the end of the 5-year award period, the goals are to have a working
stress-detection classifier algorithm ready for R01 testing, and for the principal investigator to be established as
an independent investigator. This award is consistent with NIH's goal of increasing and maintaining a strong
cohort of investigators to address the nation's clinical research needs.
