Melvin McInnis
University of Michigan, MI·Medicine & Clinical Research
Bipolar disorder causes swings between depression and mania. Spotting a coming swing early could help patients and clinicians act sooner.
- Fields
Medicine & Clinical Research
Psychology
Computer Science
- Topics
bipolar disorder
mood disorders
digital phenotyping
- Grants
- 1 known grant, through Oct 2028
What they work on
- Test whether average levels of emotional valence and arousal, measured by phone self-ratings (EMA) and by the PRIORI speech algorithm, predict self-reported and clinician-rated mood severity in 160 adults with bipolar disorder.
- Test whether day-to-day instability in valence and arousal predicts later mood symptoms, and whether PRIORI's passive estimates from ambient audio improve predictions beyond EMA alone.
- Test whether large deviations from a person's usual emotion levels (anomalies) predict mood episodes, including in weeks when only passive PRIORI data are available.
- Examine how environmental noise, location, social factors and psychosocial functioning influence models that predict mood from digital phenotyping data.
More about the research
Digital phenotyping uses data from phones and other devices to measure a person's behavior and emotions in daily life. The team's earlier work used ecological momentary assessment (EMA), in which people rate their feelings on a phone several times a day. That work showed that emotion patterns predict mood symptom severity. These patterns include average emotional valence (how positive or negative a person feels) and arousal (how energized they feel), day-to-day instability, and sudden large changes, called anomalies. Frequent EMA is a burden for participants, however. The team therefore developed PRIORI, an app and set of machine learning algorithms. PRIORI samples ambient audio every 15 minutes and estimates valence and arousal from speech, and these estimates track EMA ratings.
Why it matters
Passive monitoring of emotion from speech could give people with bipolar disorder and their clinicians an early warning of worsening mania or depression, without the burden of constant self-reports. The results could support personalized, real-world models of mood and better tools for tracking mental illness over time.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Joshua Curtiss
Northeastern University, MA·Psychology·Early-career
Evaluate whether forecasting machine learning using dynamical systems and network features from ecological momentary assessment data can predict future symptoms across emotional disorders.
Leslie Adams
Stanford University, CA·Public Health & Epidemiology·Early-career
Modify an existing smartphone digital phenotyping protocol, combining ecological momentary assessment surveys with passive GPS and accelerometer data, so it is appropriate for Black youth.
Madhukar Trivedi
University of Texas Southwestern Medical Center, TX·Medicine & Clinical Research
Investigate how the amount of social media use and positive or negative online experiences affect depression, anxiety and suicidal ideation trajectories over six months in 100 at-risk Texas youth.
The school
University of Michigan: 124 labs on PI Pages, $2.1B a year on research, 963 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.