CMU Technology Helps People Maintain Health at Home Collaborative App Uses Passive Sensing to Understand Mental Health, Chronic Disease
Amanda SapioTuesday, September 16, 2025Print this page.
An app designed by Carnegie Mellon University researchers uses artificial intelligence to assess depression and fatigue. The goal? To better understand mental health in people living with multiple sclerosis (MS).
The app was part of a study by CMU's School of Computer Science (SCS) and colleagues at the University of Pittsburgh and UPMC. Mayank Goel, an associate professor in the Software and Societal Systems Department and Human-Computer Interaction Institute in SCS, led CMU's efforts with Ph.D. student Prerna Chikersal. The study found that using existing sensors on smartphones and fitness trackers could give patients the data they need to monitor their own symptoms and provide clinicians with information to better respond to patients when treating MS or, potentially, other chronic neurological disorders.
Goel explained that this study is part of his broader work at CMU dedicated to providing individuals with first-hand access to their health data.
What made you focus on MS and mental health?
We chose to focus our research on depression because many studies have focused on the physical complications of MS, but there isn't a lot of research on the mental impact of the disease.
What have you learned?
Overall, we had around a 35 to 40% incidence rate for depression and a 50 to 55% incidence rate for fatigue and sleep deprivation in the people we studied who had MS. Most of our data collection happened during the pandemic, particularly during the stay-at-home period, which affected patients' mental health as well. For some patients, the pandemic was actually a positive because they could work from home and reduce job stress. Others became extremely lonely and isolated, and their mental health was far worse during the stay-at-home period.
What was the app you developed, and how did it work?
We used a Fitbit paired with a custom app that patients installed on their smartphones to collect the data. The app and Fitbit collected the number of steps patients took in a day as well as heart rate, geographic location, sleep duration and how socially active they were. To identify social activity, we looked at their GPS location to see if they were inside their home or in social spaces. We also used the patient's Bluetooth to scan the number of other Bluetooth devices nearby, indicating if they were around other people in a social setting. We also studied how many apps they used in a day, as well as how many text messages and phone calls they received. Finally, we sent patients a message three times each day asking them to rank their depression and fatigue on a scale of one to five. Most people didn't answer that question every time we asked, but on average, we had around 80% compliance.
What was the motivation for creating the app originally?
I have always been motivated by making applications using devices that we already have, such as smartphones and Fitbits. Health has remained a priority for me given the promise of immediate impact on people's lives.
Do you think you'll conduct another study similar to this one now that we're a few years out from COVID?
We have been doing a lot of studies on depression and fatigue in other populations. For example, we recently collected data with breast cancer survivors. And we collected similar data for LGBTQIA+ teens struggling with depression and high school girls dealing with daily stress.
At CMU, we have been building a machine learning model that works across populations. The challenge is that everyone's "normal" is different, so we have been working on a model that can adapt to various users and contexts. That is one of the reasons why we have been collecting the data from different populations.
How are doctors using the data you collect and provide?
The goal is to make the data part of their decision-support system. We don't want our algorithms to act as a medical device. We are determining what information should be shared with doctors so they can use it in their decision-making without relying too heavily on it.
Another option we have been testing is showing the data to the patient but not the doctor. That way, the patient is more empowered when they go to the doctor. They have more information and know what to talk about.
Do you have any other research projects you're working on?
Overall, my main objective is to help people maintain their health at home. In some cases, that means building technology that helps screen your symptoms for depression, stress, fatigue, etc. I envision the app also being used to improve patient-doctor communication. We are looking at how we can better measure people's physiology, including their heart rate, calories they burn, blood pressure, glucose levels, and how what we eat affects our body.
Aaron Aupperlee | 412-268-9068 | aaupperlee@cmu.edu