Smartphones and mental illness: How an LSU professor is using AI technology to empower patients and clinicians | News

Can smartphones detect mental health patterns and help clinicians evaluate patients?

Psychology professor Alex Cohen has been researching this question since he was in graduate school 20 years ago. Today, he collaborates with psychology students and faculty from LSU and abroad to look into how smartphone technology can assist clinicians with mental health evaluations.

“The reason I got into it is because I never trusted my symptom ratings, my diagnoses, you know, I just saw people as being too complex to reduce down to a single number. And so, using objective technologies, whether it be from a smartphone or from audio or video recording or something else, gives you much richer and more flexible data for understanding people,” Cohen said.

Cohen developed an app that uses self-report assessments, geolocation tracking and video diaries to look into various mental health problems. People submit data from all over the world, and Cohen and his team analyze different factors like language patterns, vocals and facial expressions.

“We can...use computational modeling to try and predict how [people are] doing, how they will be doing, how they might respond to medication or treatment, how they might respond to stress and things like that,” Cohen said.

According to Cohen, symptoms can be tracked through analyses. For example, Cohen said if someone was interested in depression, they could look at a person’s behavior throughout the day, how often they leave their house, the expressions and words they use in video diaries and how they refer to themselves and others. 

On the other hand, if someone wanted to look into mania, Cohen said they would do a similar analysis with “opposite expectations.” When looking at someone’s geolocation, Cohen said he would expect more activity late at night, as well as animated expressions and a faster speech rate during video diaries.

Cohen said his research so far has shown that people express themselves in different ways, so even the best clinicians’ diagnoses aren’t always accurate.

“There's one big finding, and no one likes it – but that is that clinicians aren’t nearly as reliable as they think they are,” Cohen said. “Clinicians...that are tasked with dealing with people that have serious mental illnesses, [have] the very difficult task of evaluating them, trying to observe their behavior and compare it to some kind of imaginary norm.”  

Cohen recently partnered with Capital Area Human Services District, a mental health and developmental disabilities public services provider for Baton Rouge and surrounding areas, to test the research and technology.

He said there was still a “tremendous amount of work to be done,” and the research faces many challenges, including figuring out how to translate the analyses across different types of people.

“We don't all talk the same, we don't all express things the same way, we don't all have the same, you know, patterns and schedules and things like that,” Cohen said. “The other big [challenge] is figuring out how to make this sustainable."

He said that clinicians can benefit from this research by receiving more accurate data to work with rather than relying solely on a patient's responses. The technology also has the potential to reduce wait times in emergency rooms and improve patients’ responses to treatment.

“I think we could develop tools that clinicians could use…[and] empower patients, and their families and peer support groups, so that they can better understand their illness and better predict what happens [and] better manage treatments so that they're reducing side effects…if they had at their fingertips, you know, readily accessible objective data, if patients were able to supply that…we'd see better collaboration between patients and their doctors,” Cohen said.

Clinical psychology graduate student Kamil Warren has been working with Cohen since she was an undergraduate student. 

Warren’s research focuses on language analysis of speech patterns through mobile phone assessments and looking at how often patients say certain words. Warren stressed the importance of collecting a diverse set of data and not using basic research models as a way to stereotype.

“We shouldn't try to like, break down these individuals in analyzing them against each other, such as African Americans versus Caucasian Americans…you have to make sure you have accurate samples, and like a wide variety of samples for the model to train on so that it won't find correlations that aren't really beneficial,” Warren said.

Warren said she wants to conduct research with greater diversity to determine if current assessments are accurate. She mentioned how African Americans are more likely to be diagnosed with schizophrenia than their white counterparts, and she wants to see if there is something that researchers aren’t looking at that causes this.

“I don't want to just explain these things, but I want to improve treatment. But before you can even do that, you have to understand where the variability is first coming from,” Warren said.

In the future, Warren said the lab was going to collect a greater variety of samples to train their research model on.

Clinical psychology graduate student Edvard Granrud also does research in Cohen’s lab and looks specifically at how what patients say affects how they sound. To do this, Granrud looks at keywords patients use and their vocal acoustics to detect any patterns.

Like Warren, Granrud noted that diversity is an important factor when collecting data, and one of the challenges they face in the lab is how to use their findings for a broader population.

“How do we develop something that is useful for a lot of people? That's difficult, especially when you then look at different cultural backgrounds like ethnicities and countries,” he said.

Granrud’s research also takes into account how people’s speech patterns can change depending on their current environment, noting that there would be a difference in how someone talks to their family members versus a clinician.

While the research is still a work in progress, Granrud believes it has the potential to create greater accessibility to mental healthcare.

“If this does work, I think you'd have this big positive effect, not just like mental healthcare in America, but also other countries that have less resources,” Granrud said. “What this does is potentially levels out the playing field so that more people can have access to help and that whatever resources the system has can be allocated in a fair way to where people who need help get it.”

Comments are closed.