View Full Version : Getting wired could help predict emotions

08-01-05, 09:00 AM
Experiment tracks state of mind

By Carey Goldberg, Globe Staff | June 13, 2005

Psychiatric inpatients are always closely watched, but not this closely.

In a groundbreaking experiment at Massachusetts General Hospital, a handful of patients battling depression have agreed in recent weeks to be wired up for 24-hour-a-day, mobile monitoring of their palm sweat, heart rate, voice dynamics, movements, and location.

The study aims to show that such measures can reliably reflect a patient's state of mind as treatment progresses, researchers say.

But more broadly, it seeks to prove that technology has reached the point where it can monitor a person's basic emotional tenor through ordinary days.

To that end, researchers are developing an easy-to-wear system that continuously gathers masses of information, then quickly crunches and transmits it, providing objective data to help with diagnoses and provide early warning when crisis nears.

''The Holy Grail of what we're trying to do is to develop an automated system that can just look at the physiology and be able to predict if [patients] are depressed and if they're going to become more depressed," said Michael Sung, a media lab researcher at Massachusetts Institute of Technology who is working on the study.

Researchers on aging have long been developing ways to monitor elders' physical well-being from afar, but work on using physiological sensors to track emotions in people of all ages -- 24 hours a day -- is still in its infancy.

Only in the past two or three years, researchers estimate, has the technology gotten small and powerful enough to make such monitoring possible, inspiring Sung and his colleagues to grand visions of possible future uses.

Imagine, they say: You are having a breakdown and reaching for a drink when your cellphone rings: ''Hold on!" says your Alcoholics Anonymous sponsor from across town, alerted by a device that monitors the subtle sweating of your palms.

Or: You are feeling just fine, but suddenly your palmtop computer flashes: ''Warning! You seem to be entering the manic phase of your bipolar disorder."

Or: You tell your therapist you're feeling vaguely bad. ''Let's wire you up and see whether it's more anxiety or depression," is the response.

Monitoring physical signs of emotion could add highly objective elements to the often-subjective business of psychiatric diagnoses, said Dr. Carl Marci, the Mass. General psychiatrist leading the study. The study is expected to last about a year and observe 20 patients. The patients are considered competent enough to decide whether to enroll.

Ultimately, physiological monitoring could become a key component of a typical psychiatrist's toolkit, said Alex Derchak, principal scientist at VivoMetrics, a California-based firm that makes the ''LifeShirt", a monitoring system that combines a variety of sensors in a shirt that patients wear all day and night.

LifeShirt was federally cleared for marketing as a medical device in 2002 and is used mainly for medical studies. But Derchak said it is also at work in about 25 studies ''that are specifically focused on different aspects of mental health: issues such as anxiety, panic, autism, bipolar disorder -- they're all different flavors of the mental health story."

Rhode Island researchers are using LifeShirt on autistic children to determine when they are in a state conducive for learning, Derchak said. At the University of California at San Diego, researchers are using the shirt to record the chaotic movements that patients with bipolar disorder make during a manic phase.

VivoMetrics officials say the shirt is also meant to be used during clinical trials of psychiatric drugs, to help track their effects.

In the experiment in Mass. General's inpatient psychiatric unit, researchers are using a system called LiveNet that employs cheaper, off-the-shelf technology contained mainly in a fanny pack rather than a shirt.

They can follow the patients' voice dynamics, motions, and heart rates, along with tracking their movements: Do they stay in bed? Are they wandering the ward?

They also monitor the patients' skin conductance -- subtle changes in palm sweat that reflect nervous system activity and levels of arousal. And four times a day, patients record how they feel in a palmtop computer.

The patients suffer from such persistent depression that they are hospitalized to undergo a series of electroshock treatments, a last-ditch measure that is often effective. A typical course lasts about two weeks. The researchers check what the monitoring data tell them against the traditional measures of discerning improvement: standard interviews with psychiatrists.

In a highly preliminary analysis of the data from the five patients who have been followed so far, it appears that the system can pick up the post-shock improvement in mood that patients tend to experience, Sung said.

The ultimate goal, Marci said, is to find ''signatures of depression that can aid in diagnosis, relapse-prevention, and choice of treatment."

In particular, ''Can we predict who's going to respond to electroshock therapy? And can we find objective measures that are early indications of relapse?"

One patient in the study, a 34-year-old woman who asked to remain anonymous, said she recalls little inconvenience from the monitors during her two-week stay for electroshocks. She said she was pleased to contribute to work that could someday help fight depression: ''If I could save one person from going through how I feel, I would jump over the moon."

In her case, she said, she had long assumed she disliked leaving her house because of depression, but her psychiatrist told her she had an anxiety disorder. It would be helpful ''if there were a way for them to figure that out faster without guesswork," she said.

Also, she said, ''It's kind of hard when somebody asks you what's going on in your mind -- it's kind of hard to explain it."

For all their computerized sophistication, systems such as LiveNet and LifeShirt are not mind readers. They cannot, for example, say definitively that a patient is feeling anger or happiness, said Lisa Feldman Barrett, a Boston College professor who researches emotion.

Rather, they measure things such as levels of arousal and whether a person is experiencing broadly positive or negative emotion at a given point, she said. But even that crude level of detection can prove useful in research and in therapy, she noted.

Marci recently wrote a paper about an experiment in which a patient in therapy, an unemotional-seeming woman, was wired up for skin-conductance monitoring during her session.

The monitoring turned up a high level of arousal that indicated undisclosed anxiety. When told of this, the patient said that it was the first anyone ''had seen her true very high level of anxiety," and that at last someone could understand her constant suffering. Considerable progress in therapy followed.

''The technology facilitated this incredibly accurate empathic interpretation that then led to this 'Aha!' moment," Marci said.

Marci demonstrated the LiveNet system last month at the exhibit ''Getting Emotional" at the Institute of Contemporary Art. As he wandered from painting to sculpture to video, he recorded his feelings on a palmtop computer while the system monitored his heart rate, skin conductance, motion, and voice dynamics.

A few days later, the initial results were in. His emotional state had varied considerably as he viewed the works. When he watched ''Black Mouth," a cryptic, upsetting video clip featuring contorted sound and a mud-spattered, gaping-mouthed girl, his skin conductance and heart rate indicated he was semi-anxious, while his low voice features suggested helplessness.

But later, in a section of the exhibit featuring portraits of intimacy, he said his skin conductance and heart rate indicated he was balanced and relaxed.