Those diagnosed with bipolar disorder have notorious difficulty adhering to their medications. Not only do mood modulations require medication adjustments, but people who experience episodes of mania often develop, as one of their symptoms, the conviction that medication is unnecessary. This makes patients often noncompliant at the moment they have the greatest need for the stabilizing effects of psychopharmaceuticals.
More globally, bipolar disorder often remains under-treated -- particularly in communities where mental health resources are scarce. But intervention and treatment in the early stages of a period of mania or depression can have a big impact on long-term health, preventing long episodes of mania and depression and even suicide.
With this in mind, a team of computer scientists, engineers and psychiatrists from the University of Michigan are working to develop an app that can catch mood modulations before they become full-blown episodes. "The ability to predict mood changes with sufficient advance time to intervene would be an enormously valuable biomarker for bipolar disorder," psychiatrist and study co-leader Melvin McInnis, M.D., said in a statement.
Using a smartphone app, they will set out to learn if they can predict early warning signs of a shift in mood via pattern changes in a patient's speaking voice and cadence.
So far, the team has worked with six patients who suffer from a rapid-cycling form of Type 1 bipolar disorder, which means that their depressive and manic episodes have historically been more frequent than many other types of bipolar disorder patients. The small pilot study involved giving Droid smartphones with the app installed to the six patients, who also underwent weekly mood assessment meetings with their clinicians. This helped to create the data set that will be used for the next phase of research.
The app works by running in the background of a patient's smartphone, recording his or her voice during phone conversations. A week's worth of recording is encrypted for privacy and analyzed by an algorithm -- researchers never hear the actual conversation. The algorithm analyzes acoustic features of the speech, looking for patterns that correlate to mood. In the future, the team hopes to develop an algorithm that doesn't just recognize, but also anticipates a mood shift.
"These pilot study results give us preliminary proof of the concept that we can detect mood states in regular phone calls by analyzing broad features and properties of speech, without violating the privacy of those conversations," said Zahi Karam, Ph.D., a postdoctoral fellow and specialist in machine learning and speech analysis at the school's Department of Electrical and Computer Engineering, in a statement. "As we collect more data the model will become better, and our ultimate goal is to be able to anticipate swings, so that it may be possible to intervene early."
McInnis and his team presented their new app and research at the International Conference on Acoustics, Speech and Signal Processing in Italy. The team also announced a new study, funded by National Institute of Mental Health, based on their initial finding. The new study will involve 60 patients who are seeking treatment for bipolar disorder at the school's Prechter Bipolar Research Fund at the U-M Depression Center.
Despite encryption, there may still be privacy concerns, particularly for a population that has historically had their right to privacy violated.
Bipolar disorder is not the only mental health condition that can change a person's voice, according to the researchers. Their hope is that this technology could be used to monitor patients with diseases as varied as schizophrenia, post-traumatic stress disorder and Parkinson's disease.