Smartphone app predicts bipolar mood swings
An experimental smartphone app analyses changes in vocal patterns to predict the early signs of a bipolar mood swing.
An essential part of managing bipolar disorder -- which is characterised by dramatic mood swings between manic energy and depressive apathy -- is being able to predict when those changes will occur, and act accordingly.
An experimental new smartphone app called PRIORI could offer an non-intrusive way to manage the condition. Developed by a team at the University of Michigan's Depression Center, the app runs in the background, automatically kicking in to monitor the user's voice patterns, whether it be a normal phone conversation or a scheduled weekly call with their care team.
As it monitors the call, it analyses the sounds and silences of the user. A manic episode, for instance, may be heralded by rapid-fire speech with little silence, where as a depressed episode could be signalled by fewer words and longer silences.
After the individual user's patterns have been learned by the app, it will be able to signal both the user and his or her medical team to alert them of the upcoming mood change.
The team is careful to note that only the patient's side of the conversation is recorded, and then it is encrypted and placed on a secure server so that no one can listen to it -- the researchers only have access to the computer analysis of the recordings.
So far, the app has been tested on six patients with rapid-cycling Type 1 bipolar disorder. The researchers have found that the app is able to detect in everyday conversations the vocal modulations that predict varying levels of bipolar mood changes. The next step is to test the app further, with the eventual hope that it will help not just bipolar individuals, but those living with disorders such as depression, schizophrenia, post-traumatic stress and even Parkinson's disease.
"These pilot study results give us preliminary proof of the concept that we can detect mood states in regular phone calls by analysing broad features and properties of speech, without violating the privacy of those conversations," said study co-leader Zahi Karam, a postdoctoral fellow and specialist in machine learning and speech analysis. "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."
The app may not be much use for those who text rather than make calls, but, once the analysis framework is in place, we imagine it would not be too much of a stretch to use software similar to Swype's behaviour-learning technology to the same end.
In the meantime, PRIORI still needs a fair bit of work to get it off the ground. If you want to help out by participating in the study, you can sign up at the University of Michigan clinical studies website.