A neural network can detect depression and mania in bipolar subjects by analyzing how they hold and tap on their smartphones.
The news: Researchers gave both bipolar and control subjects a phone with a custom keyboard that collected data on key presses and accelerometer movement. They also asked participants to self-report how much they were feeling depressed or manic.
Details: Bipolar participants, when suffering mania or depression, had a more uniform typing time compared with control subjects, who varied their typing speeds. One explanation could be that bipolar people don’t react as much to stimuli, like incoming text messages. The accelerometer data showed that subjects with depressive or manic symptoms tended to hold their phone at an angle.
Results: Using those insights, the researcher trained an algorithm called DeepMood to predict when a bipolar subject was suffering from a mood disturbance. DeepMood reached 90 percent accuracy with sessions that were usually under one minute.