Bipolar states can be identified by vocal characteristics, study finds

Google+ Pinterest LinkedIn Tumblr +

People with bipolar disorder experience at least two distinct types of states: mania and depression. Increased activity, racing thoughts and impulsivity are notable characteristics of manic states, while depressive episodes can be comparable to the prolonged periods of deep despair associated with major depressive disorder. There are variations of each state related to specific diagnoses, such as the hypomania (often reported as euphoric and less intense than typical mania) experienced by people with bipolar type II disorder. The frequency of mood changes can also vary widely between (and within) people, ranging from many times a day to weeks. Identifying and tracking these changes is key to providing effective treatments, but consistent monitoring is necessary for accuracy and often difficult to accomplish.

New research published in a 2016 issue of Translational Psychology suggests that such monitoring may be possible using cell phone based vocal analysis.

The experiment was performed by a research team led by M. Faurholt-Jepsen and included 28 outpatients recruited from The Copenhagen Clinic for Affective Disorders. A special application was installed on each subject’s Android-based phone if they had one. Participants without an Android phone were loaned one for the duration of the study. The app sounded daily alarms to remind the patient to provide the required data, which included mood ratings, sleep characteristics, activity levels and several other variables of interest. Measurements of social activity and mobility were extracted by the application from cell phone usage behaviors (calls, texts, range, etc.). Voice data was captured by recording participants’ daily conversations. The measurement phase of this study lasted twelve weeks.

Combining all sources of data for analysis revealed that bipolar states can be objectively identified using vocal features alone. However, the addition of self-monitoring report variables and automatic app measurements led to a significant increase in the accuracy, sensitivity and specificity of the model. Additionally, accuracy was discovered to be better in the classification of manic or mixed (has features of both mania and depression) states when compared to purely depressive states.

Bipolar disorder is a difficult condition to diagnose and manage due to the complexity of shifting mental states. The continuous monitoring required to effectively identify bipolar states has been a barrier because it was not practical for most people to complete daily evaluations. This research shows that personal technologies like cell phones can perform such tasks without seriously disrupting daily life. Voice analysis is an ideal platform for conducting personal monitoring because it is easy to collect data from everyday interactions without interrupting the natural flow of events, though issues may arise in the future if voice-based communication becomes less popular.