Framework to Predict Bipolar Episodes - Sensor Fusion of Electrodermal Activity Heart Rate Variability Sleep Patterns

Patients suffering from Bipolar disorder (BD) experience repeated relapses of depressive and manic states. The extremity of this disorder can lead to many unpleasant events, even suicide attempts, which make early detection vital. Presently, the primary method for identifying these states is evaluation by psychiatrists based on patient’s self-reporting. However, ubiquitous use of mobile devices in combination with sensor fusion has the potential to provide a faster and convenient alternative mode of diagnosis to better manage the illness. This paper proposes a continuous, autonomous sensor fusion based monitoring framework to identify and predict state changes in patients suffering from bipolar disorder. Instead of relying on subjective self-reported data, the proposed system uses sensors to measure and collect, Heart Rate Variability, Quantity and Quality of sleep and Electrodermal activity data as predictors to discern between the two bipolar states. Using classification techniques along with a fusion algorithm, a prediction algorithm can be derived based on all the sensor modalities, gathered via a mobile application, is used to set alerts and visualize the information and results efficiently.