Predicting mood changes in bipolar disorder through heartbeat nonlinear dynamics: A preliminary study

Bipolar disorder is characterized by mood swings alternating from depression to (hypo-)manic, including mixed states. Currently, patient mood is typically assessed by clinician-administered rating scales and subjective evaluations exclusively. To overcome this limitation, here we propose a methodology predicting mood changes using heartbeat nonlinear dynamics. Such changes are intended as transitioning between euthymic state (EUT), i.e., the good affective balance, and non-euthymic state. We analyzed Heart Rate Variability (HRV) series gathered from four bipolar patients involved in the European project PSYCHE, undergoing 24h ECG monitoring through textile-based wearable systems. Each patient was monitored twice a week, for 14 weeks, being able to perform normal (unstructured) activities. From each acquisition, the longest artifact-free segment of heartbeat dynamics was selected for further analyses. Sub-segments of 5 minutes of this segment were used to estimate trends of HRV linear and nonlinear dynamics. Considering data from a current observation at day t₀, and past observations at days (t_-1, t_-2,...,), personalized prediction accuracies in forecasting a mood state (EUT/non-EUT) at day t₊₁ were 74.18% on average. This approach is intended as a proof of concept of the possibility of predicting mood states in bipolar patients through heartbeat nonlinear dynamics exclusively.