Introduction: Estimation of sympathetic-driven arousal state (SDAS) is an important problem in many conditions. Traditionally, frequency-based metrics of heart rate variability, such low and high frequency power (LF and HF), are used. However, the unimodal nature of these indices can make them misleading, particularly in situations where the breathing rate is a con-founding factor. We hypothesized that a multimodal approach to SDAS es-timation would be more accurate and robust.
Methods: In this study, we collected multimodal data from five healthy volunteers during a three-stage paced breathing task. The first and slowest breathing frequency fell directly within the LF range. We used a state space framework to estimate underlying SDAS.
Results: A unimodal model based on only LF and HF derived observa-tions estimates the highest SDAS at the slowest stage of breathing, likely due to the breathing rate. On the other hand, a multimodal model based on time and frequency domain HRV measures and electrodermal activity ob-servations estimates a low SDAS during this stage that increases with breathing rate. This better reflects known physiology associated with deep, slow breathing.
Conclusion: These results support a multimodal paradigm for SDAS estima-tion, especially when there is no information about breathing rate.