Multivariable relationships between autonomic nervous system related indices in hyperbaric environments.

The main aim of this work is to model the relationships between parameters extracted from the heart rate variability (HRV) signal, which is derived from the electrocardiogram (ECG), at different stages of a simulated immersion in a hyperbaric chamber. The response of the Autonomic Nervous System is known to be affected by changes in atmospheric pressure, reflected in changes in the HRV signal. A dataset consisting of ECG signals from 17 subjects exposed to a controlled hyperbaric environment, simulating depths from 0 m to 40 m, was used. Both linear and nonlinear dependences of HRV parameters were analysed using linear regression and Mutual Information (entropy-based) techniques. Furthermore, relationships between parameters of the HRV signals, biophysical variables of the subjects, and atmospheric pressure changes were characterized by artificial neural networks. In particular, self-organizing maps (SOM) were trained for modelling and clustering all the data. In the mid-term, these models could be the basis to create predictive models of HRV parameters at high depths in order to increase the safety for divers by warning them if some abnormal body response could be expected just by processing the ECG signal at sea level before immersion.