Breathing detection from tracheal sounds in both temporal and frequency domains in the context of phrenic nerve stimulation.

Abstract

Electrical stimulation of the phrenic nerves via implanted devices allows to counteract some disadvantages of mechanical ventilation in patients with high tetraplegia or Ondine's syndrome. Existing devices do not allow to monitor breathing or to adapt the electroventilation to patients' actual needs. A reliable breathing monitor with an inbuilt alarm function would improve patient safety. In our study, a real-time acoustic breathing detection method is proposed as a possible solution to improve implanted phrenic stimulation. A new algorithm to process tracheal sounds has been developed. It combines breathing detection in both temporal and frequency domains. The algorithm has been applied on recordings from 18 healthy participants. The obtained average sensitivity, specificity and accuracy of the detection are: 99.31%, 96.84% and 98.02%, respectively. These preliminary results show a first positive proof of the interest of such an approach. Additional experiments are needed, including longer recordings from individuals with tetraplegia or Ondine Syndrome in various environments to go further in the validation.