Abstracts

Rationale: Apnea events carry a high risk of death and are believed to cause some cases of sudden unexpected death in epilepsy (SUDEP) and sudden infant death syndrome (SIDS). In the UK alone these two conditions account for over 1000 deaths per year. Respiratory monitoring is a norm in intensive care units but systems for long term out-of-hospital monitoring are not adequate. Current devices for respiratory monitoring sense movement, airflow or blood oxygen levels. There are important drawbacks with these techniques for their ambulatory implementation; large size, sensor movement artifacts and in high power requirements. There is therefore an urgent necessity for a miniaturized, battery operated system capable of reliably detecting apnea, for long term use. Methods: The presented monitoring system was designed to detect apnea of over 25 seconds and consequently trigger an alarm. 25 seconds was selected as an apnea time period because it is long enough so that it does not occur physiologically or involuntarily and is short enough for timely interventions to be provided. To achieve this, breathing is monitored via acoustic sensing at the suprasternal notch. The system’s sensor consists of a miniature microphone placed within a small dome, which is fixed to the subject’s skin with a flexible medical adhesive guaranteeing sensor position for extended periods of time. The new signal processing algorithm for apnea detection is based on a multi-characteristic analysis of the acoustic signals and is implemented in a microchip. Results: In order to test the algorithm’s capability of detecting apnea a total of 630 minutes of breathing data and 182 minutes of voluntary apnea, in 30 second periods, were recorded. The acoustic signals were obtained from 13 different subjects, 11 male and 2 female, aged 18 to 33 years old, weight 50 to 100Kg and height 1.55 to 1.90 meters. In order to test for system robustness part of these recordings were carried out with loud ambient music and the subject continuously moving their head. The sensitivity and specificity of the algorithm to detect apnea sections of over 25 seconds were 99.2% and 99.92% respectively. In absolute terms these corresponds to having a single false alarm over 13.5 hours of breathing and missing 3 apnea events out of the total 363 voluntary 30 second apnea periods. Conclusions: We present a system composed of a robust acoustic sensor and algorithm capable of very high sensitivity and specificity detection of apnoea events. Detection of longer periods of apnoea than 30 seconds would be even more reliable. This system represents a ground breaking advance in the field of ambulatory respiratory monitoring.