An Open Source Classifier for Bed Mattress Signal in Infant Sleep Monitoring

Abstract

ObjectiveTo develop a non-invasive and clinically practical method for a long-term monitoring of infant sleep cycling in the intensive care unit.MethodsForty three infant polysomnography recordings were performed at 1–18 weeks of age, including a piezo element bed mattress sensor to record respiratory and gross-body movements. The hypnogram scored from polysomnography signals was used as the ground truth in training sleep classifiers based on 20,022 epochs of movement and/or electrocardiography signals. Three classifier designs were evaluated in the detection of deep sleep (N3 state): support vector machine (SVM), Long Short-Term Memory neural network, and convolutional neural network (CNN).ResultsDeep sleep was accurately identified from other states with all classifier variants. The SVM classifier based on a combination of movement and electrocardiography features had the highest performance (AUC 97.6%). A SVM classifier based on only movement features had comparable accuracy (AUC 95.0%). The feature-independent CNN resulted in roughly comparable accuracy (AUC 93.3%).ConclusionAutomated non-invasive tracking of sleep state cycling is technically feasible using measurements from a piezo element situated under a bed mattress.SignificanceAn open source infant deep sleep detector of this kind allows quantitative, continuous bedside assessment of infant’s sleep cycling.

An openly available wearable, a diaper cover, monitors infant's respiration and position during rest and sleep

“…According to international guidelines of respiratory infection management, the RR levels are considered to vary in the order of tens of breaths per minute between normal and pathologically fast breathing, 1,10,24 which indicates that NAPPA's accuracy is clearly within the required limits. Substantial differences in individual RR levels are due to many more physiological factors, such as age, 1 altitude, 25 vigilance state, 16,26 position 27 or other acute medical conditions. 28,29 Respiratory assessment of subtle changes in such conditions may still need more stable and well-established clinical methodology.…”

“…The RRr was estimated using two computational ways 3 both of which were computed from the capnography signal. First, we assessed RR from the peaks in the autocorrelation function as described earlier 16 . In addition, we also estimated RR using a complementary method based on peak detection 16 .…”

“…First, we assessed RR from the peaks in the autocorrelation function as described earlier 16 . In addition, we also estimated RR using a complementary method based on peak detection 16 . For the final RRr , we excluded epochs where difference between autocorrelation‐based and peak detection‐based estimates was more than 5 breaths per minute (bpm) and the autocorrelation was <0.3.…”

“…15 The sensor computed RRn during recording as follows: a 30-s epoch of the signal was read into buffer, then band-pass filtered with second-order Butterworth in 0.1-1.5 Hz, followed by using autocorrelation function to define its peaks as described earlier. 16…”

“…Our pilot studies comparing respiration signals from such sensor and concurrent recordings of polysomnography (PSG) signals have demonstrated that these sensor type and placement are able to track infant's abdominal respiratory movements with high accuracy 15 . We have then implemented a computational algorithm 16 directly into the sensor to allow direct estimation of respiration rate (RR). Here, we wanted to test how reliable is the RR estimated from long‐term recordings with our NAPPA wearable as compared to an established RR estimate obtained from the capnography signals of polysomnography (PSG) recordings.…”