1ResMed Sensor Technologies, ResMed Ltd, Dublin, Ireland, 2Ventures and Initiatives, Resmed Ltd, San Diego, CA, United States,. 3Schlafmedizinisches ...
Automated Sleep Staging Classification Using a Non-Contact Biomotion Sensor A. Zaffaroni1, L. Gahan1, L. Collins1, E. O'Hare1, C. Heneghan2, C. Garcia3, I. Fietze3, T. Penzel3 1ResMed Sensor Technologies, ResMed Ltd, Dublin, Ireland, 2Ventures and Initiatives, Resmed Ltd, San Diego, CA, United States, 3Schlafmedizinisches Zentrum, Universitaetsmedizin Berlin, Berlin, Germany
1. Introduction The use of a Non-Contact Biomotion Sensor (NCBS) allows for a non-invasive method of sleep staging estimation. The method presented here allows for the detection of wake, light (N1+N2), deep (N3) and REM (Rapid-Eye-Movement).
2.1 Data Collection 40 simultaneous PSG and NCBS recordings were carried out on 40 healthy subjects in a sleep lab. An expert scorer manually scored each PSG recording. Of these recordings, 20 were used for algorithm development and 20 were retained for validation.
Age(yrs) Gender
Mean
Std
31.9
9.3
Male
Female
21
19
Table 1 - Dataset Demographics for 40 Recordings
Figure 6 – Sleep Onset, PSG vs NCBS
Both the NCBS 30s epoch three and four sleep state stage decisions had good comparison against PSG for the validation dataset.
3. Results
Author/year Figure 1 - Schematic of NCBS setup
2. Methods An ultra-low power radiofrequency sensor was used to measure: o Respiration amplitude and frequency. o Gross Objectives body movement. A combination of respiration and activity features were then used to estimate wake and sleep stages.
Figure 4 - Sample 4 states hypnogram, NCBS vs PSG
The NCBS 30s epoch sleep/wake decision showed good performance when compared against PSG for the validation dataset, as shown in Table 2.
S-W
Acc
Sens
Spec Kappa
90.6%
52.6%
95.6%
Signal
Acc Kappa
This paper NCBS 79.2% W Long, 2014[2] RE 76.2% N [3] Kortelainen, 2010 BCG 79.0% R ECG,RE 76.1% R Redmond, 2007[4] This paper NCBS 64.1% W Long, 2014[2] RE 63.8% L [5] Isa, 2011 ECG 60.3% D Hedner, 2011[6] PAT,PR,OS,AC 65.4% R
0.53 0.45 0.44 0.46
0.45 0.38 0.26 0.48
Table 4 - 3 & 4 State Results Comparison
0.51
Table 2 - Sleep-Wake Validation Results
Table 4 – 3 & 4 State Validation Results
Figure 2 – Comparison of Hypnogram, Respiration Rate & Movement flags
For validation, software was used to obtain respiration rate from Polysomnography (PSG) [1]. The respiration rate calculated by our system exhibited a 0.92 correlation with PSG, with a 95% confidence interval of [– 0.9 1.7] breaths per minute. The correlation between the NCBS and PSG respiration rates is shown in Fig 3. In this figure blue indicates a high density of values.
Figure 7 – Sleep Stage Duration Correlation
4. Conclusion
Figure 5 – Sleep Eff., PGS vs NCBS
Sleep Efficiency is defined as total sleep time expressed as a percentage of total sleep time plus total wake time. Sleep Onset Latency (SOL) is defined as the duration of wake before sleep. The performance for sleep efficiency and SOL is provided in Table 3 below.
[1] http/ Figure 3 – Respiration Rate Comparison – NCBS vs PSG
Sleep Eff(%) SOL(mins)
Bias
Std
-2.4 -6.4
5.4 7.3
Table 3 - Sleep Eff. & SOL Error: Validation Results (PSG-NCBS)
These results show that an algorithm based on the combination of movement and breathing information is capable of detecting multiple sleep states with a high degree of accuracy. The usage of a non-contact biomotion sensor allows for the low cost monitoring of sleep macrostructures over successive nights in an unconstrained environment.
References [1] http://somnomedics.eu/
[2] Long, Xi, et al. "Analyzing respiratory effort amplitude for automated sleep stage classification." Biomedical Signal Processing and Control 14 (2014): 197205 [3] Kortelainen, Juha M., et al. "Sleep staging based on signals acquired through bed sensor." Information Technology in Biomedicine, IEEE Transactions on14.3 (2010): 776-785. [4] Redmond, Stephen J., et al. "Sleep staging using cardiorespiratory signals."Somnologie-Schlafforschung und Schlafmedizin 11.4 (2007): 245-256. [5] Isa, Sani M., Ito Wasito, and Aniati Murni Arymurthy. "Kernel Dimensionality Reduction on Sleep Stage Classification using ECG Signal." International Journal of Computer Science Issues (IJCSI) 8.4 (2011). [6] Hedner, Jan, et al. "Sleep staging based on autonomic signals: a multi-center validation study." Journal of clinical sleep medicine: JCSM: official publication of the American Academy of Sleep Medicine 7.3 (2011): 301.