Clustering of Human Sleep Recordings Using a Quantile Representation of Stage Bout Durations - Biomedical Engineering Systems and Technologies - page 357

“crawly” sensation in the legs that is relieved by walking. It is found that these vari-
ables are significantly different in the bout duration group characterized by the highest
mean duration of wake bouts. This finding provides a specific manner in which sleep
dynamics reflects the values of variables that are not specific to sleep. It is noted that
multivariate linear regression captures less than 1% of the variance associated with the
muscular paralysis variable. Thus, machine learning provides access to findings beyond
those available with more traditional statistical tools in this case.
The results presented in this paper are based on a highly compressed representation
of the bout duration distributions, utilizing only the three quartile values of the cumula-
tive bout duration distribution for each sleep stage. It is possible that this compression
limits the capacity of the clustering technique to identify important dynamical features.
Increasing the number of quantiles provides greater representational accuracy, but was
found to also reduce stability of the clustering results. Future work should investigate
alternative representations of sleep dynamical information that simultaneously provide
important detail in the distributions and stability of the machine learning results.
A limitation of the current work is that it only considers the duration of each stage
bout, without regard for what stage occurs immediately afterwards. It would be desir-
able to consider stage transitions. Work in progress by the authors examines the use of
Markov-type variants for this purpose. A major obstacle in this direction is the sparsity
of stage transitions available within a single night of sleep. More accurate modeling of
the sleep stage transition statistics will require the use of multiple nights' sleep data, or
ambulatory monitoring of physiological signals over extended periods of time.
References
1. Aserinsky, E., Kleitman, N.: Regularly occurring periods of eye motility, and concomitant
phenomena, during sleep. Science 118(3062), 273-274 (1953)
2. Benjamini, Y., Hochberg, Y.: Controlling the false discovery rate: a practical and powerful
approach to multiple testing. J. Royal Statistical Soc., Series B 57(1), 289-300 (1995)
3. Berger, R.J., Phillips, N.H.: Energy conservation and sleep. Behav. Brain Res. 69(1-2), 65-73
(1995)
4. Bianchi, M.T., Cash, S.S., Mietus, J., Peng, C.-K., Thomas, R.: Obstructive sleep apnea alters
sleep stage transition dynamics. PLoS ONE 5(6), e11356 (2010)
5. Borg, I., Groenen, P.J.F.: Modern Multidimensional Scaling: Theory and Applications, 2nd
edn. Springer Series in Statistics. Springer, Berlin (2005)
6. Burns, J.W., Crofford, L.J., Chervin, R.D.: Sleep stage dynamics in fibromyalgia patients and
controls. Sleep Medicine 9(6), 689-696 (2008)
7. Brezinova, V.: Duration of EEG sleep stages in different types of disturbed night sleep. Post-
grad. Med. J. 52(603), 34-36 (1976)
8. Cohen, W.W.: Fast effective rule induction. In: Twelfth International Conference on Machine
Learning, pp. 115-123. Morgan Kaufmann (1995)
9. Dang-Vu, T.T., Schabus, M., Desseilles, M., Sterpenich, V., Bonjean, M., Maquet, P.: Func-
tional neuroimaging insights into physiology of human sleep. Sleep 33(12), 1589-1603
(2010)