STATISTICAL CHARACTERIZATION OF ACTIGRAPHY DATA DURING SLEEP AND WAKEFULNESS STATES

Alexandre Domingues, J.M. SanchesInstituto de Sistemas e Robótica

Instituto Superior Técnico,Lisboa, Portugal

Teresa PaivaCentro de Electroencefalografia e Neurofisiologia Clínica

Faculdade de Medicina da Universidade de Lisboa Lisboa, Portugal

Abstract

Results

• The diagnosis of Sleep disorders, highly prevalent in the western countries, typically involves sophisticated procedures and equipments that are intrusive to the patient. • Wrist actigraphy, on the contrary, is a non-invasive and low cost solution to gather data which can provide valuable information in the diagnosis of these disorders.• In this work, actigraphy data, acquired from 20 healthy patients and manually segmented by trained technicians, is described by a weighted mixture of two distributions where the weight evolves along the day according to the patient circadian cycle. • Several mixture combinations are tested and statistically validated with conformity measures. It is shown that both distributions can co-exist at a certain time with varying importance along the circadian cycle.

Motivation

32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society

• Afect a significative percentage of adult population.

• Related with diabetes, obesity, depression, cardiovascular diseases.

• The diagnosis envolves complex and intrusive procedures such as the ones in a Polysomnography (PSG).

• Many individuals never seek medical care.

Methods (cont.)

• Estimation of the SW state, from the variation of the weight α

along the circadian cycle, from non-segmented data .

• Sleep stateGamma Distribution and Inverse Gaussian lead to a minimum fitting error

• Wakefulness stateRician and Maxwell – Boltzan distributions lead to a minimum fitting error.

Ondrej AdamecFaculty of Electrical Engineering

VSB Technical University,Ostrava, Czech Republic

Sleep disorders

Alternative diagnosis tools

Typical Polysomnography setup

Actigraph

• Sleep diary

• Portable ECG monitoring

• Temperature monitoring

• Actigraphy

These tools do not replace the accuracy of a PSG!!

Raw data analysis

Single distribution

References [1] João Sanches, Pedro Pires and Teresa Paiva, ”Cell Phone based Sleep electronic Diary (SeD),” in 20th Congress of the European Sleep Research Society, 14-18 September, Lisbon, Portugal 2010.[2] J. Lotjonen at al., ”Automatic Sleep-Wake and Nap Analysis with a New Wrist Worn Online Activity Monitoring Device Vivago Wrist- Care,” SLEEP, vol. 26, pp 86-90, 2003.[3] C. A. Kushida at al., ”Comparison of actigraphic, polysomnographic, and subjective assessment of sleep parameters in sleep-disordered patients,” leep Medicine, vol. 2, no. 5, pp 389 - 396, 2001.[4] P. Pires, T. Paiva, J. Sanches, ”Sleep/Wakefulness state from actigraphy,” in 4th Iberian Conference on Pattern Recognition and Image Analysis, Póvoa de Varzim, Portugal, 2009, pp. 362-369.[5] G. Jean-Louis et al., ”Sleep estimation from wrist movement quantified by different actigraphic modalities,” Journal of Neuroscience Methods, vol. 105, no. 2, pp 185 - 191, 2001.[6] J. Paquet, A. Kawinska, J. Carrier, ”Wake Detection Capacity of Actigraphy During Sleep,” SLEEP, vol. 30, pp 1362-1369, 2007.[7] V. Gimeno et al., ”The Statistical Distribution of Wrist Movements during Sleep,” Neuropsychobiology, vol. 38, no. 2, pp 108-112, 1998.[8] H. Shinkoda et al., ”Evaluation of human activities and sleep-wake identification using wrist actigraphy,” Psychiatry and Clinical Neurosciences, vol. 52, no. 2, pp 157-159, 1998.[9] Vasickova, Z., Augustynek, M., ”New method for detection of epileptic seizure,” Journal of Vibroengineering, pp 209, 2009.[10] Augustynek, M., Penhaker, M., Korpas, D., ”Controlling Peacemakers by Accelerometers,” The 2nd International Conference on Telecom Technology and Applications, Bali Island, Indonesia, pp 161-163, 2010.

•Actigraphy data collected from 20 healthy patients during a

period of approximatly 14 days in normal quotidian life.

• Segmentation into sleep/wakefulness (SW) periods by trained

technicians

•Data segments were grouped according to the corresponding

sleep or wakefulness state and the two histograms were

computed.

Conclusions• This work proposes a description for actigraphy data based on the different statistical characteristics of the movements during sleep and wakefulness states. •It is shown that the global activity can be described by a mixture of two distributions; one associated with movements during sleep state (Gamma distribution) and other associated with movements during wakefulness state (Rician distribution ). •The weight coefficient of the mixture,α, evolves along the circadian cycle and may be used to help in the estimation of the Sleep/Wakefulness state. • Future work will join α with other physiological/environmental features to build a robust tool to aid the diagnosis of sleep disorders

Single distribution fitting:

• Several different distributions were fitted to assess the best

representation for each data type, per patient, according to:

•The fitting error was computed according to:

Mixture distribution fitting:

• Probability distribution function for the mixture distribution:

2||)(),(||min argˆ xhxp kkk

),(

|)(),(|

xp

xhxpe

k

kkk

ws ppxp ).1(.)(

Data sets

Methods

Data processing:

Single distribution fitting

Mixture distribution fitting

Mixture distribution

• The mixture of Gamma (sleep state) and Rician (wakefulness state) distributions lead to a minimum fitting error

Non-segmented data

Evolution of α along the circadian cycle