On the effects of signal processing on sample entropy for postural control

Research output: Contribution to journalArticle

Abstract

Sample entropy, a measure of time series regularity, has become increasingly popular in postural control research. We are developing a virtual reality assessment of sensory integration for postural control in people with vestibular dysfunction and wished to apply sample entropy as an outcome measure. However, despite the common use of sample entropy to quantify postural sway, we found lack of consistency in the literature regarding center-of-pressure signal manipulations prior to the computation of sample entropy. We therefore wished to investigate the effect of parameters choice and signal processing on participants’ sample entropy outcome. For that purpose, we compared center-of-pressure sample entropy data between patients with vestibular dysfunction and age-matched controls. Within our assessment, participants observed virtual reality scenes, while standing on floor or a compliant surface. We then analyzed the effect of: modification of the radius of similarity (r) and the embedding dimension (m); down-sampling or filtering and differencing or detrending. When analyzing the raw center-of-pressure data, we found a significant main effect of surface in medio-lateral and anterior-posterior directions across r’s and m’s. We also found a significant interaction group × surface in the medio-lateral direction when r was 0.05 or 0.1 with a monotonic increase in p value with increasing r in both m’s. These effects were maintained with down-sampling by 2, 3, and 4 and with detrending but not with filtering and differencing. Based on these findings, we suggest that for sample entropy to be compared across postural control studies, there needs to be increased consistency, particularly of signal handling prior to the calculation of sample entropy. Procedures such as filtering, differencing or detrending affect sample entropy values and could artificially alter the time series pattern. Therefore, if such procedures are performed they should be well justified.

Original languageEnglish (US)
Article numbere0193460
JournalPLoS One
Volume13
Issue number3
DOIs
StatePublished - Mar 1 2018

Fingerprint

Entropy
entropy
Signal processing
sampling
Pressure
Virtual reality
Time series
time series analysis
Sampling
Outcome Assessment (Health Care)
sensory evaluation
Research

ASJC Scopus subject areas

  • Biochemistry, Genetics and Molecular Biology(all)
  • Agricultural and Biological Sciences(all)

Cite this

On the effects of signal processing on sample entropy for postural control. / Lubetzky, Anat; Harel, Daphna; Lubetzky, Eyal.

In: PLoS One, Vol. 13, No. 3, e0193460, 01.03.2018.

Research output: Contribution to journalArticle

@article{6308a57ee5874eb7a21a4d64c9a23265,
title = "On the effects of signal processing on sample entropy for postural control",
abstract = "Sample entropy, a measure of time series regularity, has become increasingly popular in postural control research. We are developing a virtual reality assessment of sensory integration for postural control in people with vestibular dysfunction and wished to apply sample entropy as an outcome measure. However, despite the common use of sample entropy to quantify postural sway, we found lack of consistency in the literature regarding center-of-pressure signal manipulations prior to the computation of sample entropy. We therefore wished to investigate the effect of parameters choice and signal processing on participants’ sample entropy outcome. For that purpose, we compared center-of-pressure sample entropy data between patients with vestibular dysfunction and age-matched controls. Within our assessment, participants observed virtual reality scenes, while standing on floor or a compliant surface. We then analyzed the effect of: modification of the radius of similarity (r) and the embedding dimension (m); down-sampling or filtering and differencing or detrending. When analyzing the raw center-of-pressure data, we found a significant main effect of surface in medio-lateral and anterior-posterior directions across r’s and m’s. We also found a significant interaction group × surface in the medio-lateral direction when r was 0.05 or 0.1 with a monotonic increase in p value with increasing r in both m’s. These effects were maintained with down-sampling by 2, 3, and 4 and with detrending but not with filtering and differencing. Based on these findings, we suggest that for sample entropy to be compared across postural control studies, there needs to be increased consistency, particularly of signal handling prior to the calculation of sample entropy. Procedures such as filtering, differencing or detrending affect sample entropy values and could artificially alter the time series pattern. Therefore, if such procedures are performed they should be well justified.",
author = "Anat Lubetzky and Daphna Harel and Eyal Lubetzky",
year = "2018",
month = "3",
day = "1",
doi = "10.1371/journal.pone.0193460",
language = "English (US)",
volume = "13",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "3",

}

TY - JOUR

T1 - On the effects of signal processing on sample entropy for postural control

AU - Lubetzky, Anat

AU - Harel, Daphna

AU - Lubetzky, Eyal

PY - 2018/3/1

Y1 - 2018/3/1

N2 - Sample entropy, a measure of time series regularity, has become increasingly popular in postural control research. We are developing a virtual reality assessment of sensory integration for postural control in people with vestibular dysfunction and wished to apply sample entropy as an outcome measure. However, despite the common use of sample entropy to quantify postural sway, we found lack of consistency in the literature regarding center-of-pressure signal manipulations prior to the computation of sample entropy. We therefore wished to investigate the effect of parameters choice and signal processing on participants’ sample entropy outcome. For that purpose, we compared center-of-pressure sample entropy data between patients with vestibular dysfunction and age-matched controls. Within our assessment, participants observed virtual reality scenes, while standing on floor or a compliant surface. We then analyzed the effect of: modification of the radius of similarity (r) and the embedding dimension (m); down-sampling or filtering and differencing or detrending. When analyzing the raw center-of-pressure data, we found a significant main effect of surface in medio-lateral and anterior-posterior directions across r’s and m’s. We also found a significant interaction group × surface in the medio-lateral direction when r was 0.05 or 0.1 with a monotonic increase in p value with increasing r in both m’s. These effects were maintained with down-sampling by 2, 3, and 4 and with detrending but not with filtering and differencing. Based on these findings, we suggest that for sample entropy to be compared across postural control studies, there needs to be increased consistency, particularly of signal handling prior to the calculation of sample entropy. Procedures such as filtering, differencing or detrending affect sample entropy values and could artificially alter the time series pattern. Therefore, if such procedures are performed they should be well justified.

AB - Sample entropy, a measure of time series regularity, has become increasingly popular in postural control research. We are developing a virtual reality assessment of sensory integration for postural control in people with vestibular dysfunction and wished to apply sample entropy as an outcome measure. However, despite the common use of sample entropy to quantify postural sway, we found lack of consistency in the literature regarding center-of-pressure signal manipulations prior to the computation of sample entropy. We therefore wished to investigate the effect of parameters choice and signal processing on participants’ sample entropy outcome. For that purpose, we compared center-of-pressure sample entropy data between patients with vestibular dysfunction and age-matched controls. Within our assessment, participants observed virtual reality scenes, while standing on floor or a compliant surface. We then analyzed the effect of: modification of the radius of similarity (r) and the embedding dimension (m); down-sampling or filtering and differencing or detrending. When analyzing the raw center-of-pressure data, we found a significant main effect of surface in medio-lateral and anterior-posterior directions across r’s and m’s. We also found a significant interaction group × surface in the medio-lateral direction when r was 0.05 or 0.1 with a monotonic increase in p value with increasing r in both m’s. These effects were maintained with down-sampling by 2, 3, and 4 and with detrending but not with filtering and differencing. Based on these findings, we suggest that for sample entropy to be compared across postural control studies, there needs to be increased consistency, particularly of signal handling prior to the calculation of sample entropy. Procedures such as filtering, differencing or detrending affect sample entropy values and could artificially alter the time series pattern. Therefore, if such procedures are performed they should be well justified.

UR - http://www.scopus.com/inward/record.url?scp=85042931274&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85042931274&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0193460

DO - 10.1371/journal.pone.0193460

M3 - Article

C2 - 29494625

AN - SCOPUS:85042931274

VL - 13

JO - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 3

M1 - e0193460

ER -