From polysomnography to sleep parameters indexing sleep quality and sleep related physiological and psychometric factors Roman Rosipal, Achim Lewandowski, Georg Dorffner Section of Medical Cybernetics and Artificial Intelligence Center for Medical Statistics, Informatics and Intelligent Systems Medical University Vienna Vienna, Austria ESRS 2010, September, 14.-18., Lisbon
30
Embed
From polysomnography to sleep parameters indexing sleep ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
From polysomnography to sleep parametersindexing sleep quality and sleep relatedphysiological and psychometric factors
Roman Rosipal, Achim Lewandowski, Georg Dorffner
Section of Medical Cybernetics and Artificial IntelligenceCenter for Medical Statistics, Informatics and Intelligent Systems
Medical University ViennaVienna, Austria
ESRS 2010, September, 14.-18., Lisbon
Objectives Methods Results Conclusions
Objectives of the Study
How to define and objectively measure sleep quality?
How to balance objective and subjective measures of the sleepquality? Questionnaires versus PSG, behavioral testing ...
How the subjective perception of sleep relates to the objectivemeasures of the day-time behavior or subject’s physiologicalchanges? For example, does poorly rated and perceived sleepnecessarily mean impaired cognitive ability, increased sleepinessor reduced vigilance?
Objectives Methods Results Conclusions
Objectives of the Study
How to define and objectively measure sleep quality?
How to balance objective and subjective measures of the sleepquality? Questionnaires versus PSG, behavioral testing ...
How the subjective perception of sleep relates to the objectivemeasures of the day-time behavior or subject’s physiologicalchanges? For example, does poorly rated and perceived sleepnecessarily mean impaired cognitive ability, increased sleepinessor reduced vigilance?
Objectives Methods Results Conclusions
Objectives of the Study
How to define and objectively measure sleep quality?
How to balance objective and subjective measures of the sleepquality? Questionnaires versus PSG, behavioral testing ...
How the subjective perception of sleep relates to the objectivemeasures of the day-time behavior or subject’s physiologicalchanges? For example, does poorly rated and perceived sleepnecessarily mean impaired cognitive ability, increased sleepinessor reduced vigilance?
Objectives Methods Results Conclusions
Dataset
Subjects (the Siesta project database): 148 healthy volunteers,67 males and 81 females, age between 20 and 86, spending twoconsecutive nights in the sleep lab
List of 22 tests and measured variables collected during the twoconsecutive days in the sleep lab:
Abbreviation Explanationage Age of a subjects qua Self-rating Questionnaire for Sleep Qualitya qua Self-rating Questionnaire for Awakening Qualitys tot Self-rating Questionnaire for Somatic Complaintsnum m Numerical Memory Test (morning)wb e Well-being Self Assessment Scale (evening)wb m Well-being Self Assessment Scale (morning)pul m Pulse Rate (morning)pul e Pulse Rate (evening)sys m Systolic Blood Pressure (morning)sys e Systolic Blood Pressure (evening)dia m Diastolic Blood Pressure (morning)dia e Diastolic Blood Pressure (evening)vas drive Visual Analogue Scale Test for Drivevas mood Visual Analogue Scale Test for Moodvas aff Visual Analogue Scale Test for Affectivityvas drows Visual Analogue Scale Test for Drowsinessad ts Alphabetical Cross-out Test (total score)ad sv Alphabetical Cross-out Test (variability)ad errp Alphabetical Cross-out Test (percentage of errors)fma r Fine Motor Activity Test (right hand)fma l Fine Motor Activity Test (left hand)
Objectives Methods Results Conclusions
Dataset
Subjects (the Siesta project database): 148 healthy volunteers,67 males and 81 females, age between 20 and 86, spending twoconsecutive nights in the sleep lab
List of 22 tests and measured variables collected during the twoconsecutive days in the sleep lab:
Abbreviation Explanationage Age of a subjects qua Self-rating Questionnaire for Sleep Qualitya qua Self-rating Questionnaire for Awakening Qualitys tot Self-rating Questionnaire for Somatic Complaintsnum m Numerical Memory Test (morning)wb e Well-being Self Assessment Scale (evening)wb m Well-being Self Assessment Scale (morning)pul m Pulse Rate (morning)pul e Pulse Rate (evening)sys m Systolic Blood Pressure (morning)sys e Systolic Blood Pressure (evening)dia m Diastolic Blood Pressure (morning)dia e Diastolic Blood Pressure (evening)vas drive Visual Analogue Scale Test for Drivevas mood Visual Analogue Scale Test for Moodvas aff Visual Analogue Scale Test for Affectivityvas drows Visual Analogue Scale Test for Drowsinessad ts Alphabetical Cross-out Test (total score)ad sv Alphabetical Cross-out Test (variability)ad errp Alphabetical Cross-out Test (percentage of errors)fma r Fine Motor Activity Test (right hand)fma l Fine Motor Activity Test (left hand)
Objectives Methods Results Conclusions
Factor Analysis Towards Parsimonious Sleep Quality Indexing
Spearman rank correlations between sleep parameters andthree factor scores were computed
Objectives Methods Results Conclusions
Factors vs. Individual Variables
Correlations between sleep parameters for the second(physiological) and third (psychometric) factors were found to behigher or comparable with the correlations computed using theindividual variables they consist of (two sample t-test)
This was not true for the first factor where s qua was higher(s qua - 7 questions self-rating sleep quality, Saletu et al. (1987))
Objectives Methods Results Conclusions
Factors vs. Individual Variables
Correlations between sleep parameters for the second(physiological) and third (psychometric) factors were found to behigher or comparable with the correlations computed using theindividual variables they consist of (two sample t-test)
This was not true for the first factor where s qua was higher(s qua - 7 questions self-rating sleep quality, Saletu et al. (1987))
Objectives Methods Results Conclusions
Age Effect
Strong age effect was found for the physiological andpsychometric factors⇒ restriction to age group 20 - 40 yearswhere the effect is not significant
20 30 40 50 60 70 80 90−3
−2
−1
0
1
2
3
4
5
Age [years]
Phys
iolog
ical fa
ctor s
core
s
20 30 40 50 60 70 80 90−3
−2
−1
0
1
2
3
Age [years]
Psyc
hom
etric
facto
r sco
res
Linear fit:R2 = 0.196ρ = 0.443
Linear fit:R2 = 0.373ρ = −0.611
Objectives Methods Results Conclusions
R&K versus PSM - s qua
s qua (subjective sleep quality questionnaire)
Comparable results between R&K and PSM for general sleepparameters (e.g. eff , tst , . . . ), wake, S1 and REM|ρ| ≈ 0.3− 0.36
In addition, PSM shows significant correlations for S2 and SWS(auc, entropy )|ρ| ≈ 0.24− 0.27
Objectives Methods Results Conclusions
R&K versus PSM - s qua
s qua (subjective sleep quality questionnaire)
Comparable results between R&K and PSM for general sleepparameters (e.g. eff , tst , . . . ), wake, S1 and REM|ρ| ≈ 0.3− 0.36
In addition, PSM shows significant correlations for S2 and SWS(auc, entropy )|ρ| ≈ 0.24− 0.27
Objectives Methods Results Conclusions
R&K versus PSM - 2nd factor
physiological factor
R&K: significant correlations for two general sleep parametersfw q4, fs and two SWS parameters tst , tst q2|ρ| ≈ 0.26− 0.39
PSM: significant (and in comparison to R&K higher) correlationsfor general sleep parameters and also significant correlations forparameters representing all sleep stages and wake|ρ| ≈ 0.30− 0.44
Objectives Methods Results Conclusions
R&K versus PSM - 2nd factor
physiological factor
R&K: significant correlations for two general sleep parametersfw q4, fs and two SWS parameters tst , tst q2|ρ| ≈ 0.26− 0.39
PSM: significant (and in comparison to R&K higher) correlationsfor general sleep parameters and also significant correlations forparameters representing all sleep stages and wake|ρ| ≈ 0.30− 0.44
Objectives Methods Results Conclusions
R&K versus PSM - 3rd factor
psychometric factor
R&K: only sleep latency to REM and average duration of REMcycles are significant|ρ| ≈ 0.31,0.26
PSM: significant for parameters representing all sleep stages butnot wake|ρ| ≈ 0.30− 0.43
Objectives Methods Results Conclusions
R&K versus PSM - 3rd factor
psychometric factor
R&K: only sleep latency to REM and average duration of REMcycles are significant|ρ| ≈ 0.31,0.26
PSM: significant for parameters representing all sleep stages butnot wake|ρ| ≈ 0.30− 0.43
Objectives Methods Results Conclusions
Beyond R&K
Higher correlation values of auc and entropy sleep parameterswere observed for combined sub-states models(e.g. ρ = 0.39 vs. 0.42; for auc q4 in wake; 2nd factor; 5.5sub-states)
Number of sub-states varies with individual sleep stages but onaverage it is less than 1/4 of all sub-states
This finding indicates that changes in substructures of thestandard R&K sleep stages may better reflect important aspectsof the sleep process related to subjective or objective evaluationof sleep
Objectives Methods Results Conclusions
Beyond R&K
Higher correlation values of auc and entropy sleep parameterswere observed for combined sub-states models(e.g. ρ = 0.39 vs. 0.42; for auc q4 in wake; 2nd factor; 5.5sub-states)
Number of sub-states varies with individual sleep stages but onaverage it is less than 1/4 of all sub-states
This finding indicates that changes in substructures of thestandard R&K sleep stages may better reflect important aspectsof the sleep process related to subjective or objective evaluationof sleep
Objectives Methods Results Conclusions
Beyond R&K
Higher correlation values of auc and entropy sleep parameterswere observed for combined sub-states models(e.g. ρ = 0.39 vs. 0.42; for auc q4 in wake; 2nd factor; 5.5sub-states)
Number of sub-states varies with individual sleep stages but onaverage it is less than 1/4 of all sub-states
This finding indicates that changes in substructures of thestandard R&K sleep stages may better reflect important aspectsof the sleep process related to subjective or objective evaluationof sleep
Objectives Methods Results Conclusions
Conclusions
PSG provides objective measures which significantly correlatewith the collected subjective and objective measures of sleepquality
The proposed probabilistic approach allows to model finermicro-structure of sleep which increases the level of the studiedcorrelations
The clinical validation of these results remains the subject of thefurther study
Objectives Methods Results Conclusions
Conclusions
PSG provides objective measures which significantly correlatewith the collected subjective and objective measures of sleepquality
The proposed probabilistic approach allows to model finermicro-structure of sleep which increases the level of the studiedcorrelations
The clinical validation of these results remains the subject of thefurther study
Objectives Methods Results Conclusions
Conclusions
PSG provides objective measures which significantly correlatewith the collected subjective and objective measures of sleepquality
The proposed probabilistic approach allows to model finermicro-structure of sleep which increases the level of the studiedcorrelations
The clinical validation of these results remains the subject of thefurther study