1. The Measurement of Sleep: A Practical Workshop for Investigators Polysomnography
Patrick J. Strollo M.D.
Martica Hall, Ph.D.
Neuroscience – Clinical & Translational Research Center Laboratory Team
Pittsburgh Mind-Body Center on Sleep Workshop
Pittsburgh, PA
April 11, 2008

2. Agenda Brief review of polysomnography (PSG)
Goal: Learn what is meant by poly (many) somnus (sleep) graphein (to write)
Tour of N-CTRC sleep laboratory
Goal: learn about the different types of studies that can be conducted in the N-CTRC
Meet sleep technician and sleep study “participant” to see what’s involved in using PSG to measure sleep
Goal: learn what all of the electrodes & monitors measure
Watch as signals are collected from participant and review how different signals change with behavior
Goal: learn what different signals look like
Review examples of sleep pathologies
Goal: learn about some of the signals that indicate sleep pathologies
Review two kinds of advanced signal processing
Goal: learn what is meant by spectral analysis of the EEG and EKG during sleep

3. Advanced Signal Processing
Spectral analysis of signals collected during sleep studies
What is spectral analysis?
Decompose a complex, multi-determined signal
Move from time to frequency domain (power is variability2)
Spectral analysis of the EEG
Example: EEG profile in patients with insomnia differs by gender and across the night
Spectral analysis of heart rate variability
Example 1: Methods
Example 2: Lab stressor affects HRV during sleep

4. Delta & REM Counts and vPSG Sleep Histogram
Compute:
Total power
Relative power
For:
All Night
Individual Sleep Cycles

5. QEEG Bandwidths Delta .5 – 4 Hz. Theta 4 – 8 Hz. Alpha 8 - 12 Hz.
Sigma
Hz.
Beta
16 – 32 Hz.

6. Delta Power and VPSG Sleep Histogram
Compute:
Total power
Relative power
For:
All Night
Individual Sleep Cycles
Example…

7. Frequency and time domain analysis of EEG power during NREM sleep in primary insomnia Supported by MH24652, RR024153, RR00052 (D. Buysse, PI)
Insomnia is a clinical disorder with sleep and waking symptoms
Etiology uncertain, but hyperarousal often felt to be a critical component
Subjective symptoms
Hypothalamic-pituitary-adrenal axis
Functional neuroanatomy using FDG PET studies
Beta power in quantitative EEG during NREM
Krystal SLEEP 2002; Perlis Sleep Med Rev 2001, Perlis J Sleep Res 2001; Merica Eur J Neurosci 1998)

8. Participants General PI and GSC recruited in 3:1 ratio
Age years, men and women
Medical history, psychiatric history (SCID), sleep disorders history, screening PSG (AHI, PLMAI < 15)
PI (n = 48)
DSM-IV Primary Insomnia
PSQI ≥ 7
No specific quantitative criteria by diary or PSG
GSC (n = 25)
No sleep disorder
PSQI ≤ 5
Equated for age and sex with PI

9. Power-frequency plots: Whole night

10. Power-frequency plots by NREM period: Women

11. Power-frequency plots by NREM period: Men

12. Heart Rate Variability: What is it?
Heart rate is rhythmic and varies dynamically in response to intrinsic and extrinsic inputs and demands (CNS activity, mechanical changes, reflex-related changes, behavior, psychological stress, affect).
Interbeat intervals (IBIs) refer to milliseconds between beats. Evaluate in the time domain or frequency domain.
Two main components of HRV
Low frequency changes (~3-9 cycles/min.)
Multiply-determined: input from PNS and SNS
High frequency changes (9-24 cycles/min.)
Related to PNS (‘vagal’ ‘RSA’)
Low-to-High frequency ratio
Index of sympatho-vagal activity

13. Frequency Domain Estimates of HRV (QEKG)
60-minute IBI sequence
10 (shaded) minute IBI sequence
Power spectral estimates
of variability in 10-minute
IBI epoch (raw, smoothed)
IBI variability is partitioned along a frequency spectrum using frequency-modeling techniques (e.g., fast Fourier Transformations (FFTs), autoregressive spectral analyses).
Amount of variability (spectral power) is estimated for given frequency components (bandwidths).
Low Frequency = Hz, High Frequency = Hz
Slide courtesy of Julian F. Thayer

14. DAILY: Fill out Sleep Diary and Wear Wrist Actigraph
Sleep SCORE: Study Protocol
(HL076379, Investigators: K. Matthews, M. Hall, D. Buysse, P. Strollo, T. Kamarck, S. Reis)
DAILY: Fill out Sleep Diary and Wear Wrist Actigraph
PSG Sleep Study
(2 nights)
Ambulatory BP
(48 hours)
DAY 1
DAY 10
EEG, EMG, EOG, EKG
GNT GMT
SCORING PSG visual sleep stage scoring in 20 second epochs
EKG HRV processing in 2-minute epochs

15. Heart rate variability: Processing & linking HRV and vPSG data
HRV records processed: 101 (Night 2)
Mean sleep duration: 7 hours (420 minutes) 
210 (2-minute) epochs 
total number of epochs (101 x 210)  > 21,120
HRV Output: LF power, HF power, LF:HF Ratio, Respiration Rate, etc.
vPSG
HRV
W W W W W W
WAKE NREM

16. NREM differs from Wakefulness & REM
HF Power LF:HF

17. How Many HF HRV Epochs During NREM Sleep Are Enough?
Number of HF HRV epochs during NREM for G = 0.8
NREM Whole Night = 5; NREM 1 = 7; NREM 2 = 7, NREM 3 = 6, NREM 4 = 3

18. How Many LF:HF HRV Epochs During NREM Sleep Are Enough?
Number of LF:HF HRV epochs during NREM for G = 0.8
NREM Whole Night = 13; NREM 1 = 13; NREM 2 = 15, NREM 3 = 17, NREM 4 = 7

19. Does HF HRV During NREM Sleep Change Across The Night?
Time F (3,100) = 2.43, p < .07
Time F(3,100)=2.43, p < .07
Time F(3,100)=2.43, p < .07

20. Does LF:HF HRV During NREM Sleep Change Across The Night?
Time F (3,99) = 3.99, p < .02

21. STUDY 1: Acute Laboratory Stress
Ambient Stress Acute Stress Heart Period Variability Sleep Quality
Experimental
Manipulation
8:00 p.m.
Sleep
a.m.
SUBJECTS:
59 healthy undergraduate men and women (50% female, mean age = 19.6 years).
Hall et al., Psychosomatic Medicine, 2004

22. NREM REM Parasympathetic Activity During NREM and REM Sleep
Parasympathetic Activity = high frequency bandwidth ( Hz)