Quantitative Sleep Monitoring - Quantitative EEG Analysis Methods and Clinical Applications - page 279

patient is awake. However, this activity decreases when the patient is in NREM
sleep, and the activity almost disappears after the patient enters the REM sleep
stage. Figures 10.18 and 10.19, adapted from Geyer et al. [43], show 30-second
epochs of an EMG channel when the patient is in stages 3/4 of NREM sleep and in
REM sleep, respectively. A decrease in the amount of activity in the EMG can be
seen as the patient progresses to REM sleep.
A sliding window variance analysis could be used to show the decrease in EMG
activity. A sliding window variance analysis over time is shown in Figure 10.20 for
stages 3 and 4 of NREM sleep. The variance is much larger in NREM stages relative
to a sliding window variance analysis of the EMG channel in REM sleep, as shown
in Figure 10.21.
The variance for a given set of points is defined as in (10.3), where is the mean

of the sampled points. At a given time t, a sliding window variance analysis calcu-
lates the variance for all points sampled within the last T seconds, where T is some
preset constant of time:

μ

N

1

( )

= ∑

2

2

σ = x i − μ (10.3)

N

i 1

Chin1-Chin2

Figure 10.18 30-second epoch of an EMG channel when the patient is in stages 3/4 of NREM sleep. (Cour-
tesy of James Geyer and Paul Carney.)

Chin2-Chin1

Figure 10.19 30-second epoch of an EMG channel when the patient is in REM sleep. (Courtesy of James
Geyer and Paul Carney.)

Stage 3-4 variance for EMG

35

30

25

20

15

10

5

0

0 20 40 60 80 100 120

Figure 10.20 A sliding window variance analysis over time is shown for stages 3 and 4 of NREM

sleep.