Hearing the sounds of the sleeping brain

If you have followed the previous chapter you will get a plot similar to that shown in figure 4.1

Figure 4.1: The sleep spectrogram with a REM period marked

The band at about 5500 Hz (5500/400 = 13 Hz in the original record) corresponds to the sleep spindles. They do not appear in REM sleep. The intermediate segments where this band does not appear grossly correspond with REM periods. One of them has been selected and appears with white background color in the upper trace.

If you pay attention to the band between 0 and 1600 Hz (0 and 4 Hz in the original record) you can see that it mainly appears in non-REM periods and its appearance is more selective than the appearance of spindles. It corresponds to slow wave sleep (the deepest part of non-REM sleep). When slow wave appear, sleep spindle decrease (this is not evident at first sight).

So the well-known signatures of the different stages are:

• Non-REM sleep (especially stages 2 and 3) is associated with spindles
• Slow wave sleep (stages 3 and 4) is associated with slow waves and attenuation of spindles
• REM sleep is associated with absence of spindles and slow waves.

Let's hear the recording. You have a nice red cursor allowing the navigation through the spectrogram and you can find the signature of each element. Of course, you can replay or hear selectively any segment. The perception can change for different listeners but this is what I hear:

After an initial part of silence (there is no recording at this part of the EDF file) we hear some noises and something like a windy day; immediately sounds like cricket sounds (sleep spindles beginning at 00:18) are heard and something like a waxing and waning intermittent storm (slow waves) appears. Segments of windy and rainy days without cricket sounds (REM periods) can be heard like calm intermediate periods in the middle of the storm.

REM periods sound more or less like light non-REM sleep without spindles but with some differences: the background noise is slightly different from the one heard in non-REM sleep and from time to time some other superimposed frequencies appear.

To see the first feature we can use the spectrum facility of Wavesurfer. We averaged the spectrum for two different segments: non-REM without slow waves (black trace) and REM sleep (red trace). As you see, both spectra are similar although they do not contain exactly the same frequencies. Both spectra, as they appear at the Spectrum section plot, are shown in figure 4.2

Figure 4.2: The spectrum of REM and non-REM segments

The second trait is shown in figure 4.3. From time to time REM sleep is interrupted by some activities different from the background noise. We can hear some of them between 1:27 and 1:28 minutes at the original trace as a chirp. The segment is shown together with the stage-2 spindles (at the right of the figure)

Figure 4.3: A small chirp at the end of a REM sleep period (selected). At the right third of the screen sleep spindles appear. Notice the different time scale in relation to the previous figures

This recording does not contain long segments of alpha rhythm. But you could predict (and you will be right) that it can be heard as a long whistle.