Listen to your heart: hearing rhythmical signals in long recordings

After gaining some confidence on the hearing of a normal sleep, let's analyze the sleep of an apneic patient. Sleep apnea is a process in which the air flow ceases during 10-120 seconds. At the end of the process the patient can be awaken or aroused by the apnea although he is not aware of this if the arousal is brief. Besides cardiovascular effects in severe cases, the effect on wakefulness and drowsiness is similar to the fact of being awakened several hundreds times during the night. Coincident with the sleep apnea, even if there is no awakening, a cycle of bradycardia-tachycardia appears (bradycardia is coincident with the apnea and tachycardia is coincident with the awakening process). We are going to try to hear it. First of all we need a sleep apnea recording. It can be found at EDF site under the name "Here is a 20-minute (about 2.5M) EDF sleep recording of a severe OSAS patient during NREM- and REM-sleep".

The file such as it appears in libRASCH is shown in figure 7.1

Figure 7.1: A segment of the file Osas2002.edf. Notice the recurrent apneas at the respiratory channels. They can be seen in absence of air flow at the signal Resp oro-nasal

Without being very strict, we could think at the accelerated ECG as somewhat similar to a frequency-modulated signal, in which the carrier wave is the shape of the ECG and the signal is the heart rate. The analogy is not perfect because the internal structure of PQRST complexes are not modulated at all.

We accelerated the signal called ECG 400 times in the usual way (the details are the same as in the previous chapters). The result, as can be seen with Wavesurfer, is shown in figure 7.2

Figure 7.2: The spectrogram of the signal ECG of the file Osas2002.edf after a 400 times acceleration

The file is shorter than that the considered in the previous chapters. When we hear it we could hear something like a turbine whose revolutions were modified.

Now we could relate it to respiration. The acceleration is done at the same rate that the signal Resp abdomen of the same file. The result, which can be seen with Wavesurfer, is shown in figure 7.3

Figure 7.3: The spectrogram of the signal Resp abdomen of the file Osas2002.edf after a 400-times acceleration

We could even create a stereo file with the ECG and the respiratory signal. The result is shown in figure 7.4

Figure 7.4: The spectrogram of the signals ECG and Resp abdomen (file Osas2002.edf) after a 400 times acceleration. An apnea cycle is selected to show the relations between both signals

Once we have heard all these different signals I think that we share the impression that EEG during sleep sounds like nature: waves, cricket sounds, rain, chirps... The sound changes from time to time indicating a very sophisticated internal structure spanning from milliseconds to hours. Rhythmical signals are more monotonous and sound like an engine finely regulated. I used to hear different signals and, while I was looking the computer, my dog was close to me. Soon, it became not interested in rhythmical signals. However, the audition of EEG made it alert and my pet tried to locate the origin of these familiar sounds.