Hi all,
I'm scoring some sleep data and noticed multiple (maybe harmonic) peaks in the EEG FFT (denoted with arrows) at some points during the recording.
Can anyone explain why this phenomenon might occur and its significance?
This is in a male Swiss-Webster mouse.
I do not know anything about EEG, but I know a bit about FFT. Spikes in the FFT indicate there is something in the original signal that repeats at the frequency of the spike. For example, if you artificially added a cosine term to your EEG and calculated the FFT, you would see a spike at the frequency of the cosine.
Hi Jeremy, I think you are right about harmonics, ca. 11.4 Hz being the fundamental frequency (leftmost arrow). Usually there is a low pass filter located near the input of the signal processing chain, blocking out the 50 or 60 Hz of the electric power lines; therefore, if the harmonics were present in the EEG signal itself, harmonics above 50 Hz wouldn't be visible but they are. So, I suspect that the 11.4 Hz are genuine but the harmonics are generated in the amplifier behind the filter, either due to overdriving or more likely due to a hardware defect (e. g. decreased supply voltage; is your amplifier supplied by batteries?).
Can you stretch the signal in such a way that not more than a few seconds are displayed, and post a screenshot of it here? I expect to see plateaus at maximum or minimum signal values (the limiting has to be asymmetric, because the even harmonics are present, too).
Hi Joerg,
Here's a picture of just 10 sec of data showing a portion with multi-peaks and a portion without.
Note, that these multiple peaks are not evident in the whole recording, just sporadically throughout.
Thanks,
Jeremy
Hi Jeremy,
obviously, I was wrong about the plateaus. Two details catch my attention:
First, the frequency of the spikes is not constant, e. g. the fourth spike is at about 47 Hz in the FFT attached to your question, but at 50 Hz in the second multipeak FFT. So, that's another reason excluding external sources like railway power supplies etc.
Second, each time your FFT produces spikes, there is an additional peak at about 2 to 3 Hz which exceeds the top of the scale. This peak does not fit into the pattern of the remaining peaks (different distance, so the remaining peaks are not harmonics of the 2..3 Hz peak) but it might be the cause.
If the FFT is done on dedicated hardware, it might be implemented using fixed point numbers, and the peaks could result from an overflow condition during execution of the FFT. Can you attenuate the signal (resp. reduce its amplification) before it is fed into the FFT? If so, you could try to contain the output of the FFT below, say, 5 myV as in the "good" result, and see whether the peaks disappear.
BTW, do you think this large 2..3 Hz spike could be caused by muscle potentials rather than the brain? (In humans, facial muscles often interfere with the EEG data.) In this case, you could just take the spikes as a "warning signal" because the affected data are not reliable anyway.
Hi Jeremy,
Where were the electrodes placed?
Was it telemetry? Could it be simply an ECG interference in the signal since the first peak with the highest amplitude is at around 5 Hz (rr interval mouse is 6-8 Hz)?
I noticed you used Neuroscore for the analysis. How often does it happen to have such a pattern?
Thanks.
Hi Stefano,
One electrode was placed 1 mm lateral to the sagittal suture and 1 mm rostral to Bregma. The other was placed contralaterally 2 mm from the sagittal suture and 2 mm caudal to Bregma.
This is telemetry recording (F20-EET).
This does not happen often, this is the only instance I have noticed this.
- Badges
- Science topic