I recently started running in-vivo electrophysiology experiments in mice, and unfortunately many of my recordings have been plagued by a persistent noise artifact which I cannot seem to get rid of (picture 1). It is characterized by small sharp peaks in the raw EEG with a frequency roughly between 5-15Hz, but seems to affect the power spectral density in higher frequencies (described below).
To briefly describe our setup:
- We implant mice with 6 skull screws connected to copper wire (PVC hook-up wire) soldered to electrode interface boards (Neuralynx EIB16). Four channels (two anterior, two posterior) serve as our EEG channels, with the other 2 channels as ground (occipital).
- We do 5hr recordings in a sound-proof chamber at least 5 days post-implant and then compute power spectral density and connectivity & relate to behavior.
The resulting PSDs look something like this (picture 2), with most of the noise affecting higher frequencies (varying in severity), perhaps due to harmonics. We have used independent components analysis with some success to remove artifacts before computing PSDs, but I’d much prefer to collect clean data.
We’ve tested our hardware with known input signals, and nothing seems amiss. So, we originally thought the issue was related either to our electrode construction or surgical technique. After switching from solid-core silver wire to stranded copper, the artifact appears less frequently, though it still reappears in some animals. We also record during implants to anticipate any noise issues before cementing the electrodes in place. Additionally, the artifact appears independent of movement.
In summary, I haven’t been able to diagnose the source of a noise artifact in my mouse EEG recordings after various troubleshooting efforts. We believe it may be heart rate noise (mouse HR is ~300-800bpm, so 5-12Hz), but I’m open to ideas about what this could be and how to fix it.
Has anyone dealt with this issue (or similar artifacts) before? If so, what could I try to avoid seeing such noise in the future?
Happy to provide any other details (a few more technical aspects are described below). Please help! Thank you for your time.
...
More technical details: our connectors have 16 channels, so remaining 10 channels (non-EEG) are tied to ground. The boards are connected to an omnetics adapter (ZCA-OMN16), clipped into an analog head stage (ZC16), which is then connected to a preamplifier (PZ5) and then our DSP (RZ5D).
It looks like you have a good deal of 60hz noise. I don't know about the spikes, but maybe consider shielding.
Just check it against the heart rate and see if the frequency matches. I have only ever dealt with human EEG, and this suspiciously looks like an ECG artefact.
See the article below I'd written a few years back.
I think varying the position of the reference might help. Some subject's ECG vector is in a weird direction causing it to be picked up with standard electrode placements. Try setting it 90 degrees to the imaginary vector.
If you're covering the head with electrodes, it would be easy to see which electrodes are picking it up most and can give you a generalish direction of the ECG vector.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Electrodes