3
$\begingroup$

I recently start to work on sleep study.
For my research i download sleep EEG data from physionet. The EEG data has 100 HZ sampling rate and was recorded from 2 bipolar EEG site.

When i start the preprocessing stage, i encounter a simple problem, how should i know if my signal has an artifact or noise?

It should be noted, based on nyquist theorem and my signal's sampling rate, the maximum frequency of my signal is 50 HZ, so i did not filter unnecessary EEG frequency.
In general i only used a simple notch filter at 50 HZ, and used simple threshold method in order to remove the epochs that were grossly contaminated by muscle and/or eye movement artifacts.

Back to the main question, how should i know if i need to uses more complicated method for removing EMG or EOG artifact from my signal?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
3
$\begingroup$

Your recording necessarily has noise. That's a property of any physical apparatus, and you are recording the activity of noisy neurons. The other property of noise is that, by definition, it will average out. So don't do anything about that. 100Hz is quite a low frequency. Artifacts (e.g. blinks, but I doubt you would have that in sleeping subjects) would typically have a higher frequency than that so (1) it's unlikely it has been recorded by your system at all, (2) you can low-pass filter your signal to remove both the 50Hz artifact and possible other artifacts. Other artifacts such as sleeping position would have very low frequencies. So you can high-pass your signal to get rid of them. I'm not familiar with sleep EEGs but I would not worry too much about it. You can also check few papers and see how they filtered their signal.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Thank you, i checked some related papers and most of them used higher sampling rate, so they high-pass and low-pass filtered their signal, as you mentioned i don't need to low pass my signal but i think as your recommendation i should uses a high-pass filter at least.In some of the paper,they used ICA in order to remover blink and movement artifact but some other only checked that epochs were not grossly contaminated by artifacts by using threshold method which i used too but my main concern was that if i should used complicated method such as ICA to remove noise in first place or not? $\endgroup$
    – maia
    Dec 3, 2017 at 7:08
  • 1
    $\begingroup$ Sorry just saw that. Filtering and ICA are 2 very different techniques. ICA tries to find underlying dimensions that explain your data better. Because noise will be mainly independent on your electrodes (thus have a different effect on each electrode) while blinks will have a similar effect on all electrodes, blinks will strongly contribute to the variance and will be easily picked up by the ICA. Then you can subtract that dimension from your data and get a "blink-free" signal. In principles, it is sound although I would personally trust filtering more. $\endgroup$
    – user17122
    Aug 2, 2018 at 1:14
  • 1
    $\begingroup$ In any case, ICA requires multiple dimensions (electrodes) and you have only 2. So even if you wanted to, it's not sure you could use ICA at all. If your dataset is manageable, I think thresholding and manually checking trials is your easiest solution. $\endgroup$
    – user17122
    Aug 2, 2018 at 1:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.