Artifact detection, thresholds, algorithms

edited August 2017 in Software
Hi there!

I'm using dry electrodes with the ultracortex, and keeping the impedance to less than 5-7 kOhm per site. At what range of uVrms should I set the threshhold to consider the signal an artifact? Can regular brainwaves evoke a band power that exceeds the average to a large degree?

Thanks a lot in advance!
Cheers

Luc

Comments

  • wjcroftwjcroft Mount Shasta, CA
    edited March 2020
    Luc, hi.

    Generally eye blink or scalp muscle EMG signals are huge, sometimes even in the millivolt range (versus microvolt for EEG). You could generally say anything over 100 microvolts is suspect. This type of one dimensional thresholding is simple, but there are more involved methods. Below is a clip from a post I made in 2014 with a few links you can explore.

    ====

    Check out these interesting and simple algorithms for artifact rejection.  In Appendix 1 of this document:

    https://qeeg.pro/wp-content/uploads/2014/10/qEEG-pro-Manual.pdf



    They call it S.A.R.A (Standardized Artifact Rejection Algorithm), developed by Dr. Andre Keizer.  Summary below,

    De-artifacting with S.A.R.A (Standardized Artifact Rejection Algorithm)
     
    When a raw EEG is uploaded to the qEEG-Pro report service portal, it will automatically be de-artifacted using the standardized artifact rejection algorithm (S.A.R.A). Contrary to what some believe, S.A.R.A has nothing to do with ICA / PCA or other reconstruction techniques. S.A.R.A has been developed by dr. Andre Keizer and has gone through a thorough testing process in which the algorithms and parameters were systematically evaluated and refined (download the qEEG-Pro manual for a detailed description of the methods used). S.A.R.A evaluates four different EEG artifact categories:

    Eye blinks
    Horizontal eye movements
    Low frequency artifacts (e.g. head movement)
    High frequency artifacts (e.g. muscle tension)

    When the de-artifacting procedure is complete, you receive a pdf report that shows the raw EEG in which the detected artifacts are marked in the time series. Moreover, the de-artifacted EEG is provided in .edf format so you can view and process your de-artifacted EEG in the software of your choice.

    As part of the de-artifacting procedure an epileptiform episode detection (E.E.D.) algorithm was developed that detects very high-powered, low frequency artifacts that can be elicited by epileptiform activity. When such an episode is detected, a warning will be present in the summary of the S.A.R.A results.

    ----

    There are some much more complex algorithms using ICA (Independent Component Analysis), but the SARA idea looks to be superior and easier.

    http://sccn.ucsd.edu/~jung/Site/EEG_artifact_removal.html

    William
  • wjcroftwjcroft Mount Shasta, CA
    edited April 2020

    I've uploaded a copy of the pdf mentioned in the previous post, to our local forum server. In case it is ever removed from the original link above. Appendix 1 at the end of the pdf shows the "S.A.R.A (Standardized Artifact Rejection Algorithm)".

    William

  • PrageethdaPrageethda Colombo, Sri Lanka

    What type of threshold approaches that can be used to identify the predefined thresholds mentioned in SARA?

  • wjcroftwjcroft Mount Shasta, CA

    Prageethda, hi.

    What I have seen in other signal processing apps (such as Brainmaster, Neuroguide or BioEra), is that the 'noise' threshold amplitude values (in microvolts), can be adjusted during the recording or playback or review -- via one or more slider controls. So that you can dynamically adjust this to suit the individual subject's EEG. You can have a default value(s), that would work in most cases. Then allow dynamic adjustment if needed in a particular case. If you don't have a real-time GUI with adjustable user interface elements -- then the threshold values would be held in easily accessible constants or defines.

    In the case of some of the apps mentioned above, the time series EEG graph shown on the screen during review of the EEG, shows (via dimming or background color), the time slices of the EEG that are being discarded because of artifact. Thus if the thresholds are changed, you can visibly see what additional sections of EEG will be removed or added.

    I do not currently know of any open source SARA code. If you develop such, please post your Github repo.

    Regards, William

  • yundayunda Singapore

    @lagidigu said:
    Hi there!

    I'm using dry electrodes with the ultracortex, and keeping the impedance to less than 5-7 kOhm per site. At what range of uVrms should I set the threshhold to consider the signal an artifact? Can regular brainwaves evoke a band power that exceeds the average to a large degree?

    Thanks a lot in advance!

    Cheers

    Luc

    Hi Luc, can I check how we’re you able to get 5-7kOhm impedance using dry electrodes with Ultracortex? Do you have a link to the electrodes you used? I’m getting in the range of 200-1000+

  • wjcroftwjcroft Mount Shasta, CA

    @yunda, please see this thread for expected impedance range of Ultracortex / dry electrodes.

    https://openbci.com/forum/index.php?p=/discussion/2517/expected-impedance-range-with-ultracortex-cyton-resolved

Sign In or Register to comment.