[Eeglablist] Consequences of manual rejection on continuous data

[Alexis Leiva]

Hi Bryan,

Thank you so much for your answer, I appreciate it. Firstly, I want to
clarify that my principal doubt is how the manual rejection of a noisy
segment from the continuous data in EEGLAB works. I ‘m not sure if this one
is a kind of “cut and splice” implementation or if it is an algorithm that
smooths the clean edges that splice after the removal of noisy segments to
avoid possible future distortions, new sharp nonstationary segments in the
signal, phase-slip data, etc. Secondly, I’ve been performing this manual
rejection of bad segments and plotting my power spectrum as a routine
analysis; however, a few researchers who are not so familiar with
continuous data analysis (i.e., they are used to epoching data) asked me if
this manual rejection method is not a cut and splice implementation due to
possible negative impact on the power spectrum outcomes. Therefore, that
observation was the origin of my doubts. Thirdly, answering your questions:



1)      Noisy segments last between <1 second up to 2 or 3 seconds. In the
3-second case, noise originated from coughing or the participant’s
movement. Moreover, the noisy segments are within the 4-minute block. An
example is shown in this link: https://urldefense.com/v3/__https://ibb.co/wZzd46wN__;!!Mih3wA!FOfODmcqJp_J1Gv4ROz2n7uQ2DseZE_sZR8ImoiDEYk04Gqb0PVPtFE8d5-cY1b7PCovNslvV5m7eVJ3iCoQgvE$ 

2)      Yes, the rejection method is used within a 4-minute block.

3)      To calculate Power spectrum analysis, I have been using a
continuous 1s Hann window with 50% overlap through Welch’s method across
the entire 4-minute block.

4)       To answer this question, I performed a mean PSD calculus from
different cortical areas from 12 participants after EEGLAB manual rejection
on the continuous data. Apparently, the power spectrum curve is
acceptable. However,
I want to be sure that what I have been doing is right and that there are
no unnoticeable artifacts with the level of analysis I’ve performed.

Link PSD curves: https://urldefense.com/v3/__https://ibb.co/Zzqd50ky__;!!Mih3wA!FOfODmcqJp_J1Gv4ROz2n7uQ2DseZE_sZR8ImoiDEYk04Gqb0PVPtFE8d5-cY1b7PCovNslvV5m7eVJ3Pcv6tLc$ 


Thank you for your time

Best,

Alexis




El mar, 4 feb 2025 a las 15:20, Bryan Hall (<bryathlon at gmail.com>) escribió:

> Hello Alexis,
>
> Can you provide a little more information about your method and analysis?
>
> 1. You have 4 minute blocks…what is the resolution of a segment? Are the
> noisy segments within the 4 minute block?
> 2. When you remove noisy segments and then “cut and splice” is this within
> a 4 minute block?
> 3. What resolution are you targeting for power spectrum analysis?
> 4. What does your power spectrum analysis process look like in detail?
> Even band pass filters can add artifacts around edges.
>
> The short answer is that I think you are definitely adding noise with “cut
> and splice” but it may not cause you an issue depending on what and how you
> are doing your analysis. One way to largely avoid this issue to calculate
> PSD for each continuous segment (cut) and don’t splice. Instead recombine
> the PSD for each segment back into a single answer for your 4 minute block.
>
> There are much smarter EEG minds here than me however so take my advice
> with a grain of salt.
>
> Talk to you soon,
>
> Bryan
>
> > On Feb 3, 2025, at 3:23 PM, Alexis Leiva via eeglablist <
> eeglablist at sccn.ucsd.edu> wrote:
> >
> > Dear all,
> >
> > I am performing an experiment with 4-minute blocks to analyze the power
> > spectrum from the entire block but not in epochs. Then, when I manually
> > reject a noisy segment from the continuous data in EEGLAB, a boundary
> mark
> > event appears in the channel data.
> >
> > Now, my question: Is this method of rejection and the boundary mark
> > apparition in EEGLAB a “cut and splice” implementation, inserting a kind
> of
> > imbalance into the signal after splicing the pre- and post-noise
> segments?
> > Knowing the possible consequences of manual rejection of my continuous
> data
> > is critical because I don’t know if this implementation might cause a
> > non-evident discontinuity of my data, some distortion, unreliable
> outcomes,
> > or another kind of issue on the power spectrum calculus through Welch’s
> > method on the entire blocks.
> >
> > I appreciate your help.
> >
> > Regards,
> >
> > Alexis
> >
> > --
> > Alexis Leiva Catalán
> > Tecnólogo Médico  mención en ORL Universidad de Chile.
> > Candidato a PhD Neurociencias Pontificia Universidad Católica De Chile.
> > _______________________________________________
> > To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu or visit
> https://sccn.ucsd.edu/mailman/listinfo/eeglablist.
>
>

-- 
 Alexis Leiva Catalán
Tecnólogo Médico  mención en ORL Universidad de Chile.
 Candidato a PhD Neurociencias Pontificia Universidad Católica De Chile.