[Eeglablist] Fw: Re: ECG Channel Inclusion in ICA and ICLabel Steps

Tue Aug 6 06:59:12 PDT 2024

Hi all, 
Just chiming in to say that this discussion is very timely (I sent an email to Arnaud Delorme about a week ago asking for guidance re: whether or not to include ECG channels in ICA), and I hope a consensus opinion from the ICA gurus will be posted to the EEGlab wiki page!
Thanks
Amanda McCleery, Ph.D. (she/her/hers)
Assistant Professor, Department of Psychological and Brain Sciences
The University of Iowa
473 Psychological and Brain Sciences Building, Iowa City IA 52242 

-----Original Message-----
From: Makoto Miyakoshi <mmiyakoshi at ucsd.edu> 
Sent: Monday, August 5, 2024 4:54 PM
To: EEGLAB List <eeglablist at sccn.ucsd.edu>
Subject: Re: [Eeglablist] Fw: Re: ECG Channel Inclusion in ICA and ICLabel Steps

Hi Cedric and John,

> It does not inherently assume that all sources must be of the same 
> nature
or originate from EPSP/IPSPs from pyramidal neurons in cortical gyri, does it?

I mean "If ICA were to inherently assume that all sources must be of the same nature or originate from ERSP/IPST from pyramidal neurons in cortical gyri", it would not identify non-brain artifacts.

If the heartbeat-evoked potential is completely synchronized with the ECG, ICA must capture the mixed cortical activation pattern as a single component and there is no way to separate the cortical contribution from the heart's contribution.

John, what's the relation of the timings between ECG and HEP--are they always synchronized (i.e. the functional connectivity), or the latter delays/jitters the onset of ECG (the effective connectivity)? ICA may capture the former under favorable conditions, but it cannot capture the latter.

Makoto

On Mon, Aug 5, 2024 at 4:55 PM Cedric Cannard via eeglablist < eeglablist at sccn.ucsd.edu> wrote:

> Hi Makoto,
>
> Sorry for the typo. As you can guess, I meant "I do not believe 
> including auxiliary channels violates ICA assumptions."
>
>
> > Tony Bell is a legend
>
> haha :)
>
>
> > For example, concatenating millivolt-range signals with 
> > microvolt-range
> signals for decomposition may cause unexpected trouble.
>
> Yes, I generally rescale the ECG signal to the EEG range if the ECG 
> scale is vastly different, so ICA does not prioritize the larger ECG 
> signals over the smaller EEG ones, which gives better results.
>
>
> > Then ICA would not identify non-brain artifacts, right?
>
> Why? Non-brain artifacts have independent sources. My point was that, 
> as you summarized, ICA assumes sources to be independent, mixed, and 
> non-gaussian, not that the sources of all signals are EPSP/IPSPs from 
> pyramidal neurons.
>
>
> > I'm rather curious to know if heartbeat-evoked potential is a real
> thing. For example, if I decompose a heartbeat IC which typically has 
> a broad and dipolar projection from the depth. Is there any neuronal 
> contribution there? If yes, then the heartbeat IC is contributed by 
> both the heart and the brain. If no, it is generated only by the 
> heart. Which is the case?
>
> I agree and am wondering the same thing. And I don't think there is an 
> answer to this yet. removing heart ICs may remove both the CFAs and 
> the relevant neuronal-related HEP. Ideally, heart ICs would capture 
> the cardiac field artifacts (CFA) which travel by volume conduction 
> (e.g., skin) and are picked up by scalp EEG electrodes. The broad and 
> dipolar spatial properties of the heart ICs reflect these CFAs (i.e., 
> the source is non-neuronal). And HEPs (after removing these CFAs) 
> would only reflect neuronal processing of heart activity.
> I think one potential way of answering this would be to have 
> intracranial electrodes time-locked to scalp electrodes, mark the 
> R-peaks in both, and compare. But even then, heart activity can travel 
> to the brain via various direct and indirect pathways: directly via 
> the vagus nerve, indirectly via volume conduction through the spinal 
> cord (surrounded by conductive CSF) and conductive arteries, or via 
> blood pressure-related activity detected by baroreceptors/chemoreceptors.
>
> I am currently developing a multivariate method that would provide 
> more information on the synchronization (coherence & partial 
> coherence), system response (transfer function matrix, spectral 
> matrix, inverse spectral matrix), and causal interactions (directed 
> coherence, directed transfer function, partial directed coherence) between heart and EEG signals.
> hopefully, that can provide some insights (direct, indirect 
> interactions and direction of the interaction).
>
> I think what provides some answer is the growing body of literature 
> showing meaningful associations between HEP and interoception and 
> integration of internal bodily states, highlighting brain areas known 
> to be involved in these processes from the neuroimaging literature, 
> e.g. insula, ACC, somatosensory cortex.
>
> Some good references for HEP:
>
> https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/31051293/_
> _;!!Mih3wA!A7XPAnqNlQiz9pvotlbJTJbG2_n0TrNLB20H1cJLZDjdmMz8JsfHFIdXHQu
> fx-wO2xcvGmCGkuOn8VQfYj8XKL4HCQ$
>
>
> https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/22541740/_
> _;!!Mih3wA!A7XPAnqNlQiz9pvotlbJTJbG2_n0TrNLB20H1cJLZDjdmMz8JsfHFIdXHQu
> fx-wO2xcvGmCGkuOn8VQfYj8D2qUf-A$
>
>
> https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/28651745/_
> _;!!Mih3wA!A7XPAnqNlQiz9pvotlbJTJbG2_n0TrNLB20H1cJLZDjdmMz8JsfHFIdXHQu
> fx-wO2xcvGmCGkuOn8VQfYj9BgOZjfw$
>
>
> Cedric
>
>
>
> ------- Forwarded Message -------
> From: Makoto Miyakoshi via eeglablist <eeglablist at sccn.ucsd.edu>
> Date: On Monday, August 5th, 2024 at 7:59 AM
> Subject: Re: [Eeglablist] ECG Channel Inclusion in ICA and ICLabel 
> Steps
> To: EEGLAB List <eeglablist at sccn.ucsd.edu>
>
>
> > Hello John and
> >
> > Cedric
> >
> >
> > ,
> >
> > This double negative makes it difficult to understand what you are 
> > trying to say, Cedric.
> >
> > > I do not believe including auxiliary channels does not violate ICA
> >
> > assumptions.
> >
> > You are saying that including ECG violates ICA's assumption, but 
> > your claim seems opposite.
> >
> > ICA's assumptions are, off the top of the head,
> >
> > - The sources are non-Gaussian
> > - The sources are linearly independent
> > - The sources are linearly and instantly mixed
> > - Tony Bell is a legend
> >
> > So ECG channels may be included in ICA with no problem in theory.
> However,
> > if empirical observations recommend otherwise, there may be some 
> > reason that is specific to each application which is worth 
> > investigating. For example, concatenating millivolt-range signals 
> > with microvolt-range
> signals
> > for decomposition may cause unexpected trouble.
> >
> > > It does not inherently assume that all sources must be of the same
> nature
> >
> > or originate from EPSP/IPSPs from pyramidal neurons in cortical 
> > gyri,
> does
> > it?
> >
> > Then ICA would not identify non-brain artifacts, right?
> >
> > I'm rather curious to know if heartbeat-evoked potential is a real thing.
> > For example, if I decompose a heartbeat IC which typically has a 
> > broad
> and
> > dipolar projection from the depth. Is there any neuronal 
> > contribution there? If yes, then the heartbeat IC is contributed by 
> > both the heart and the brain. If no, it is generated only by the heart. Which is the case?
> >
> > Makoto
> >
> >
> >
> > On Mon, Aug 5, 2024 at 12:58 AM Cedric Cannard via eeglablist < 
> > eeglablist at sccn.ucsd.edu> wrote:
> >
> > > Dear John,
> > >
> > > This is the same debate as including EOG for separating and 
> > > extracting ocular components (isn't that common practice?). The 
> > > only reason why
> you
> > > can get away without EOG channels is that ocular activity is so 
> > > large compared to EEG, so easy to identify as an independent 
> > > source. However, cardiac artifacts are difficult to identify 
> > > without an explicit
> reference
> > > signal (except occasionally in some individuals where you can see 
> > > them visually in the raw time series).
> > >
> > > Someone will correct me if I'm wrong, but I do not believe 
> > > including auxiliary channels does not violate ICA assumptions. ICA 
> > > is an
> algorithm
> > > designed to separate a mixture of signals into statistically
> independent
> > > components. It does not inherently assume that all sources must be 
> > > of
> the
> > > same nature or originate from EPSP/IPSPs from pyramidal neurons in
> cortical
> > > gyri, does it?
> > >
> > > By providing a clear reference for ECG artifacts, ECG channels can
> enhance
> > > the accuracy of this separation in my experience, in line with 
> > > Sofia's observation. In fact, this is a method supported by my new 
> > > BrainBeats plugin (along with HEP/HEO and HRV analysis which 
> > > implements validated algorithms and guidelines via command line or 
> > > GUI). See here for a step-by-step tutorial:
> > >
> https://urldefense.com/v3/__https://www.jove.com/t/65829/author-spotli
> ght-advancing-study-brain-heart-interplay-with__;!!Mih3wA!CWCJFZKu2zYD
> 1n7sKo1ur_bO9g_iWhAy9vSTS18Bf-jfmH6errEXOVCDaXANh_5ukcj1Wajfh_MIn3CPFG
> GRV09qjA$
> > > And Github repo:
> > >
> https://urldefense.com/v3/__https://github.com/amisepa/BrainBeats__;!!
> Mih3wA!CWCJFZKu2zYD1n7sKo1ur_bO9g_iWhAy9vSTS18Bf-jfmH6errEXOVCDaXANh_5
> ukcj1Wajfh_MIn3CPFGGKS9yf5A$
> > >
> > > Sofia, my question is: Why are you using ICA in the context of 
> > > HEP? You generally want to keep ECG-related activity for HEP. Do 
> > > you suspect
> the HEP
> > > is a CFA? I have wondered about that but haven't found a work-around.
> > >
> > > Curious to hear more on this!
> > >
> > > Cedric Cannard
> > >
> > > On Thursday, August 1st, 2024 at 7:08 AM, Richards, John via
> eeglablist <
> > > eeglablist at sccn.ucsd.edu> wrote:
> > >
> > > > Don't include ECG in the EEG signal for ICA. For various 
> > > > technical reasons (amplitude, frequency, signal spikes, 
> > > > location) and
> theoretical
> > > > ones (its NOT a brain signal; you want to correlate ECG and EEG).
> > > >
> > > > See poster by Wanze Xie at ICIS, or contact him. He has 
> > > > synchronized cardiac cycles and ERPs. His techniques might be 
> > > > useful and he has
> some
> > > > very interesting VEP/cardiac cycle data.
> > > >
> > > > John
> > > >
> > > > ***********************************************
> > > > John E. Richards
> > > > Carolina Distinguished Professor Department of Psychology 
> > > > University of South Carolina Columbia, SC 29208 Dept Phone: 803 
> > > > 777 2079
> > > > Fax: 803 777 9558
> > > > Email: richards-john at sc.edu
> > >
> > >
> https://urldefense.com/v3/__https://jerlab.sc.edu/__;!!Mih3wA!B4rXaMGV
> hHYr_2oHmmQc94PJcfhiYYPWsZNqJtpm0CRKhOTeHaXWrkUCGI-mcxaW1b2qR1YWJv1lJr
> PrNV29FIao1z4gCQ$
> > >
> > > > ***********************************************
> > > >
> > > > -----Original Message-----
> > > > From: eeglablist eeglablist-bounces at sccn.ucsd.edu On Behalf Of 
> > > > Sofia Amaoui via eeglablist
> > > >
> > > > Sent: Thursday, August 1, 2024 8:54 AM
> > > > To: eeglablist at sccn.ucsd.edu
> > > > Subject: [Eeglablist] ECG Channel Inclusion in ICA and ICLabel 
> > > > Steps
> > > >
> > > > Dear all,
> > > >
> > > > We are conducting a Heart-Evoked Potential (HEP) study using EEG
> signals
> > > > that are time-locked to R-peak via ECG. Our preprocessing 
> > > > pipeline
> is being
> > > > implemented with EEGLAB, including the ERPLAB and HEPLAB 
> > > > plugins. As
> part
> > > > of this process, we employ Independent Component Analysis (ICA) 
> > > > and
> ICLabel
> > > > to identify ocular and cardiac field artefacts (CFA).
> > > >
> > > > A key question has arisen regarding including the ECG channel in 
> > > > the
> ICA
> > > > and ICLabel steps. Our preliminary findings show that when we
> include it,
> > > > CFA classification is much higher.
> > > >
> > > > We are seeking your insights on whether the inclusion of the ECG
> channel
> > > > is aiding in the identification of CFA components or if it might 
> > > > be compromising EEG information of interest by focusing on 
> > > > identifying
> and
> > > > removing CFA.
> > > >
> > > > Thank you for your time and assistance.
> > > >
> > > > Best regards,
> > > >
> > > > Sofia Amaoui,
> > > > _______________________________________________
> > > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > > To unsubscribe, send an empty email to 
> > > > eeglablist-unsubscribe at sccn.ucsd.edu
> > > > For digest mode, send an email with the subject "set digest 
> > > > mime" to eeglablist-request at sccn.ucsd.edu 
> > > > _______________________________________________
> > > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > > To unsubscribe, send an empty email to 
> > > > eeglablist-unsubscribe at sccn.ucsd.edu
> > > > For digest mode, send an email with the subject "set digest 
> > > > mime" to eeglablist-request at sccn.ucsd.edu 
> > > > _______________________________________________
> > > > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > > > To unsubscribe, send an empty email to 
> > > > eeglablist-unsubscribe at sccn.ucsd.edu
> > > > For digest mode, send an email with the subject "set digest 
> > > > mime" to eeglablist-request at sccn.ucsd.edu
> >
> > _______________________________________________
> > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> > For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to 
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to 
> eeglablist-request at sccn.ucsd.edu