The original method that produced opposite tracings for left and right EOG channels in sleep recordings (e.g., Rechtschaffen & Kales, 1968) was based upon the placement of the EOG electrodes next to the outer canthi with a reference electrode, between them, just above the nasion.  This placement of the reference electrode causes the two EOG channels to have opposite polarities when the eyes move (but not for delta waves).  Your tracing suggests that the reference being used for the EOG channels is not located between the EOG electrodes.  instead, it is a lateral reference and the signals have the same polarity with regard to that lateral reference.

As you probably know, EOG leads must be linked to the same reference so that we could see "out of phase" movements. Generally, the preferred reference is M2 (right mastoid).

One more thing to remember is that in PSG tracings, the EOG also capture EEG signals.

Considering that the picture is an actual sleep epoch, one possibility is that the deflections that you are seeing are in fact physiological sleep elements, like K complexes.

That's why we should analyze both signals  (EEG and EOG) together.

Dear Dr. Giorelli,

Thank you for your response. You are right about the EEG signals interfering with the EOG signal. I have attached a picture (Capture1) showing this (no visualisation filters applied). How do i fix this? The reference is on the M2. STARSTIM has two reference electrodes one called CMS and the other DRL. I had placed them both on M2 (CMS right above DRL following instructions from STARSTIM wiki).

The reference option on the top corner that you see in the picture lets you choose one of the eight channels in total for the device (in my case Ch1 = LEOG, Ch2= T7, Ch3 = F3, Ch4 = F4, Ch5 = T8, Ch6 =C3 ,Ch7 = VREOG, Ch8= REOG) NOT INCLUDING CMS and DRL. The option also allows you to average electrodes together that you choose to display in the middle but clicking on the right corner for which electrodes you want on display.

I believe that STARSTIM automatically uses CMS and DRL electrodes for reference but i may be wrong.

I am conducting an experiment for my masters thesis and it is imperative that i distinguish sleep stages in real time as i have to apply tACS stimulation at F3,F4,T7,T8 during REM sleep. How would you advice me to go about it since EEG and EOG signals affect each other?

I'm also attaching real time EEG analysis options that the STARSTIM offers in case it can give away when REM stage is reached

I would greatly appreciate if you could please offer some expert advice. I fear that Neuroelectrics' STARSTIM may be unsuitable for the sleep study that i am conducting. It does not allow for direct electrode placement using collodion glue and therefore I'm conducting the study using the EEG cap provided with the machine.

The experiment does not necessarily require an EEG analysis, i am more interested in the effect that the stimulation has in regards to inducing dream lucidity (Voss et al., 2014) as well as the effect of auditory stimulation on dream content. At minimum i require EEG only for distinguishing sleep stages to apply my stimulation protocol

Dear Achilleas,

Just to add a few alternative thoughts. At first sight, my "gut feeling" tells me there's an impedance issue (which reference have you used for the eogs?) with those signals.

In addition, you might also consider that lachrymation during sleep spreads electrolytes over the face, which may interfere with the eog electrodes if not properly isolated. 

Best, 

Amadeu. 

Dear Amadeu,

Thank you for your response. I have updated my response above with this

" I would greatly appreciate if you could please offer some expert advice. I fear that Neuroelectrics' STARSTIM may be unsuitable for the sleep study that i am conducting. It does not allow for direct electrode placement using collodion glue and therefore I'm conducting the study using the EEG cap provided with the machine.

The experiment does not necessarily require an EEG analysis, i am more interested in the effect that the stimulation has in regards to inducing dream lucidity (Voss et al., 2014) as well as the effect of auditory stimulation on dream content. At minimum i require EEG only for distinguishing sleep stages to apply my stimulation protocol "

Is there any way i can reduce this lachrymation effect on the cap?

I would also like to add that EOG signals show up fine when participant woke up. Its during sleep that they exhibit weird signals. The photo i've attached is a snapshot of EOG (LEOG,REOG) activity at 5am of an awoken participant after 3.5 hours of sleep.

Dear Achilleas,

By the look of it, you have properly inverted eog signals when the participant is awake. Is there any other power source (or even an antenna) plugged on exclusively during sleep? You may need to play around with a few things, one at a time. I would start by figuring out how do those reference channels work at different sites.

Regarding lachrymation during sleep, I would expect it to affect the eog electrodes but not a big deal of change on the eegs. Try using Cz as a ref and see what you get.

Best,

Amadeu.

There are no extra signals during sleep... Its such a bizarre issue.

I think it may be time to try a classic brand eeg device and see if there is a difference. I've had 3-4 tries and I get the same issue everytime. Neuroelectrics is a new Spanish startup company and  it might have issues that haven't been worked out yet. Also they are releasing a big software update soon.

Is there some kind of signal processing option/filter that i am neglecting to put on that may be causing this?

When conducting a PSG what software options/parameters have to be ticked/set in order to ensure correct display of EEG/EOG signal?

First I would try to attach the reference electrodes on both mastoids, just as we do on clinical PSGs. If that doesn't work, then try to use Cz as a reference.

secondly,  I would format the montage in a way that both EOG leads are displayed next to each other  (this way is easier to recognize eye movements instead of having EEG leads in the middle....)

then check the impedance and apply the filters (LFF 0.3 Hz and HFF 35 Hz according to AASM ). sampling rate of 500.

Now I have a question to you: if you are using an EEG cap and the machine doesn't accept collodion attached electrodes, which electrodes are you using for EOG? and where in the face?

because standard EOG placement suggests that you attach EOG 1cm out and above (and under on the other side....) of the outer can thus. 

best regards

Andre 

Thank you for your response. Are the reference electrodes you use in PSG called CMS and DRL? CMS is an active electrode and DRL is a passive electrode and together they form a loop. They need to be as close as possible at least to my knowledge.

I agree on formating the montage unfortunately Starstim software doesn't allow you to do that. Afterwards during analysis you can by using other EEG reading software. The reason i have Ch1 as LEOG and Ch8 as REOG is because of their location and wire length. This becomes more clear with this picture http://wiki.neuroelectrics.com/images/thumb/f/fb/Testboardclips.png/250px-Testboardclips.png

I checked mpedance was good and i was getting good signal.

For the EOG cables i use EMG sticky electrodes pads. The placement of the EOG electrodes was put on just like you describe.

It is a weird issue as eog activity behaves weirdly as soon as participant enters the SWS. As soon as participant wakes up without even moving from the bed,  EOG activity returns to normal. I'm currently speaking with Neuroelectrics tech support.

In the meantime I will be now using NEXUS device for EOG activity and HR activity and STARSTIM for tACS stimulation only. Lets see if the issue is corrected.

Dear Achilleas,

I was thinking about your problem again...

I came up with 2 suggestions:

1 - EEG jumpboxes only work fine when all the electrodes plugged are made of the same conductive material (gold, silver, lead, etc..).

You have to check the cap electrodes and the sticky pads and see if they are made of the same conductive material.

The other thing is that commercial PSG machines can use monopolar or bipolar references. When using monopolar references, the devices often have a specific input on the jackbox (an active reference most of the time) and this electrode should be attached "away" from the head (mostly on the neck), but a drawback would be the reference capturing EKG signal that could contaminate the whole tracing.

Generally, when we use an active reference is just one channel and not two, as you described.

2 - The artifact that you are questioning is the presence of "slow waves-like" potentials on EOG, right?

The question is: What do you expect to see in EOG in SWS?

In SWS, the EOG behaves much like any other EEG lead and displays much of the slow waves seen on EEG channels...you don't usually see much eye movements in it like in wake, N1 or REM sleep.

Take a look on this epoch of SWS .

The EOG leads usually follows the same frequencies of the EEG.

Well, I hope this helps...

Andre

The original method that produced opposite tracings for left and right EOG channels in sleep recordings (e.g., Rechtschaffen & Kales, 1968) was based upon the placement of the EOG electrodes next to the outer canthi with a reference electrode, between them, just above the nasion.  This placement of the reference electrode causes the two EOG channels to have opposite polarities when the eyes move (but not for delta waves).  Your tracing suggests that the reference being used for the EOG channels is not located between the EOG electrodes.  instead, it is a lateral reference and the signals have the same polarity with regard to that lateral reference.

Dr. Andre,

I am not looking for EOG signals in SWS sleep. I am looking to distinguish REM sleep by looking at EOG activity.

I have an update on my experiment. I've given up on using STARSTIM for EOG observation. I've been now given Nexus-10 Biotrace to measure EOG activity and HR to distinguish stages of sleep in conjunction with STARSTIM (only for stimulation).

The NeXus doesnt require reference electrodes for EOG (or so i have been told). 

Now i have a different issue, but solving this should finally get clearer EOG data (one can hope).

The interface is very different to the one i am used to. Where exactly do i input the LF and HF filters? Specifically looking at Digital filters

I have attached screenshots of the settings.

Thank you for your help and time

"The NeXus doesnt require reference electrodes for EOG (or so i have been told)."

Thus, it must use an average of all electrodes, or similar, as the reference,

Apparently the machine does have a ground electrode if needed but i think it is used for other measurements such as ECG

The machine uses positive and negative electrodes and thus i've assigned them to VEOG and HEOG electrodes. Is this why it does not need a ground reference?

Any tips/advice on what type of filters and settings to put so as to make rem sleep recognition in real time as easy as possible?

I am attaching more screens of the software

> The machine uses positive and negative electrodes and thus i've assigned them to VEOG and HEOG electrodes. Is this why it does not need a ground reference?

The descriptors "positive and negative" are too vague for an answer.

> Any tips/advice on what type of filters and settings to put so as to make rem sleep recognition in real time as easy as possible?

I assume that this is the low-pass filter that will minimize interference from the mains:  50 Hz for you, and 60 in the U.S.  be sure to check this setting.  I'd use Bessel due to the flat phase delay across the band-passed EEG spectrum.  See http://www.simonbramble.co.uk/techarticles/active_filters/active_filter_design.htm

By positive i mean there are red and black electrodes EOG that average together to form either an HEOG or VEOG signal.

Thank you for all the advice i will implement them all tonight and try.. I will be using EOG and chin EMG. Through looking at the data i will also try use neurofeedback programming markers to alert me for muscle atonia in conjuction with REMS to make monitoring easier.

The biotrace+ software is really flexible but i am still learning it. It would be great if i could manage to implement some sort of algorithm for automatic detection of sleep stages. I have already implemented an algorithm that alerts me when the EOG amplitude is higher/lower than +/-75 as most REMS exceed that. It is not perfect as movement during sleep triggers as well as EEG activity. Its still helpful. I've made a chin EMG neurofeedback function that alerts me when EMG is less than 1mV (http://www.ncbi.nlm.nih.gov/pubmed/18275559) .

The filters i am talking about are the ones recommended for sleep EOG which is 0.3-35Hz. I will use the Besel bandpass filter that you recommended. In addition to this i will apply 50Hz interference filter (as im in Europe).

For Chin EMG i will use 10-100Hz Besel bandpass.

Again, thanks you and everyone who has taken time to help me with my questions, it is very much appreciated. Each time i'm getting a little bit closer to a successful trial and my data begins to get clearer and clearer.

Hi Achilleas

EOG electrodes are useful for identifying REM but you really should be relying primarily on the EEG background. REM has two components, tonic and phasic REM. Phasic REM involves the rapid eye movements and (although not seen in all people) EEG sawtooth waves and phasic bursts on the submental EMG.  Tonic REM involves the "wake" like EEG background and the significant reduction in submental EMG activity.  It will help you study if you are able to identify both. 

Hi Achilleas,

Regarding the Starstim traces you show, you are using "Reference: none", which means that your are using the CMS electrode (presumably attached to a mastoid).  So, each trace provides you the potential difference between the electrode and the mastoid. If you want another reference (such as the one discussed by James above), you should place another electrode there, e.g., above the nasion. 

I suggest you contact [email protected] for further help. There is no reason to use another EEG system (providing EEG and stim is the nice thing about Starstim) unless you need more channels (I understand you have a Starstim 8).

Hope that helps!

Giulio

Yes i realized this might have been a big issue. I'm still not understanding why even with the referencing on the mastoid, EOG/EEG activity is normal when participant is awake (or eyes closed), and goes haywire as soon as participant enters first few sleep stages (and goes normal again when participant is woken up again and still in bed in same position).

Another issue issue with STARSTIM is that it doesn't allow for different filters for each electrode during real-time observation. Changing the filter changes it for all electrodes. This could potentially be a problem if i want to monitor Chin EMG activity as filters should be 10-100Hz (while EOG and EEG is 0.5 - 35Hz). Unfortunately NIC doesnt allow for a lot of customizability. I doesn't really show clearly amplitude levels across different electrodes clearly. Nexus allows you to build your own screens and add customizable neurofeedback functions.Additionally the STARSTIM 8 channel we have is an issue as i need more channels. Finally it has crashed on me while monitoring participants. I've had to come up to the participant while sleeping and turn on and off the device (while on their head and trying not to wake them up) for it to work again. Hopefully the upcoming NIC 2.0 addresses at least the crashes.

Unfortunately my University does not have Neurosurfer so i cant use feedback to make monitoring easier. What would be great is if Neurosurfer could be made to start stimulation automatically if certain parameters are met. E.g REM EOG activity, Muscle atonia and EEG REM activity. It would make my sleep study so much easier!

This is the current set-up i have now. Please inform me if you see any issues with it:

I use NeXus-10 to monitor EOG (0.3-35Hz) and chin EMG (10-100Hz) activity. LEOG AND REOG on outer canthus. NeXus-10 ground electrode just above nasion.

I use STARSTIM to monitor EEG activity and stimulation. F3, F4,T7,T8 tACS stimulation (and EEG as STARSTIM allows that), and C3,C4 EEG and CMS/DRL on mastoid and Cz for reference (filter 0.5-35).

I will be monitoring C3,C4,HEOG and Chin EMG for REM signs. . I will be testing on Thursday. Last week i tested the Nexus-10 with just EOG and EMG and was able to detect REM during the night and got confirmation from my participant. Hopefully it'll all go well!

Still working on this? It's trivial to detect REM using eye movements provided you place the leads correctly. I have a tool that may help you. I PMd you in case you are still working on this.