Neuroscan FAQ - Electrical and Optical Neuroimaging

During acquisition the bandpass window should be opened as wide as possible. Sampling at 10 kHz, the filter band should be DC-2000. This is the main reason for sampling at such a high rate, is to enable complete sampling of the EPI artifact and prevent aliasing of this artifact.
If manually inserting markers into a CNT file, never use "Annotation" - these are basically worthless. Either use keyboard markers (by using the "Mark Point" function and then using one of the function keys) or the Ctrl-V keys (this will let you insert a stimulus code).
Recording Sleep with a 64 electrode Quikcap or MagLink cap:
There is no limitation in the Synamp2 hardware that will prevent recordings longer than 3 hours; in fact with a fully-configured Neuroscan acquisition system (PC from us, amplifier, Scan 4.3.3) we have been able to run several acquisitions overnight. We have dealt with instances where Synamp2 users would encounter "read data errors" or "read device errors" after a period of time, but from our experience this is related to a seemingly-insignificant attribute on the PC.
For example, in one instance we discovered that turning off the local HD's fast-indexing service enabled overnight recordings. Another way to bring recordings to more manageable file sizes is to record at 5kHz or 10kHz, rather than 20kHz, but this depends to some degree on the kinds of artifact or EEG features you're specifically interested in.
This is not to say that we automatically believe this issue is purely indicative of the PC; it is very possible that some failure exists in your headbox or system unit. However, this should be repairable; there is no redesign nor modification necessary of the existing hardware, in order to enable overnight recordings. They should be possible with your current system.
I can't see sleep spindling or k-complexes. Why?
Well, if we take the sleep spindle, and decimate it, we essentially are changing the sampling rate of the spindle. Thus we are attempting to redraw the same waveform using a subset of data points. The resolution will not be as good, but hopefully we will have enough after decimating to redraw accurately. However, if that sleep spindle is made of frequencies up to, say, 700 Hz, and we decimate it to 1000 Hz, we will not be able to redraw the oscillating waveforms above 500 Hz. Thus we now are no longer accurately redrawing that same line.
In other words, the effect of decimation should not be problematic if you maintain the Nyquist rule: sample at least 2x that of the highest frequency needing to be measured. Thus, to answer your question correctly, we need to know the frequency characteristics of the sleep spindles and the k-complexes.
The decimation effects would basically depend on the frequency response of k-complexes and sleep spindles. Specifically if you did a Fourier decomposition of that waveform (the time series), this would reveal the contribution of waveforms, each oscillating at a particular frequency, in terms of its amplitude. This amplitude would be shown for multiple oscillatory frequencies, e.g. 1, 2, 5, 10, 27, 62 Hz etc. Essentially if you took each oscillatory wave, multiplied it by its amplitude, and summed those together, you should see the original time series. Now, when you perform this Fourier decomposition, there is a requirement on the number of points necessary in order to recreate an oscillatory waveform at a given frequency. This requirement is basically the oscillation frequency multiplied by two. If you picture a 1 Hz sine wave, starting with an amplitude of 0 uV at 0 seconds and ending at 0 uV at 1 second, you can see that two points are necessary in order to draw the sine wave: one at the initial "valley (occurring around 0.25 s) and one at the "peak" (occurring at 0.75 s). Basically in any oscillatory waveform, in order to construct it, you need a point for each "valley" and a point for each "peak." Since there is one peak and one valley for each Hz, or revolution, this equals two necessary points for each Hz. So for a 5 Hz waveform to be drawn, you need...10 points.

Edit

"Scan was unable to detect a valid software license on this computer. The application has been halted until it is activated. You will need to contact support in order to Update your software lock." I have a valid License Key but it's not accepted! (1) Unplug your license key and close all local instances of Scan. Log in to your local machine as admin (can't be done on a networked machine right now.) (2) Download the hotfix for Scan 4.3.3 here and the Sentinel License Key Reader upgrade here. (3) Run the executables (as admin!), then reboot the machine. (4) Plug in your license key. (5) Open Scan. IF IT STILL DOESN'T WORK, contact tech support.
Can Edit track the steps used to create a certain file? Beyond the information contained in the history bar (which can be scrolled through, or expanded, during a session of Edit though if you close the program it disappears, no. Unfortunately. However, filenames can be very long, and so our lab tracks processing by appending suffixes to each output filename to prevent overwrite, and recall what was done.
When should I decimate? fMRI gradient reduction --> advanced options has a dropdown to let you decimate the data during the gradient reduction step, to get the best anti-aliasing possible. We used to decimate before reducing gradient, and ran into trouble. Beware your Nyquist limits.
Gradient Reduction
The Neuroscan Recommended technique for performing fMRI artifact reduction goes as follows, not to say that other techniques are invalid; as long as a given technique has both scientific and mathematical validity, it can be classified as a viable method for artifact reduction. Thus, my goal is to explain not only our technique, but the logic behind the procedure. This should help to clarify whether any differences exist between what we recommend, and what you have been implementing.
Essentially, the artifact introduced into an EEG signal from the MR from a given slice can be broken down into particular components (i.e. from rapidly-changing gradients), where each component is relative to a particular action taken by the MR in the process of performing the slice measurement. These components remain rather consistent from trial to trial. If you were to take the artifact signal produced from a single slice, or even those inside an entire TR block, and performed a FFT you would see that there is no distinct range of frequencies that comprise this artifact. In fact, you would notice that the Fourier distribution would contain several very narrow-band spikes. Thus, it is difficult to identify a cutoff frequency, or a filter in general, that would be ideal in assisting with removal of the scanner artifact.
Note that the artifact signals are separate and distinct from the EEG. Therefore we have the ultimate goal of how to eliminate one, and keep the other intact. This is another reason why filtering the data beforehand is not recommended, because several of the aforementioned "artifact spikes" shown in the Fourier transformation can be found inside EEG bands of interest. That alone causes difficulty in determining an appropriate cutoff. Also, filtering the signal beforehand can affect the morphology of the signal produced by the artifact, making further removal of the artifact problematic.
This latter point is the most important. In the fMRI artifact reduction used by Scan, the user enters in necessary parameters that enable Scan to know the onset of TRs, along with the duration of TRs and the number of slices within each TR. The user can also indicate the number of TRs/slices to be used in the construction of an averaged artifact template. For example, you may have TRs 2 seconds in duration, and decide to select 9 TRs to be used in the calculation of the artifact template. The act of averaging 9 TRs together enables a twofold accomplishment:
1.) It averages to zero the underlying EEG;
2.) It also averages out any noise, thus leaving a "pure" artifact signal.
This averaged artifact template is most accurate when the original signal is sampled at a very fast rate, typically 10 or 20 kHz. The filter parameters set during acquisition should be as wide open as possible, enabling a pure recreation of the artifact from the gradients. If the artifact is filtered beforehand, the windowing affects of the filter may cause the average template to become smeared, thus preventing maximal removal of the artifact. This is the main reason why we do not recommend filtering the signal before applying the artifact reduction.
The artifact reduction in this version of Scan also encapsulates two follow-up steps: filtering of the artifact-removed signal, and decimation of data. These two steps are actually related; in order to decimate the data, the data must also be filtered to prevent aliasing.
This filter step also removes any components of artifact, and noise at higher frequencies, not completely removed by the artifact template subtraction. This last step may be problematic in attempting to measure ABRs during a continuous, simultaneous EEG-fMRI acquisition, because some artifact parameters still remain even after the subtraction takes place. I will be bluntly honest: I have not performed a subsequent analysis on the data post-correction, but pre-filtering, to see if the remainder artifact/noise has a more narrow distribution, which would allow for a bandpass filter. Normally people using the Maglink are looking at ERPs whose frequency distribution is much lower than that of remainder noise/artifact; thus a filter with a cutoff of 70 Hz, or even 40 Hz, is typically sufficient.

What are *.ev2 files about? Basically, an event file is a text format list of events that Scan can use for epoching in lieu of actual events within the file itself. In your case, I'll assume that you have a CNT file with zero events, i.e. there are no event markers at the bottom of the data window. While you could use a batch (.TCL) file to insert events, an EV2 file simply contains a list of events that can be used across multiple data files for epoching.
Essentially, Scan reads the location of an event from this EV2 file; Scan then goes to that exact location within the CNT file, and uses this as the zero point for an epoch. The epoch is then made based on your X-min/X-max criteria. It's basically the same as if you had specified "internal/port" as the mode for epoching, except that in this case Scan is not reading for the events in the CNT file itself; it's using the event file as the list of events.
For example, let's say you have a CNT file sampled at 1000 Hz. In your case, you want to parse the entire file into sweeps of 1024 data points each, with 25% overlap (an overlap of 256 points). Thus, you can dictate in the event file that the 1st event occurs exactly at the end of your 1st sweep (1024 points). Then, you set your X-min to -1024 and your X-max to 0. Your first sweep then is the first 1024 milliseconds (time 0 to time 1024). Now, in the event file, you would set the onset of your 2nd event as 1792 (garnered by adding 768 to 1024). When you epoch that data point using -1000 to 0, your 2nd epoch starts at time 768 and ends at 1792. Thus, you have a 25% overlap of the 2nd sweep to the 1st sweep. Your 3rd event would then be indicated with an offset of 2560 (1792 + 768). Each succeeding event would have an incremental offset of 768.
The event file contains 5 columns: first: ordinal number of the event (1st event =1, 2nd event =2, etc.)
2nd: actual event code. Thus you can specify multiple codes in the file, although in this case we only need 1 code.
3rd: The next 3 columns can be used to associate response information with the event; this column is the code of the response itself.
4th: This column is the accuracy of the response (1=correct response; 0=incorrect response; -1=incorrect rejection to respond)
5th: This indicates the latency of said response to the stimulus, in seconds.
6th: The most important - this lists the actual location of the event. This can be in either seconds or data points, and indicates the EXACT onset of the event from the beginning of the data. if you sampled at 500 Hz, an event with an offset of 5000 data points would be read at the 10-second mark within the CNT file.
With this format, you can easily make a file in Excel, save it as a text tab-delimited file with an .EV2 extension. Column headers are not necessary. Attached to this email is a full event file made with the parameters described above; it can then be adjusted for use on CNT files with a different sampling rate.
Another nice feature of using EV2 files for this purpose is that it lets you only specify a certain section of a CNT file for analysis. Using the above example, if we only wanted to epoch within a time block starting at 10 seconds and ending at 4 minutes (inside a 10 minute file) then the first event offset would be 10240 instead of 1024. Also, if the event file lists event onsets not possible within a CNT file (e.g. an onset of 10240 in a file only 5 seconds long), those extra events are simply ignored.
The linked file here took about 60 seconds to make in Excel, and will enable epoching of a file with 25% sweep overlap, up to 12 minutes. The file can be a lot longer, and the same EV2 file can be used to epoch multiple CNT files as long as the CNT files have the same sampling rate. Epoching using an EV2 file is detailed in the Edit manual, in case you have any questions about it. However if you have further questions, feel free to contact tech support.

Partway through a batch file execution it quit. How do I get Edit to work again? Close all Scan programs. Search for a file called Acquire43.ini (usually in a C:\Windows or C:\WinNT folder). Once found, delete the file. Restart Scan and Edit.

All about event triggers There are actually a couple of different options that will let you insert events so you can make epochs and average. I think it will help if I describe the difference between epoching using "Port/Internal" as the mode, and using "Event File" as the mode.
Basically the event markers you see at the bottom of a CNT file are actually stored in an "event table," and this table is located in a footer after the continuous data. If you were to open a .CNT file using custom code), you would notice that a CNT file essentially has three parts:
1.) Header, which contains basic file information such as number of channels, filter and A/D settings, subject information, etc.
2.) Data, which is the actually numbers/time series data;
3.) Footer, which contains the event table.

When Scan opens a CNT file, it reads all three; when it reads the event table, it knows where to place the event markers within the CNT file. Essentially, "Port/Internal" epoching tells Scan to use information in the event table for epoching the file.
The "Event File" transform essentially outputs the information in the event table into an ASCII file (the .EV2 file), and this ASCII file can be read in a multitude of programs, e.g. Excel, Notepad, etc. This EV2 file is then used (or any EV2 file, actually) when you select "Event File" as the epoching mode. What the latter tells Scan to do is to ignore any information in the event table, for that particular CNT file, and to use the event information listed in the EV2 file itself. This was highly useful in situations where there were NO event markers in the CNT file (i.e. the event table was empty).
The above was highly useful before Scan 4.2. Before 4.2, there was no way of modifying the event table directly within Scan; you needed to be a skilled programmer, and write a custom script that could read the CNT file and change this information. Thus, if you were a pre-4.2 user and wanted to modify the event information easily, for epoching purposes, the best way to do this was to create an event file, modify it, and epoch using that modified event file.
It is important to note that when this is performed, it does not actually change the event table itself. For example, let's say you open a CNT file that shows an event marker every exact 2000 milliseconds, and the code of that marker is always a 1. Let's say for whatever reason, you wanted to change every OTHER marker to a code of 2, and change the latency of that marker by 500 ms (thus if the 2nd marker had a latency of 4000 ms from the onset of the file, you would change it to 4500). Thus you would create an event file, change the information for those markers within Excel, save it as a new event file, and then epoch the CNT file using that modified event file. When you finish, you close Scan. The next time you open that same CNT file, it will still have a "1" every 2000 ms, because we did not change any information in the event table itself.
Note that the above is still possible for just about any version of Scan, up through 4.4.
In 4.4 however, we implemented a new transform called "Import Event File." What this enables you to do is to output an event file, modify it, and then import it back into a CNT file...thus, you actually overwrite that CNT file's original "event table" with the information in the new EV2 file. So the next time that CNT was opened, the event markers displayed would reflect the changes you made in the EV2 file. Thus you could then epoch the file, and use "Port/Internal" as the mode. Note that 4.4 can be downloaded from http://www.neuroscan.com , is free to all Scan 4.x users, and will install directly over top of Scan 4.3.1, 4.3.2 and 4.3.3 (the latter three versions are necessary in order to install and run 4.4).
Anyway, now that I have gone through all that, I DO realize I have yet to answer your question. In each row of the EV2 file, it lists information for a single stimulus/event marker. The first column is simply a line number (irrelevant in this case); the 2nd column shows the stimulus event marker code; the 3rd column is the code of any response associated with that stimulus; the 4th column is the accuracy of that response (1 is a correct response/rejection, 0 if an incorrect response, -1 if subject fails to respond when required to); the 5th columns is the latency of that response, relative to the stimulus; and the final column is the offset of that stimulus, either in data points or time, from the beginning of the data (in your example, it looks like sampled data points). If you are inserting rows into the EV2 file it would be something like this, let's say I want an event between your 1st and your 2nd event:
1 1 0 0 0 714
2 1 0 0 0 1675
So I would put in (in red):
1 1 0 0 0 714
1 2 0 0 0 1250
2 1 0 0 0 1675
I am also setting the stimulus code to a "2" so I can distinguish it as my inserted events (not necessary, just for purposes of this example). I would then re-save the EV2 file. Then, when you epoch (using the "Epoch File" transform), select "Event File" as the mode, and point to the modified EV2 file. Then, set "Type" as your Sort Criteria, and put in a 2. You would not need to select anything else. It should then go to the 1100th data point in your file, and create an epoch based on your parameters, and assign it a type code of 2.
Right now, my question would be if you are setting the offsets of the new events correctly. If you are outputting them as data points, you would need to divide that offset by your sampling rate in order to get the actual timestamp of that event. So, if you know you want to insert an event at 2500 milliseconds, and the sampling rate of the CNT file was 500 Hz, then you would need to divide the time by 1000 to convert into seconds, and then multiply that by the sampling rate: (2500/1000*500=1250).

You can also manually insert an individual stimulus event by typing CTRL-V , and giving it a stimulus code. Be careful not to overlap different kinds of events with the SAME stim code.

A third way to insert events is through Transforms-->Voltage Treshold-->Add Events.

Matlab and EEGlab can read *.cnt files but my home software can't. Can I "Save As" something else? The main issue with ASCII export is that it is HUGE. I think the size of an ASCII-exported file is about 2.5-3 times the size of the original CNT. The only thing that may limit whether Scan can then re-import this file is if the ASCII is too large (I have no empirical evidence to note that is valid, but if you are getting an "out of memory" message when trying to import, then that is likely the issue); or if some extra "marks" have been written into the file. An example of the latter is that some programs will write out positive values with a "+" sign in front, and this can cause Scan to not be able to read the export.
16 or 32-bit format? Scan 4.3.1 and up can read both. Anything earlier than 4.3.1 (this would include 4.3) cannot read 32-bit files. Also, anything Scan 4.3.1 reads in or produces is automatically converted into 32-bit. Any Synamp2 file is stored in 32-bit format. There is no flag in the output file that notates whether a file is 16 or 32-bit, either.

How can I display group averages? At the bottom of the Transforms list, go to Group/Individual Avg. You can then select the AVG files to include in a group average. For the "weighting" process, you can select either "Group" or "Individual." Group means that each AVG file is treated the same way in the averaging process, i.e. the waveforms (within each channel) are added together, and divided by the number of files. Thus, each file contributes equally to the resultant average. With Individual weighting, the number of accepted sweeps in each file are summed; then, a scaling factor is determined by dividing the number of sweeps in each file by the total sweeps. That scaling factor is then multipled by that AVG file, and the sum is calculated, giving the average. So let's say you have three files:
File1: 50 accepted sweeps
File2: 30 accepted sweeps
File3: 20 accepted sweeps
so that is 100 total sweeps. Individually weighted, File1 would have a scaling factor of 50/100 = 0.5; File2 = 30/100 = 0.3; and File3 = 20/100 = 0.2. Thus, your resultant average would equal:
group AVG = (File1*0.5)+(File2*0.3)+(File3*0.2)
and File1 would contribute more strongly than the others to the group's average. Which method to use? Depends on your rationale.

3DSpace

I can't Record, I think my digitizer isn't communicating. How do I check? There are three places to set the baud rate and communication of your digitizer. The (hardware) dipswitches on your Polhemus, the 3DSpace setup pane, and your PC's COM Port settings (control panel-->management-->Device manager -->com/lpt-->Properties-->Port settings).
There are four things to check;
Bits per second (57600) (or with the dipswitch settings below, 115200)
data bits (8)
Parity (none)
Stop bits (1)
flow control (Xon/Xoff)
The Polhemus manual page 4 (of rev.E) has the default I/O dipswitch settings; I recommend DOWN UP UP DOWN UP DOWN DOWN UP.

My 3DSpace reconstruction looks inaccurate. Why? Make sure the Polhemus Transmitter is within 50cm of your subject (and the three sensors).
The further you can get from ANY metal objects, the better the digitization will be.
It seems to help if the rubber cap is off the stylus to get positions for electrodes.
It seems to help if the stylus is PERPENDICULAR to each position you record.

Stim^2

This is how we import our TTL pulse from the MRI scanner.Pin 3 from Stim talks to pin 24 on SCAN parallel extender. In Acquire setup, Triggers --> External, Holdvalue = 0, Use inverted values.
When Stim2 systems are purchased from Neuroscan, there are two options:
1.) Stim2 Complete: includes a Stim2 software license, the audio box, PC, graphics card, P I/O card, sound card. This system's audio output is calibrated, and the trigger timing verified.
2.) Stim2 Software Only: includes only the Stim2 software license - no other hardware.
In the latter option, there is no way we can guarantee any kind of calibrated audio output, nor can we verify the trigger timing. Thus, we generally do not recommend any analogous hardware that may be compatible in other PCs.
Can I run Stim^2 from my laptop, like ePrime?
Basically, you would need a sound level meter to measure the output, if exact intensities are important to you. If these will be free-field playbacks, it may not be as necessary to know the exact intensity past "Can you hear it now? Good." It is also a little easier to verify trigger timing with sound files; you can attach the analog output of the WAV file playback into the BNC connector of the Synamp2 headbox's shorting plug, and actually record the sound waveform itself (as if it were an EEG channel). You would then also record the trigger, and thus could see if any jitter existed between the two.
Thus, you would need to perform some verification work, but it is definitely possible to run Stim2 on a laptop.
Responses are not reaching the Scan PC: After you've checked the cabling and made sure power is on where power shouldbe on, (1) Turn off and unplug the Stim PC. (2) Open the case, reseat the Stim card in the motherboard (take it out, push it back in GENTLY but solidly). (3) Close up and boot the machine. IF IT STILL DOESN'T WORK, contact tech support.
I hear loud white noise instead of my stimulus. What happened?
Any time you hear noise coupled into the auditory output, check the audio in and mic out cables (that connect the audio box to the PC's soundcard), and the digital control cable (that connects the audio box to the P I/O card installed in the PC). If any of these three cables are near or in contact with a source of noise, e.g. a power supply from another device or the isolation transformer, the noise from that device will couple into the auditory output. Thus, you want to isolate these three cables as much as possible from all other cables.
Really, the audio box should plug into the soundcard itself, not into the mixer panel on the front of the PC - although I am guessing that by plugging into the mixer, it basically isolated those cables from the noise generators. Optimally, you should plug into the soundcard, but if there is no way to achieve good signal quality by doing so, then the mixing panel sounds (unintentional pun) like the best bet. IF IT STILL DOESN'T WORK, contact tech support.
Basic ABR acquisition parameters here. Also, you will want to check the presentation parameters once in Sound. The information should be preserved from the files I sent you, but it may not preserve the set ITI for playing between sweeps. This should be 85 (85 + 4.9 quiet time in the file = 89.9, which gives an 11.1 Hz frequency) .
Also, when playing in Sound, you will need to use the slider scales at the bottom of the main Sound window, to set the dB intensity appropriately in each channel. Basically, each file is 5 ms in duration, with a click onset at 3 ms, and the click duration is 0.1 ms. The onset of the trigger pulse is at 4 ms; thus given the 0.9 ms delay, the click onset would be 3.9 ms, offset at 4. Thus, the trigger should present at the end of the click.
You will also want to use the bipolar channel to connect your electrodes; and you will need to jump from the negative pole of the bipolar into the REF input on the headbox (when using Synamp2).
Can I be certain "rare" stimuli are not presented back-to-back? Yes, with this format of genCPT file, that ran in the DOS version of Stim. This legacy snippet should import correctly to a newer version. Contained in the zipfile is a textfile explaining the format and the genCPT.cpt example. What you want to do is edit the genCPT.cpt to what you'd like (with timing in seconds, not msec) according to the parameter restriction in the genCPT.txt file, then run the executable. It will prompt you for a CPT (don't include path or file extension) and the name you'd like to call your new *.SEQ in Stim-for-Windows. Review your newly converted file in the Gentask Editor.

Quikcaps

Maglink cap electrodes and wires that need repair should be returned to Neuroscan.

Non-MR quikcap electrodes can be repaired by the user with the repair kit available from Neuroscan. The electrode repair kit is comprised of 4 pieces:

1.) solder tube
2.) heat shrink
3.) o-ring
4.) replacement electrode with short length of lead.

Trim the broken wire on the cap, and slide the heat shrink down the wire. Then, insert the end of the wire into one end of the solder tube. The wire coming off the replacement electrode should be inserted into the other end of the solder tube. The two wires should "meet in the middle" where the solder ring is located. Then, using a heat gun (optimally) or a soldering iron (a little trickier but should work) melt the tube. This will join the two wires together. Then you place the electrode in the hole of the cap, and secure around the rim of the electrode with the o-ring.

Electrode impedances, over time, degrade. We found that limiting the soak time in cleaning solution after each session to 30 minutes staves off repair by a few months.

Is my electrode broken? If the trace looks "off," check that the impedance is acceptable. If early in your recording session, sometimes the electrode gel/paste has dried or moved far enough that connectivity to the scalp is lost. This might be fixed by inserting more gel/paste, but beware that you don't bridge your "trouble" channel to a known good channel in the process. Alternately, you can (offline) touch the metal of the electrode you are checking with one probe of a multimeter on Resistance mode; the other probe would go into either (a) the pin of the 80-pin connector that is wired to that channel (this would be in the documentation that comes with the Quik-cap); or (b) if the cap is plugged into the headbox (system should be off), then you would contact the pin in the physical channel on the top of that headbox which is assigned to that electrode (as shown in the Channel Assignment Table).
If you discover that your "trouble" channel is rather different than known good channels, you can either send your quikcap (MagLink caps must be sent to Neuroscan for repair) to Neuroscan, or you can order a repair kit to do it yourself with a little additional hand/eye coordination and a source of hot air.

I got a "data read error" soon after I started recording, but nothing noticable had changed. What happened? Acquire records into a buffer for a certain amount of time, then pushes that data in packets to your PC. In the event any of those packets became corrupt, Acquire will notice the error, and stop the recording, rather than continue to log gibberish. One possible fix might be to swap the 20-foot long USB cable from the amplifier to the PC with a shorter one.