Dear Bill,

On Wed, Jun 24, 2009 at 4:32 AM, Bill Budd<[log in to unmask]> wrote:

> Have been following your posts with interest as we also have a 64 channel
> Biosemi (Active 2) system and would also like to start using SPM8 to do
> analyses and EEG pre-processing with Biosemi EEG data. I think Diane's
> system is probably a standard Biosemi 64channel system (like ours) so would
> be great to have the electrode configuration recognised by SPM8 as a
> default. I have attached the Biosemi electrode 10-20 type naming conventions
> and standard spherical coordinates for the standard Biosemi systems (32-256
> channels) which I downloaded from the CorTech website (see
> http://www.cortechsolutions.com/FAQ_ActiveTwo.htm for other Biosemi related
> info).

This is the same file that Diane sent me. You can download the updated
SPM from  https://download.yousendit.com/UmNMV0p3T01reENGa1E9PQ (just
do it soon before the link expires). All these changes will be in the
next SPM update.

>
> One issue I would like to get some advice regarding SPM8/Biosemi regards the
> active reference format (i.e reference free) used by Biosemi. This differs
> from other EEG systems and I wondered how this is dealt with in SPM8
> generally, and in particular for source analyses. Specifically, does this
> impact on the preprocessing if no reference is specified or should I
> re-reference to an explicit reference in SPM8 before doing any processing?


I'm not sure what this 'reference free' really means. Voltage is
always defined in reference to some point and your other remark about
huge values probably implies that this point might be the amplifier
ground. At this stage for 3D source reconstruction and DCM I'd
recommend to convert the data to average reference. For sensor-level
analysis or time-frequency analysis you can use any reference that you
think will better amplify the effects of interest as long as you can
justify it in your paper.


> A related issue is that often when using the 64-bit Biosemi data in some EEG
> analysis packages (i.e. BESA) the actual amplitude values when imported into
> BESA can be HUGE as a result of large DC offsets and dynamic range of the
> Biosemi data. This can be readily resolved by application of a high-pass
> filter, however I did wonder whether this would cause any problems using
> this data in SPM8 (not sure if this is relevant to Diane's original issue
> below)?
>


By default SPM uses float32 (single in Matlab) to store the data. This
might lead to quantization errors in your case. It is possible to
change the data type to float64 (double)  in batch or script. Perhaps
you should try both ways and see if there is any noticeable
difference. It is also possible to do the initial conversion in
float64, filter and then convert the filtered dataset (just as any
other raw file) this time choosing float32.

Best,

Vladimir


> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
> Behalf Of Vladimir Litvak
> Sent: Monday, 22 June 2009 8:55 PM
> To: [log in to unmask]
> Subject: Re: [SPM] SPM8 EEG inverse: Matrix is singular, close to singular
> or badly
>
> Dear Diane,
>
> Things are becoming clearer and I think we are quite close to resolving your
> problems.
>
>> We have a 64-channel standard Biosemi system. From what I understand,
>> you can order Biosemi with 32, 64, 128, or 256 channels. Can SPM work
>> with 64 channels if they have standard 10-20 names?
>
> We haven't seen a Biosemi system with less than 128 channels but since you
> have one, it should be supported. The question is, is the EEG cap you are
> using a standard one for 64-channel Biosemi or is it something you ordered
> separately and other labs with 64-channel Biosemi systems are likely to have
> other caps? Also is the mapping between Biosemi and 10-20 labels that you
> are using just something approximate or do you have some documentation from
> Biosemi saying that this is the right correspondence?
>
> The reason I'm asking you all this is as follows. If your system can be
> considered a standard Biosemi 64 system, then I can add default support for
> it in SPM. I have already added automatic recognition of EEG channels and to
> make it complete I need a file with 3D electrode positions. If you are sure
> that the channels correspond exactly to 10-20 locations, then just send me
> the full label mapping you are using and I'll make this file myself. If not,
> then perhaps you can find somewhere a locations file with fiducials for your
> cap or you can make a good quality measurement in your lab with a 3D
> digitizer.
>
> If your cap is something that is not likely to come with other
> Biosemi-64 systems then we still can make a locations file for you, just it
> won't be included in SPM distribution. Then you won't have to rename your
> channels with the GUI every time.
>
>> Ok, this sounds good. It's strange to me that the filtering function
>> requires channel names to work.
>
> The problem here is that not all channels should be filtered. For instance
> trigger channels should not. The filtering function doesn't use channel
> names but it does use channel types so the types should be set correctly
> before filtering.
>
>> Is it necessary to include a negative baseline? We want to exclude the
>> visually evoked potentials, which is why we are interested in 0.5
>> seconds to 4.0 seconds.
>>
>
> In principle it's not necessary but most epoched files have negative
> baseline and SPM5 required it. There might still be some unfixed bugs
> related to that, like the one I found in spm_eeg_filter. I think you should
> continue the way you are working now and report any problems to me.
>
>
>> There is a strong VEP at 0-0.5 seconds, but we are not interested in that.
>> Actually, we'd like to perform time-frequency analysis on a
>> single-trial basis rather than ERP analysis. Can SPM calculate an
>> inverse solution for non-averaged data?
>
> SPM can look at induced activity based on single trials (see the
> manual) but this is something that we don't yet have much experience with.
> There are now several people testing this functionality and I can put you in
> touch with them if you want. What is not possible is to perform
> time-frequency analysis at the channel level and then source-reconstruct the
> power data, because once you discard phase information source reconstruction
> is not possible any more.
>
>> Shouldn't the data be filtered prior to down-sampled? Otherwise, the
>> effects of aliasing will distort the data?
>
> No, it's OK. The downsampling function in Matlab is aware of aliasing so it
> performs filtering prior to downsampling.
>
>> One thing I'm not clear about, however, is whether the data should be
>> downsampled and filtered before or after epoching.
>
> It depends, but I'd say if your continuous data is not much bigger than your
> epoched data (i.e. if you don't have too much dead time in your recordings)
> perhaps you should downsample and filter before epoching. That's always a
> safer option.
>
>
> So I'd suggest the following sequence:
>
> Convert -> Prepare (get the labels, types, channel locations etc.
> right) -> Downsample -> Filter ->Epoch ->Artefact rejection->3D source
> reconstruction
>
> You don't need averaging because to look at induced activity you need a file
> with single trials. You can try switching filtering and downsampling to see
> if it makes any difference but I don't think it should.
>
> Regarding the 'Prepare' stage I need some more information about the cap and
> then I'll tell you exactly what to do.
>
> Best,
>
> Vladimir
>
>
>
>
>>>
>>> >>
>>> >> On Fri, Jun 19, 2009 at 11:01 AM, Diane
>>> >> Whitmer<[log in to unmask]>
>>> >> wrote:
>>> >> > Vladimir,
>>> >> > The problem of losing my data is arising at the time of filtering.
>>> >> > Are the filter functions supposed to work on epoched data? If
>>> >> > so, is there anything special I should do to get the filtering
>>> >> > to work? e.g. I am trying to high pass above 2 Hz using the gui.
>>> >> > My sampling rate is 2048 Hz and my epochs are 3.5 seconds in
>>> >> > length.
>>> >> > Thank you,
>>> >> > Diane
>>> >> >
>>> >> >
>>> >> > On Thu, Jun 18, 2009 at 1:47 PM, Vladimir Litvak
>>> >> > <[log in to unmask]>
>>> >> > wrote:
>>> >> >>
>>> >> >> Display->M/EEG
>>> >> >>
>>> >> >> On Thu, Jun 18, 2009 at 12:37 PM, Diane
>>> >> >> Whitmer<[log in to unmask]>
>>> >> >> wrote:
>>> >> >> > Hi Vladimir,
>>> >> >> > How do I access the Reviewing tool?
>>> >> >> > When I look under Prepare > Review, I cannot find the EEG traces.
>>> >> >> > Thanks,
>>> >> >> > Diane
>>> >> >> >
>>> >> >> >
>>> >> >> >
>>> >> >> > On Thu, Jun 18, 2009 at 12:57 PM, Vladimir Litvak
>>> >> >> > <[log in to unmask]> wrote:
>>> >> >> >>
>>> >> >> >> Dear Diane,
>>> >> >> >>
>>> >> >> >> The file you sent me contains just zeros in the data so it's
>>> >> >> >> not surprising that SPM can't do much with it. Also the
>>> >> >> >> montage you applied is just an identity matrix so it doesn't
>>> >> >> >> do anything, but that's not the reason why the data is
>>> >> >> >> invalid. You should use the reviewing tool to see at which
>>> >> >> >> stage in the processing your data disappears.
>>> >> >> >>
>>> >> >> >> Best,
>>> >> >> >>
>>> >> >> >> Vladimir
>>> >> >> >>
>>> >> >> >> On Thu, Jun 18, 2009 at 11:42 AM, Diane
>>> >> >> >> Whitmer<[log in to unmask]>
>>> >> >> >> wrote:
>>> >> >> >> > Hi Vladimir,
>>> >> >> >> > I changed the montage to average reference, and then
>>> >> >> >> > averaged my trials.
>>> >> >> >> > The same error message appears when I try to run the
>>> >> >> >> > Inverse Solution.
>>> >> >> >> > I just sent both the .dat and the .mat via yousendit.
>>> >> >> >> > Please let me know if you are able to figure out the
>>> >> >> >> > source of the errors. I *very much* appreciate your help.
>>> >> >> >> > Sincerely,
>>> >> >> >> > Diane
>>> >> >> >> >
>>> >> >> >> >
>>> >> >> >> >
>>> >> >> >> > On Wed, Jun 17, 2009 at 12:29 PM, Vladimir Litvak
>>> >> >> >> > <[log in to unmask]> wrote:
>>> >> >> >> >>
>>> >> >> >> >> Dear Diane,
>>> >> >> >> >>
>>> >> >> >> >> Your new montage is quite good, but you should decide
>>> >> >> >> >> what you want to do about the 'EXG1' channel. Presently
>>> >> >> >> >> it appears in the montage, but is actually set to zero. I
>>> >> >> >> >> understand this is your reference.
>>> >> >> >> >> I
>>> >> >> >> >> saw
>>> >> >> >> >> in
>>> >> >> >> >> the history that you are using SPM's default channel
>>> >> >> >> >> locations.
>>> >> >> >> >> SPM
>>> >> >> >> >> wouldn't have a location for EXG1 and wouldn't mark it as
>>> >> >> >> >> 'EEG'
>>> >> >> >> >> by
>>> >> >> >> >> default. So you have two options:
>>> >> >> >> >>
>>> >> >> >> >> 1) If you want to use it as an extra EEG channel you need
>>> >> >> >> >> to either change its name to some 10-20 name and assign
>>> >> >> >> >> default channel types and default locations again or at
>>> >> >> >> >> least change its type to 'EEG'
>>> >> >> >> >> in
>>> >> >> >> >> the 'prepare' interface and load a locations file where
>>> >> >> >> >> this channel appears. Then you should make a montage like
>>> >> >> >> >> what you did just with
>>> >> >> >> >> 65
>>> >> >> >> >> channels including EXG1 so tra = eye(65) - ones(65)/65;
>>> >> >> >> >>
>>> >> >> >> >> 2) If you can live without it, you can use the montage
>>> >> >> >> >> you sent me, or even simpler just don't include EXG1 in
>>> >> >> >> >> the montage at all.
>>> >> >> >> >> You
>>> >> >> >> >> can
>>> >> >> >> >> either keep it in the file as a non-EEG channel or not
>>> >> >> >> >> keep it, for source reconstruction it doesn't matter.
>>> >> >> >> >>
>>> >> >> >> >> Once you convert the data to average reference, try the
>>> >> >> >> >> source reconstruction again. If there is still a problem,
>>> >> >> >> >> average your data and see if for averaged data you get
>>> >> >> >> >> the same error message.
>>> >> >> >> >> If
>>> >> >> >> >> so
>>> >> >> >> >> send it to me, the file will be much smaller. If you only
>>> >> >> >> >> get the error for epoched data, but not for averaged let
>>> >> >> >> >> me know and we'll think of a way for you to send me the
>>> >> >> >> >> epoched file.
>>> >> >> >> >>
>>> >> >> >> >> Best,
>>> >> >> >> >>
>>> >> >> >> >> Vladimir
>>> >> >> >> >>
>>> >> >> >> >>
>>> >> >> >> >>
>>> >> >> >> >> On Wed, Jun 17, 2009 at 11:12 AM, Diane
>>> >> >> >> >> Whitmer<[log in to unmask]>
>>> >> >> >> >> wrote:
>>> >> >> >> >> > Vladimir,
>>> >> >> >> >> > I tried to send the .dat file, but it was over 100MB
>>> >> >> >> >> > and therefore cannot be sent via yousendit without a
>>> >> >> >> >> > subscription. Do you have another suggestion?
>>> >> >> >> >> > Thanks,
>>> >> >> >> >> > Diane
>>> >> >> >> >> >
>>> >> >> >> >> >
>>> >> >> >> >> > On Tue, Jun 16, 2009 at 9:38 PM, Vladimir Litvak
>>> >> >> >> >> > <[log in to unmask]>
>>> >> >> >> >> > wrote:
>>> >> >> >> >> >>
>>> >> >> >> >> >> Dear Diane,
>>> >> >> >> >> >>
>>> >> >> >> >> >> The error might have something to do with the fact
>>> >> >> >> >> >> that the montage matrix you used is wrong. To do what
>>> >> >> >> >> >> you intended you'd need the transpose of what you
>>> >> >> >> >> >> specified, but as I mentioned to be on the safe side
>>> >> >> >> >> >> it'd be better to convert the data to average
>>> >> >> >> >> >> reference.
>>> >> >> >> >> >> In
>>> >> >> >> >> >> your
>>> >> >> >> >> >> case of a 64 channel system you'd need something like:
>>> >> >> >> >> >>
>>> >> >> >> >> >> tra = eye(64) - ones(64)/64;
>>> >> >> >> >> >>
>>> >> >> >> >> >> for the EEG channels (or if you have 65, the same with
> 65).
>>> >> >> >> >> >>
>>> >> >> >> >> >> Best,
>>> >> >> >> >> >>
>>> >> >> >> >> >> Vladimir
>>> >> >> >> >> >>
>>> >> >> >> >> >> On Tue, Jun 16, 2009 at 12:37 PM, Diane
>>> >> >> >> >> >> Whitmer<[log in to unmask]>
>>> >> >> >> >> >> wrote:
>>> >> >> >> >> >> > Dear SPMers,
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > I am using SPM8 with EEG data. When I try to compute
>>> >> >> >> >> >> > the Inverse solution, I get the following error
>>> >> >> >> >> >> > message:
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > Warning: Matrix is singular, close to singular or
>>> >> >> >> >> >> > badly scaled.
>>> >> >> >> >> >> > Results may be inaccurate. RCOND = NaN.
>>> >> >> >> >> >> >
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > Does anyone have suggestions for how to debug this?
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > A subset of these data previously worked for
>>> >> >> >> >> >> > computing an Inverse Solution.
>>> >> >> >> >> >> > One difference with my current dataset is that I
>>> >> >> >> >> >> > have re-referenced to one of the EEG channels
>>> >> >> >> >> >> > (because it's not clear that SPM selects a reference
>>> >> >> >> >> >> > channel when importing from Biosemi). This may or
>>> >> >> >> >> >> > may not be related to the problem I'm having.
>>> >> >> >> >> >> >
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > Thank you,
>>> >> >> >> >> >> > --
>>> >> >> >> >> >> > Diane Whitmer, Ph.D.
>>> >> >> >> >> >> > Researcher in Applied Neuroscience
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > C. de l' Observatori Fabra s/n
>>> >> >> >> >> >> > 08035 - Barcelona - Spain tel. +34 93 254 03 68 fax.
>>> >> >> >> >> >> > +34 93 212 64 45 skype. dwhitmer email.
>>> >> >> >> >> >> > [log in to unmask]
>>> >> >> >> >> >> > ______________________________________
>>> >> >> >> >> >> > STARLAB, Living Science http://www.starlab.es
>>> >> >> >> >> >> >
>>> >> >> >> >> >> > This e-mail and the attached files contain
>>> >> >> >> >> >> > confidential information and are of exclusive use
>>> >> >> >> >> >> > for the addressee. The disclosure, copy and
>>> >> >> >> >> >> > distribution of the present message and its
>>> >> >> >> >> >> > attachments are prohibited.
>>> >> >> >> >> >> > If
>>> >> >> >> >> >> > you
>>> >> >> >> >> >> > are
>>> >> >> >> >> >> > not
>>> >> >> >> >> >> > the
>>> >> >> >> >> >> > intended recipient and you receive this
>>> >> >> >> >> >> > communication, we inform you that it is forbidden by
>>> >> >> >> >> >> > law to read, copy or use it. Please notify us at
>>> >> >> >> >> >> > [log in to unmask] and proceed immediately to delete it.
>>> >> >> >> >> >> >
>>> >> >> >> >> >
>>> >> >> >> >> >
>>> >> >> >> >> >
>>> >> >> >> >> > --
>>> >> >> >> >> > Diane Whitmer, Ph.D.
>>> >> >> >> >> > Researcher in Applied Neuroscience
>>> >> >> >> >> >
>>> >> >> >> >> > C. de l' Observatori Fabra s/n
>>> >> >> >> >> > 08035 - Barcelona - Spain tel. +34 93 254 03 68 fax.
>>> >> >> >> >> > +34 93 212 64 45 skype. dwhitmer email.
>>> >> >> >> >> > [log in to unmask]
>>> >> >> >> >> > ______________________________________
>>> >> >> >> >> > STARLAB, Living Science http://www.starlab.es
>>> >> >> >> >> >
>>> >> >> >> >> > This e-mail and the attached files contain confidential
>>> >> >> >> >> > information and are of exclusive use for the addressee.
>>> >> >> >> >> > The disclosure, copy and distribution of the present
>>> >> >> >> >> > message and its attachments are prohibited. If you are
>>> >> >> >> >> > not the intended recipient and you receive this
>>> >> >> >> >> > communication, we inform you that it is forbidden by
>>> >> >> >> >> > law to read, copy or use it. Please notify us at
>>> >> >> >> >> > [log in to unmask] and proceed immediately to delete it.
>>> >> >> >> >> >
>>> >> >> >> >
>>> >> >> >> >
>>> >> >> >> >
>>> >> >> >> > --
>>> >> >> >> > Diane Whitmer, Ph.D.
>>> >> >> >> > Researcher in Applied Neuroscience
>>> >> >> >> >
>>> >> >> >> > C. de l' Observatori Fabra s/n
>>> >> >> >> > 08035 - Barcelona - Spain
>>> >> >> >> > tel. +34 93 254 03 68
>>> >> >> >> > fax. +34 93 212 64 45
>>> >> >> >> > skype. dwhitmer
>>> >> >> >> > email. [log in to unmask]
>>> >> >> >> > ______________________________________
>>> >> >> >> > STARLAB, Living Science http://www.starlab.es
>>> >> >> >> >
>>> >> >> >> > This e-mail and the attached files contain confidential
>>> >> >> >> > information and are of exclusive use for the addressee.
>>> >> >> >> > The disclosure, copy and distribution of the present
>>> >> >> >> > message and its attachments are prohibited. If you are not
>>> >> >> >> > the intended recipient and you receive this communication,
>>> >> >> >> > we inform you that it is forbidden by law to read, copy or
>>> >> >> >> > use it. Please notify us at [log in to unmask] and proceed
>>> >> >> >> > immediately to delete it.
>>> >> >> >> >
>>> >> >> >
>>> >> >> >
>>> >> >> >
>>> >> >> > --
>>> >> >> > Diane Whitmer, Ph.D.
>>> >> >> > Researcher in Applied Neuroscience
>>> >> >> >
>>> >> >> > C. de l' Observatori Fabra s/n
>>> >> >> > 08035 - Barcelona - Spain
>>> >> >> > tel. +34 93 254 03 68
>>> >> >> > fax. +34 93 212 64 45
>>> >> >> > skype. dwhitmer
>>> >> >> > email. [log in to unmask]
>>> >> >> > ______________________________________
>>> >> >> > STARLAB, Living Science http://www.starlab.es
>>> >> >> >
>>> >> >> > This e-mail and the attached files contain confidential
>>> >> >> > information and are of exclusive use for the addressee. The
>>> >> >> > disclosure, copy and distribution of the present message and
>>> >> >> > its attachments are prohibited. If you are not the intended
>>> >> >> > recipient and you receive this communication, we inform you
>>> >> >> > that it is forbidden by law to read, copy or use it. Please
>>> >> >> > notify us at [log in to unmask] and proceed immediately to
>>> >> >> > delete it.
>>> >> >> >
>>> >> >
>>> >> >
>>> >> >
>>> >> > --
>>> >> > Diane Whitmer, Ph.D.
>>> >> > Researcher in Applied Neuroscience
>>> >> >
>>> >> > C. de l' Observatori Fabra s/n
>>> >> > 08035 - Barcelona - Spain
>>> >> > tel. +34 93 254 03 68
>>> >> > fax. +34 93 212 64 45
>>> >> > skype. dwhitmer
>>> >> > email. [log in to unmask]
>>> >> > ______________________________________
>>> >> > STARLAB, Living Science http://www.starlab.es
>>> >> >
>>> >> > This e-mail and the attached files contain confidential
>>> >> > information and are of exclusive use for the addressee. The
>>> >> > disclosure, copy and distribution of the present message and its
>>> >> > attachments are prohibited. If you are not the intended
>>> >> > recipient and you receive this communication, we inform you that
>>> >> > it is forbidden by law to read, copy or use it. Please notify us
>>> >> > at [log in to unmask] and proceed immediately to delete it.
>>> >> >
>>> >
>>> >
>>> >
>>> > --
>>> > Diane Whitmer, Ph.D.
>>> > Researcher in Applied Neuroscience
>>> >
>>> > C. de l' Observatori Fabra s/n
>>> > 08035 - Barcelona - Spain
>>> > tel. +34 93 254 03 68
>>> > fax. +34 93 212 64 45
>>> > skype. dwhitmer
>>> > email. [log in to unmask]
>>> > ______________________________________
>>> > STARLAB, Living Science http://www.starlab.es
>>> >
>>> > This e-mail and the attached files contain confidential information
>>> > and are of exclusive use for the addressee. The disclosure, copy
>>> > and distribution of the present message and its attachments are
>>> > prohibited. If you are not the intended recipient and you receive
>>> > this communication, we inform you that it is forbidden by law to
>>> > read, copy or use it. Please notify us at [log in to unmask] and
>>> > proceed immediately to delete it.
>>> >
>>
>>
>>
>> --
>> Diane Whitmer, Ph.D.
>> Researcher in Applied Neuroscience
>>
>> C. de l' Observatori Fabra s/n
>> 08035 - Barcelona - Spain
>> tel. +34 93 254 03 68
>> fax. +34 93 212 64 45
>> skype. dwhitmer
>> email. [log in to unmask]
>> ______________________________________
>> STARLAB, Living Science http://www.starlab.es
>>
>> This e-mail and the attached files contain confidential information
>> and are of exclusive use for the addressee. The disclosure, copy and
>> distribution of the present message and its attachments are
>> prohibited. If you are not the intended recipient and you receive this
>> communication, we inform you that it is forbidden by law to read, copy
>> or use it. Please notify us at [log in to unmask] and proceed immediately to
> delete it.
>>
>