Dear Gustavo,
Samu is currently having problems with posting further messages on the
list but wished the below be sent soonest possible:
> Dear Gustavo,
>
> I would still like to add a couple of comments to Jukka's reply.
> First of
> all, I think that the default correlation value (0.98) of MaxFilter is
> somewhat strict and could be lowered a bit. A single-channel
> artifact is
> recognizable by tSSS but it may have such a low signal energy that
> correlation
> value 0.98 might not be enough for it to be detected. There is now
> strong
> evidence that lower correlation values can be safely used. You could
> try
> correlation value of, e.g. 0.9 or even 0.8 in tSSSS and see if it
> helps.
>
> Second, regarding item 2 in Jukka's reply, I would like to mention
> that
> strictly speaking, we do lose the information of the excluded
> channel in
> the SSS transformation of the sensor-level signals into multipole
> moments.
> However, this loss is insignificant as we have 306 channels and thus
> the
> signal of the bad channel can be reliably reconstructed from the
> multipoles by SSS.
>
> Best regards,
> Samu
With regards,
Miikka Putaala
Product Manager
Functional Mapping
Elekta Neuromag Oy
+358-9-7562400 (voice)
+358-50-5213788 (cellular)
+358-9-75624011 (fax)
Visiting address:
Siltasaarenkatu 18–20
FIN-00530 Helsinki
FINLAND
Mail address:
P.O. Box 68
FIN-00511 Helsinki
FINLAND
The contents of this e-mail message (including any attachments hereto)
are confidential to and are intended to be conveyed for the use of the
recipient to which it is addressed only. If you receive this
transmission in error, please notify the sender of this immediately
and delete the message from your system. Any distribution,
reproduction or use of this message by someone other than recipient is
not authorized and may be unlawful.
On 2009-03-02, at 18:28 , Gustavo Sudre wrote:
> Jukka,
>
> Thanks for your answer. The behavior I'm observing makes sense now.
> I'll mark the channels as bad and count on SSS's ability to
> reconstruct them. I'll also play with MaxST and see if I get better
> results.
>
> Best,
>
> Gus
>
> On Mar 2, 2009, at 2:36 AM, Jukka Nenonen wrote:
>
>> Dear Gustavo,
>>
>> 1) The discontinuity arises due to a saturating channel. At the
>> jump the signal exceeds the
>> dynamic range, and then slowly recovers back.
>> Often the saturation arises due to bad tuning of a channel. After
>> heating you should try
>> improve the tuning. If it still persists there can be a hardware
>> problem with that channel.
>>
>> 2) During the jump and recovery, the signal at that channel does
>> not obey Maxwell equations
>> and thus the spatial SSS spreads the artifact over several
>> channels. The easiest solution
>> is to mark the channel as a static bad one. SSS can reconstruct the
>> bad channel from
>> other channels without loosing any information.
>>
>> 3) Splitting the file or using the '-skip' option is more clumsy
>> because you first need to find
>> the times of jumps and then discard these periods.
>>
>> 4) The best way is to avoid discontinuities, either by setting
>> static bad channels manually
>> or by using the temporal extension (MaxST in the user guide).
>>
>> 5) MaxST may sometimes produce discontinuities at buffer boundaries
>> (def buffer is 4 secs).
>> If the boundary is at the period when a staurated channel is
>> recovering, MaxST does not
>> guarantee continuity and in fact sees differing interference
>> contributions in the buffers on
>> the 'left' and 'right'.
>>
>> Thus, I recommend to browse the data to detect if there are jumping
>> channels, and mark
>> them as static bad ones (remembering the limitiations on maxfilter
>> 2.0 autobad function).
>> You should use your own judgement to decide if MaxST is needed or
>> not. Usually,
>> MaxST improves the result and does no harm even if there are no
>> remaining
>> sensor-space components.
>>
>> Best regards, Jukka Nenonen and Samu Taulu
>>
>>
>>
>>
>> Gustavo Sudre wrote:
>>> Hello,
>>> For many repetitions in my experiments I have observed a
>>> discontinuity in the raw MEG data. This happens in some of my
>>> individual trials, and it is not always across several channels
>>> simultaneously. The attached picture shows an example of this type
>>> of noise, for a particular trial and channel. When I notice such
>>> pattern during recordings, reheating the sensor usually works.
>>> However, I don't catch them all the time, so it's often the case
>>> that I see this pattern in recorded data. These are my questions:
>>> 1) What causes this noise? Can the channel be "trusted" even if
>>> it shows the pattern a few times?
>>> 2) If I run my data through SSS, this pattern seems to appear in
>>> many more channels. That means that I need to discard many more
>>> repetitions after running SSS. Is this something intrinsic to the
>>> SSS algorithm (i.e. to multiply this noise)? What is the reason
>>> for it to appear across more channels after SSS?
>>> 3) If I clean up my raw data prior to SSS (e.g. discard the
>>> repetitions with such pattern), and create a new FIF raw file from
>>> this new data, my data won't be continuous in time anymore. I
>>> don't think SSS will have a problem with it (unless I use the
>>> temporal extension). Is that correct?
>>> 4) Would you suggest a more accurate way to deal with these
>>> discontinuities (eg. wavelets?), instead of discarding (sometimes
>>> precious) trials?
>>> 5) Assuming SSS has no problems with this new "clean" raw file,
>>> can the SSS algorithm create such discontinuities by itself? If
>>> so, why?
>>> Thanks,
>>> Gus
>>
>>
>> --
>>
>> ================================
>> Dr. Jukka Nenonen
>> Manager, Method development
>> Elekta Neuromag Oy
>> Street address: Siltasaarenkatu 18-20A, Helsinki, Finland
>> Mailing address: P.O. Box 68, FIN-00511 HELSINKI, Finland
>> Tel: +358 9 756 240 85 (office), +358 400 249 557 (mobile),
>> +358 9 756 240 11 (fax), +358 9 756 2400 (operator)
>> E-mail: [log in to unmask]
>> http://www.elekta.com/healthcare_international_elekta_neuromag.php
|