Head movement correction as a pre-processing step

✉️ Mailing List Archive (read-only)
1

Dear MNE users,

I am trying to understand the difference between using MaxFilter (SSS
based) and using 'mne_map_data' to correct for head movement. Does anyone
know what algorithm is implemented for 'mne_map_data'?

Thanks in advance for any help!

Best,

Qiong

2

hi,

mne_map_data is based on Matti's paper from the early 90's AFAIK.

for some reason mne_map_data seems to be absent from the MNE website
and even from the original MNE manual...

Alex

3

Hi Alex,

Thanks so much for the information. I really appreciate it. I have looked
into papers during that period of time by matti H?m?l?inen, with the most
relevant material in this chapter
<https://mycourses.aalto.fi/pluginfile.php/123749/mod_resource/content/1/2002-H?m?l?inen-BrainMappingMethods.pdf>
p237
last paragraph in section 3 providing some understanding of the problem:

"The continuous records of the head movements can then be utilized in
different correction methods. If the signal-to-noise ratio is so good that
unaveraged signals can be modeled, one can simply use a different
coordinate transformation between the anatomical and device coordinate
frames at each time instant. The situation is more complicated if signals
are averaged or if continuous data segments containing signals from
supposedly identical source distributions are compared. Uutela et al.
(2001) have recently explored several alternative computational approaches
applicable to signal averaging and source modelling in situations with
significant head movements during the data acquisition.
"

As the name 'mne_map_data' suggests the use of simple transformation and is
often applied by sessions/blocks, correct me if I'm wrong, it seems to
implement different transformations between the anatomical and device
coordinate frames using data averaged within those sessions/blocks. As a
result, the sensor data is transformed as if they are recorded from the
same anatomical positions across these sessions/blocks.

Best,
Qiong

4

Hi Qiong,

`mne_map_data` uses Minimum Norm Estimate. You project your data onto a
sphere and project it back to the new sensor locations. Actually, you
should be able to find the Mathematical details in Matti's 1994 paper:
http://link.springer.com/article/10.1007%2FBF02512476?LI=true.

Mainak

5

Hi Mainak,

You're right. Section 2.4 has the details. Thanks so much for the
information. It is really helpful.

Best,
Qiong