PS I forgot to say that we realise that the first example that bails out
below may be because the radius is too large (extending outside the
sensors), but were wondering whether this is actually a problem if it is
only the centre that really matters, and even if it is a problem, our
second question still stands!
Rik Henson wrote:
> Dear Neuromeg list.
>
> We have two questions about the centre of the spherical expansions in
> Maxfilter:
>
>
> 1. Occasionally MaxFilter bails out with the error:
>
> "ERROR: sphere fitted to isotrak extends outside of the sensors!
> (r0 = (14 3 36) mm, rad = 11 cm)"
>
> However, when we run the Neuromag utiltity:
>
> /neuro/bin/util/fit_sphere_to_points
>
> (or our own sphere-fitting code), we seem to get sensible fits, eg
> Centre = [1.347 7.374 42.871] (Radius=98.608).
>
> The MaxFilter manual implies that the MaxFilter sphere-fitting is more
> sophisticated/robust than simple methods, so should we not trust the
> "fit_sphere_to_points" results above? Or is the error actually with
> MaxFilter? If the latter, should we just "bypass" this MaxFilter bug
> by passing "-frame head -origin 1.347 7.374 42.871" (in this example)?
>
>
> 2. Related to question 1 above, we have noticed that the sphere fit by
> MaxFilter and the sphere fit by fit_sphere_to_points tend to differ,
> sometimes vastly so, with the origins differing by 10mm in one or
> another dimension. (When we independently fit a sphere using SPM
> functions in matlab, the origin is consistently very close (within 1
> mm) to the fit_sphere_to_points result). Here are some examples:
>
> Subject 1:
> -0.7770 5.3440 35.9980 fit_sphere_to_points origin
> -0.9496 5.1591 36.2188 SPM origin
> 2.2 6.2 38.6 MaxFilter Inside Origin
>
> Subject 2:
> -9.5500 0.8510 39.8750
> -9.5891 1.4581 39.3087
> -10.1 5.5 36.9
>
> Subject 3:
> -0.6850 11.9050 51.6610
> -0.5834 11.6485 51.4186
> -10.4 12.5 46.6
>
> So should we ALWAYS use independent sphere fitting, rather than
> relying on MaxFilter? (we expect not, but just wanted to get your
> attention! ;-) Or are we misunderstanding what MaxFilter actually does?
>
> Many thanks
> Rik Henson, Jason Taylor, Danny Mitchell
>
--
----------------------------------------
Dr Richard Henson
MRC Cognition & Brain Sciences Unit
15 Chaucer Road
Cambridge
CB2 7EF, UK
Tel: +44 (0)1223 355 294 x522
Fax: +44 (0)1223 359 062
http://www.mrc-cbu.cam.ac.uk/~rik.henson
----------------------------------------
|