Re: [Marsbar-users] Identification of ROI

[Matthew B. <mat...@gm...>]

Hi,

> I extracted 3 regions of interest from 30 subjects using MarsBar. The m f=
ile
> program in the FAQ calculates the total number of voxels in the ROI and t=
he
> number of significantly activated voxels. For some subjects, the number o=
f
> voxels in the ROI is 0 (not the number of significantly activated voxels)
> and these regions are in the periphery of the brain. In reading the messa=
ge
> board I take this to mean that there are probably NaN values and thus I
> return an empty matrix.

Hmm.  Maybe you could try changing the default ROI resampling
behaviour to nearest neighbour; the default is linear interpolation,
and this can lose voxels at the edge of the brain with NaNs outside.=20
Try something like:

roi_file =3D 'my_roi.mat';
t_imgs =3D strvcat('spmT_0002.img', 'spmT_0003.img');
thresholds =3D [3.4 4.6];

roi_obj =3D maroi(roi_file);
roi_obj =3D spm_hold(roi_obj, 0); % setting resampling to nearest neighbour

y =3D getdata(roi_obj, t_imgs);
n_voxels =3D size(y, 2);

for i =3D 1:size(t_imgs, 1)
  pc_above_thresh(i) =3D sum(y(i,:) > thresholds(i)) / n_voxels * 100;
end

Best,

Matthew