Judd wrote:

> > That would be useful if storing data in a few masks. If I
> understand you correctly, then for my problem it would yield a
> large amount of duplication of data (as each voxel may be
> contained in hundreds of ROIs). For a few ROIs with little
> overlap it may indeed be more efficient though to use a sparse
> representation.
>
> The formats seem equivalent and inter-convertable to me so I
> either must not have explained it clearly or I really don't
> understand what you're trying to compute. I don't see why there
> would ever be any duplication using either approach (even in
> the presence of extensive overlap).

Sorry, I think I read your proposal wrong. If I understand it correctly now then you propose that an ROI can not only contain integer indices but data of any type.

This may be useful, however:
- it still complicates matters further. For example you would have to store the data type of the values that are stored.
- the proposal for masks does support different sizes of ROIs. However for the application that you mentioned (time courses in ROIs) it seems to me that is this can be stored in a simple matrix ("number of voxels" x "number of time points"). So why not store this data in the mask in a .1D file, while storing the mask itself in either a BRIK or another file that contains the indices of the selected voxels?

> I was mostly concerned about keeping track of the position of
> each mask because I was thinking about this from an image
> filtering point of view. Anyway, it sounds like you're pretty
> clear about where you want to go with this.

Indeed, I have a rather specific purpose in mind. Anyway, thanks a lot for your comments and sharing your thoughts.

cheers,
Nick