Hi Matt,
The files you reference are not official FSL files ( http://neuro.debian.net/pkgs/fsl-harvard-oxford-cortical-lateralized-atlas.html ) although with respect to the first issue: the index you find in the xml file is equal to the volume number in the 4D files, but is one less than the label number in the 3D label (or hard segmentation) files. It is not an error. This is now indicated more clearly in the FSL wiki at:
https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases
For the second error it might be worth contacting the neuro-debian team as they are the maintainers of these files.
Kind Regards
Matthew
> Hi. I found two errors in the Harvard-Oxford atlas files. What follows is a little fiddly and detailed, per necessity.
>
> Error 1. Off-by-one error between the .xml and .nii files: The index values in the .nii files count from 1, with the value 0 implicitly denoting "not labelled" (except see Error 2 below). In contrast, in the .xml files that give the region names corresponding to the indices, meaningful index values start at 0. Example: the HarvardOxford-Cortical-Lateralized.xml file says index 0 corresponds to Left Frontal Pole, whereas in the .nii files (except see Error 2 below), index value 0 occurs outside the brain and in some other unlabelled areas. Similarly, the HarvardOxford-Cortical-Lateralized.xml file says that index 6 corresponds to Left Middle Frontal Gyrus, but in HarvardOxford-cortl-maxprob-thr0-1mm.nii.gz, index value 6 occurs in Right Superior Frontal Gyrus. The HarvardOxford-Cortical-Lateralized.xml file assigns an index value of 5 to Right Superior Frontal Gyrus - i.e. off-by-one error between the .xml and the .nii files.
>
> Error 2. Label error in two of the .nii files:
> HarvardOxford-cortl-maxprob-thr0-1mm.nii.gz
> HarvardOxford-cortl-maxprob-thr0-2mm.nii.gz
> The region given index value 1 in these two .nii files should be Left Frontal Pole according to HarvardOxford-Cortical-Lateralized.xml (after correcting for the off-by-one error described in Error 1 above). However, in those two files, the entire region outside the brain as well as other unlabelled regions are given the index value of 1, whereas those unlabelled regions should actually have an index value of 0. I have built corrected versions of those two .nii files with index 0 denoting unlabelled regions and index 1 denoting just Left Frontal Pole.
>
> To the FSL maintainers: Do you want these corrected files?
>
> with kind regards
> Matt
>
> ----------
> Matthew Brown, PhD
> Research Associate, Dept. Psychiatry, University of Alberta
> Edmonton, Alberta, Canada
|