quick question about "n/a" in mean_qCBF_StandardSpace_Hammers_*

[Timothy Ellmore]

Quick question: what does "n/a” signify in the output file mean_qCBF_StandardSpace_Hammers_*.tsv? 

Does it indicate that these region are not covered in this particular atlas or does it mean that the value is not computable because of the presence of artifacts or lack of coverage? 

I’m getting a lot of “n/a” for all subjects in the regions listed below. 

For the ventricles it’s understandable there would be no meaningful values but there are other regions like substantial nigra and nucleus accumbens where all subjects have “n/a” 

Thanks,

Tim

---------------------------------- 

brainstem_excluding_substantia_nigra_B

,brainstem_excluding_substantia_nigra_L,

brainstem_excluding_substantia_nigra_R

,

nucleus_accumbens_B

,nucleus_accumbens_L,

nucleus_accumbens_R, pallidum_B, pallidum_L, pallidum_R, corpus_callosum_B, corpus_callosum_L, corpus_callosum_R, Lateral_ventricle_excluding_temporal_horn_B, Lateral_ventricle_excluding_temporal_horn_L ,Lateral_ventricle_excluding_temporal_horn_R, Lateral_ventricle_temporal_horn_B, Lateral_ventricle_temporal_horn_L, Lateral_ventricle_temporal_horn_R, Third_ventricle_B, Third_ventricle_L, Third_ventricle_R, substantia_nigra_B, substantia_nigra_L, substantia_nigra_R

[Jan Petr]

Shortly: It means that there are not enough voxels to evaluate the mean CBF.

Each TSV file only has columns with the regions that are in the atlas. However, we only output the mean CBF value if there are enough ASL voxels in that region in that particular subject (calculated in millilitres). This might not be the case for several reasons: FOV does not cover it, the MaskVascular does not contain the region (or only a few voxels), etc. Also, a general mask is used for all subjects in case of differing FOV between subjects. Voxels have to be present in the FOV of at least 95% subjects, if not (bad FOV in more than 5% of subjects), then the analysis region is restricted to all - see the ExploreASL Neuroimage paper for more information about this.

That was in general and nicely explains the ventricles as you have noticed. For the other regions, another thing comes into play and that is the GM map. The mean_qCBF is evaluated in GM only. Brainstem and corpus callosum doesn't include much GM (or at least not in the SPM segmentation), so effectively after using all the masks mentioned above and a GM mask, there are too few voxels left in the region and thus the region is assigned n/a. 

There are two ways to fix this. Either decreasing the threshold for counting a region - but that's not really a good idea, since then you would have results based on a few voxels only.

Or you can re-run the population module for whole-brain instead of for GM only (I would have to look up how to activate this). And you actually don't really want that either because then the mean_qCBF value would be mostly driven by the WM CBF (which is rather low) and by the actual amount of GM voxels in that region (which only points to imperfect segmentation/registration). 

So you just ignore these regions, unless you want to take the risks above - which can be totally justified in some cases. We can discuss this further then if needed.

best,

Jan