Hi Mehul,

Freesurfer uses a partial volume calculation to determine the volumes of structures from the aseg. Matlab is not doing this, and I'm not sure about slicer but if it matches matlab than I doubt it has a partial volume correction either. I would say that the values that are corrected for partial volumes are more accurate.

Jenni

On Wed, 13 Jun 2007, Mehul Sampat wrote:

Hi all,

We have a mismatch in the volumes given in the aseg.stats by freesurfer0 and the volumes computed from the aseg.mgz with other programs like Slicer and Matlab I will use the data of subject bert that is already segmented in freesurfer. In the aseg.stats file (for the subject bert) the volume of the Brain Stem is 24490 mm^3 ( 24490 voxels ; segmentation id = 16)

If I measure the volume of the brain-stem in matlab i get 25075 voxels the code used for this is :

---

[data,m] = load_mgh('/Applications/freesurfer/subjects/bert/mri/aseg.mgh'); temp = (data ==16); sum(sum(sum(temp)))

---

When I measure the brain stem volume with Slicer 2.6, it gives 25.075 mL which matches exactly with the Matlab result. (A colleague also tried it using Analyze and the result matches those of Matlab and Slicer) Similar trends is seen for other structures also: (for example for the left Amygdala aseg.stats reports 1723 while with matlab we get 1728).

Question is: Which measurement is more accurate ? I can see that the Matlab measurement is an approximation (for example the structure may only cover half the voxel but with matlab we assume it covers the entire voxel). Does freesurfer use a sophisticated method for measuring the volumes ? If this is true then it implies that the volume measurements in Slicer and Analyze are sub-optimal...

Thanks Mehul