Hi, 

I tried the W2MHS software and in general am very satisfied by the resulting white matter hyperintensity maps. 
I use thresholded versions of the pmap as masks for quantitative parameter maps.
One my goals is to make the distinction between periventricular lesions and lesions in deep white matter. For this purpose I derived a 'periventricular area' by inflating the ventricle mask (Vent_bin*.nii) and subtracting this from the original ventricle mask. However, the ventricle mask (Vent_bin*.nii) sometimes appears truncated and lower regions of the ventricle are excluded (this happens in the region growing part of the code: 'Constructing PV Estimates and Ventricle template'). The result is that my periventricular area is not very representative.

Is this a known issue? And would there be ways to improve the ventricle delineation?

I tried to do this by playing with some settings in the 'Constructing PV estimates and Ventricle Template' section, yet could not get a more complete ventricle mask without also obtaining CSF outside the ventricles.
I also tried the ALVIN lateral ventricle segmentation (which uses a predefined ventricle template). This way I obtained a lateral ventricle segmentation that includes the lower regions of the ventricles, but some other regions were excluded (probably because of the fact a template is used).

I attach an image of the ventricles as delineated by W2MHS and ALVIN to illustrate the issue.

Kind regards,

Thibo

The text in the attached image was not well visible. Here is a better version.
KR,

Thibo

Sorry for the delayed response. Thank you for the example. The issue is at the very least known to us now. Fortunately this doesn't happen on all images. I will look into ways to restrict the region growing algorithm from cutting away voxels that may be periventricular hyperintensities. If you leave some CSF in the assessed region of interest it ptrobaby wont cause much of a problem because CSF will very likely not be detected as hyperintense white matter.