Print

Print
Jiyang,

Yes, that could be an issue if you are attempting to quantify within structures. The GM ROI is certainly sparse and was largely aimed at getting a robust whole brain mean GM CBF with reduced PV effect. It may well be that you accept a lower GM threshold for that case. In practice, if you simply use a low GM threshold you might find more between subject variability due to PV effects - although these will also tend to average out by taking the mean over a whole lobe.

Michael


On 24 Apr 2018, at 00:47, Jiyang Jiang <[log in to unmask]> wrote:

Hi Michael,

Thanks a lot for your quick and detailed reply.

Just one more question: I had a look at the gm_roi.nii.gz in native_space folder. It seems to be somewhat sparse and discrete. wm_roi.nii.gz is better. I understand this is due to restricting to voxels with high probability of being GM. I was wondering if this would be a potential problem for segmenting CBF map into regions, especially into lobes and subcortical structures as you suggested to warp the ROI atlas to perfusion_calib which is masked by gm/wm_roi?

Kindest regards,
Jiyang

---
Michael Chappell MEng DPhil
    T: +44 1865 617657
Associate Professor, Institute of Biomedical Engineering, University of Oxford.
Director of Training, EPSRC-MRC CDT in Biomedical Imaging
Governing Body Fellow, Wolfson College, Oxford.
    http://www.wolfson.ox.ac.uk