I fully agree with Matt.
Within a certain range, the number of seed voxels and the waytotal increase fairly exponentially with the target mask size. (That is part of the effect you see in Fig. 19.11 of my chapter in Heidi's & Tim's book.) However, that is not very useful in general. If your target becaomes too big, you may not see any further increase.
>Since I use exclusion mask the waytotal value is less then number of voxels in the seed * number of samples sending out (e.g. 5000).
I think that is the normal behaviour: The total number of samples sent out should be more than the total number of samples that make it from seed to target - regardless of the use of exclusion masks.
Cheers-
Andreas
________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Matt Glasser [[log in to unmask]]
Gesendet: Donnerstag, 12. November 2009 02:46
An: [log in to unmask]
Betreff: Re: [FSL] fdt_path normalization
This is a difficult question, as the target mask size is not linearly related to the connectivity strength like the seed mask size is. What is the reason that you are concerned about the target mask size? I would stick to normalization by waytotal.
Thresholding is a difficult and perhaps both study specific and pathway specific question. I would try a number of thresholds on your subjects and pick the one that shows what you believe to be the pathway of interest and excludes other weak pathways. Unfortunately there is not a statistically rigorous way of doing this at this time.
Are you interested in FA values along your tract of interest? It is not clear to me what you are trying to do, and what question you are trying to answer.
Peace,
Matt.
________________________________
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Keihaninejad, Shiva
Sent: Wednesday, November 11, 2009 10:27 AM
To: [log in to unmask]
Subject: [FSL] fdt_path normalization
Dear FSL experts,
I’m looking at connectivity between 2 ROIs (seed –target) using exclusion mask and termination mask. The seed and target regions have different size (volume) among the subjects. I want to compare the connectivity between these 2 ROIs in my database. As I followed the previous emails related to this topic (normalization), the FSL experts have recommended to normalise the fdt_path image with the waytotal or number of voxels in the seed mask * number of samples per voxel (e.g. 5000). This type of normalization will just consider the effect of the seed mask size and not the target mask size. I reckon that the connectivity that I get in the fdt_path is affected by the target mask size as well and in the analysis (using fslstats) this mask size is important.
My question is how I can include the target mask size in my study? Since I use exclusion mask the waytotal value is less then number of voxels in the seed * number of samples sending out (e.g. 5000). I was thinking of normalizing the fdt_path image by (ROI_target+ROI_seed)*5000, but this is not a correct solution, is it?
Another thing is thresholding after the normalization:
After the normalization with waytotal value, I get this values: fslstats fdt_paths_norm.nii.gz -M -R -r
0.015307 0.000000 0.269559 0.000270 0.122380
What’s your suggestion for the threshold value for this subject?
Some of the FSl fellows use FA map in their analysis as well, could anyone help me to understand why we need to include the FA map?
Any advice, comments would be greatly appreciated.
Shiva
|