Dear all,
I was looking for correlation with age in two groups using TBSS. After transforming the skeleton voxels showing significant regression back into standard space in each group, binarising and summing up all deprojected images, a pattern emerges. Many significant voxels in one group are positioned more peripherally, and while there is still a great degree of overlap, some voxels are present in a higher number of participants from one group. This difference could potentially be related to brain atrophy.
I am wondering what would be the best way to display these results and to investigate the difference further? I was thinking of testing for significant difference between the number of participants projecting to each voxel by applying the properties of a binomial distribution in a voxelwise manner, with p=0.5 (groups are equally sized) and n being the total number of subjects whose significant skeleton voxels project to each standard space voxel.
Would such an analysis make sense? Is there a way of performing it voxelwise automatically (the alternative being splitting the image and calculating the appropriate value manually for each value of n)? Would you suggest some better alternatives?
Thank you for your help and all the best,
Rok
|