 |
 |
There is a trade-off between attempting to match the image intensities with those of the template, and keeping the deformations as smooth or as small as possible. The component that keeps the warps smooth is the regularisation, so if you regularise more, then it will give less deformation. Note that the part that assesses image similarity is based on the mean-squared difference. This relies on the pattern of intensity in the individual scans being similar to that in the template. I don't know anything about the distribution of uptake in your mystery F18 compound, but the pattern of uptake of PIB does not really resemble the pattern of O15 distribution of the scans that constitute the PET template. If you have dynamic data, then you could maybe add up the first few frames and estimate the spatial normalisation parameters from these (as their signal is likely to be more representative of typical blood-flow). Best regards, -John On Tuesday 27 January 2009 15:40, Abhinay Joshi wrote: > Hello Experts, > > > I am still using SPM 2. I was trying to normalize the PET data (data for > Alzheimer's disease F18 compound) to a PET template, I used "write to > normalize" option and used all the default settings, though the > normalization is good the intensity in some of the brain areas is > underestimated or overestimated. I thought of using the "parameter > estimations" option. > In this I use "no weighting" and then "high regularization", this time the > spatial fitting us good (same like the earlier), but the regional > assessment is better. That means the values in the regions are as per > expected. By regional assessment I mean the images for Healthy control > provides less intense values in grey matter and the AD patient images shows > high intensity in the grey matter. So I conclude that once I use high > regularization the quantification closely resemble the qualitative read. > > But I would like to know what exactly High regularization does and is it a > better approach ? > > Thank you. > > Best Regards,