 |
 |
Dear Laure, The use of global calculation or not depends on your tracer kinetics, and if you have normalized the uptake in advance (Standarized uptake value, or another parametic image obtained from pharmacokinetic modelling such as distribution volume or binding potential). For example, in our center for the animal studies we usually normalize the uptake using SUV (corrected by injected dose and body weight). Then, with tracers such as 18F-FDG we use proportional scaling using the mean uptake of the brain (which can be obtained in SPM using "global calculation"), with the assumption that the mean brain uptake, and therefore the basal metabolic brain activity between subjects is the same. So, you will obtain regional differences in metabolism. However, in other tracers such as 11C-PK11195, for TSPO overexpresion (microglia activation), we do not apply proportional scaling, and therefore the global calculation parameter remains as "omit". About the "overall grand mean scaling": this options simply multiplies your uptake signal for a fixed value (e.g. 50). This was originally intended to work with tracers such as O2-water. One example of our department, we performed a study with water injecting 500MBq per subject, the obtained images were analyzed in SPM performing proportional scaling (using the mean brain uptake obtained from global calculation : mean), and overall gran mean scaling to 50. This will give a final uptake value first corrected for the uptake (proportional scaling) and later scaled by factor 50 to mimic the physiological values that may be expected in brain perfusion. In summary: - Decide if your tracer and research question requires or not the use or proportional scaling or ANCOVA. - If so, decide how this correction will be done (whole brain uptake, specific brain regions, collateral hemisphere, etc...). Then, for whole brain uptake you can use the option 'mean'in global calculation. For most of the other cases you have to introduce manually a vector with the correction factor - Apply overall grand mean scaling only if you are going to extract raw values from the image and results, and the scaling factor have a physiological meaning. If the input images were SUV or BP images, then the use of grand mean scaling make no sense. Since this should not affect the statistics, you should not care much about this option in most of the situations. I hope this information will help you. If you have specific doubts with your project, or something was not clear enough just let us know. Best regards, David Vállez On Tue, Jun 23, 2015 at 5:38 PM, Laure SA <[log in to unmask]> wrote: > Dear all, > > I have a (quite) simple question about SPM PET analyses. > I have been running simple regression between PET data and a covariate > (let's say static amyloid PET images and a simple cognitive score > performance) in a group of participants to investigate the possible > relationship between the cerebral uptake and my variable of interest. > > I used the multiple regression design, select my scans (normalised and > smoothed) and enter my covariate vector (so far so good). > I am not quite sure of the optimal following parameters to enter in my > model: > > 1- Global calculation: I used to put "omit", but that may be a mistake? > 2- The "overall grand mean scaling" option for global normalisation: I > don't really understand the point of such scaling, but maybe this would be > important for me to use? any recommandations? and then should I add the > normalisation option (proportional/AnCOVA)? > > Thank you for your help on this. I really want to avoid silly mistakes. > And sorry if these questions have already been addressed, could you then > indicate me to the corresponding answer? > > Laure >