 |
 |
Well, >However, a corollary would be to check if they are significantly different in the areas they shouldn't be, then that supports the hypothesis that they are comparable and can be used to measure similar things, correct? I think you misworded this sentence, if I got your intentions correctly. If two different images on average do *not* have significant differences, than they can be said to be able to measure the same things. If they are significant than they're different. >But I wanted to check if they are significantly different in same patient in areas such as frontal, parietal and temporal and it turns out they are not. Btw, I am using SUVr images for each Group. Ok, I've never got the occasion to analyze a dynamic amyloid scan. Indeed there's a chance that such an image would be similar to an FDG static uptake image. However, it still seems suspicious to me that, with a sample size of 65, you don't see any significant difference. Anyway, in this case, to take into account that each patient is used in the two scans, you need to perform a paired t-test, not a two samples one. however, be careful in interpreting the results. With paired t-test the statistical power increases wildly, therefore maybe blobs with t-values of 20 or 40 maybe relate to clinically insignificant differences. Hope this helps! Luca ________________________________ Da: Abhinay <[log in to unmask]> Inviato: giovedì 17 marzo 2016 22.14 A: PRESOTTO LUCA Cc: [log in to unmask] Oggetto: Re: R: [SPM] Horzcat error in SPM12 Hmm.. However, a corollary would be to check if they are significantly different in the areas they shouldn't be, then that supports the hypothesis that they are comparable and can be used to measure similar things, correct? I haven't clarified this, so it makes sense to mention it here, I am using perfusion imaging data from amyloid tracer imaging, so first 5 min images after injection of AV45, and when that is compared with FDG in frontal, parental and temporal the correlation is high. But I wanted to check if they are significantly different in same patient in areas such as frontal, parietal and temporal and it turns out they are not. Btw, I am using SUVr images for each group. Thanks. Abhinay Sent from iPhone. Please excuse typos and brevity On Mar 17, 2016, at 5:42 AM, PRESOTTO LUCA <[log in to unmask]<mailto:[log in to unmask]>> wrote: Hi, Basically, in SPM (and similar software/setups) a two-samples t-test tells you if there are statistically significant differences between the average image of all scans in Group 1 and the average image of all scans of Group 2. I don't see a case where this kind of comparison could give you an interesting information when the two groups are two PET images of different tracers. I'm pretty sure that if you study the two available contrasts [1 -1] and [-1 1] you're going to find that every voxel is significant! Also, if you are using the same patients in the two groups you're violating the Independence assumption, which instead, most likely, is something that you want to exploit here. (Maybe you'd like to study the correlation between amyloid load and FDG damage). Luca ________________________________ Da: Aj <[log in to unmask]<mailto:[log in to unmask]>> Inviato: mercoledì 16 marzo 2016 23.46 A: PRESOTTO LUCA; [log in to unmask]<mailto:[log in to unmask]> Oggetto: Re: R: [SPM] Horzcat error in SPM12 Hi Luca, I am not an expert. However, I would like to learn why images from different tracers cannot be included in two different groups in GLM. I did a sanity check by running two groups independently in Neurostat and I do get similar findings when I ran these two groups independently against the Neurostat database. And these findings are comparable if I run a similar analysis using SPM. Abhinay, Sent from my iPhone On Mar 16, 2016, at 3:36 PM, PRESOTTO LUCA <[log in to unmask]<mailto:[log in to unmask]>> wrote: Hi, If you have only two groups, ANOVA and two samples t-test are identical. My doubt is related to the use of different tracers. Since Spm uses a linear model, I doubt that by putting different tracers in the two groups you can achieve what you're looking for. Also, you wouldn't be correctly modeling the patient effect in this design. Inviata dal mio Windows Phone ________________________________ Da: Abhinay Joshi<mailto:[log in to unmask]> Inviato: ?16/?03/?2016 18:42 A: PRESOTTO LUCA<mailto:[log in to unmask]> Oggetto: Re: [SPM] Horzcat error in SPM12 Oh. Okay thanks. I tried entering all the 130 values. 65 first and then all those 65 again. And it ran through. I understand your concern. So, I moved to two-sample t-test with covariates. Is that what you were trying to suggest Or were raising a questions related to the comparison of the tracers? Abhinay On Wed, Mar 16, 2016 at 12:38 PM, PRESOTTO LUCA <[log in to unmask]<mailto:[log in to unmask]>> wrote: Still, in an ANOVA design you are going to define two cells, each with 65 scans. Therefore, when you define the covariate, you need to enter the age for every PET image you entered, not for each patient. Therefore, assuming you entered in each cell the patients in the same order, you need to copy two times the patients age vector in the covariate (i.e.: a 130 numbers vector). BTW, I'm not so sure that the ANOVA option performs the kind of statistical analysis that you mean to. Luca ________________________________ Da: Abhinay Joshi <[log in to unmask]<mailto:[log in to unmask]>> Inviato: mercoledì 16 marzo 2016 15.22 A: PRESOTTO LUCA Oggetto: Re: [SPM] Horzcat error in SPM12 Thank you Luca. I must clarify same patient had an FDG scan and an amyloid imaging on different days. So, the age column will have just 65 rows and 1 column, as I am trying to compare the FDG vs. Amyloid uptake in different brain areas w.r.t age. At some point I would also like to add gender as a covariate. Hope that makes sense? Abhinay On Wed, Mar 16, 2016 at 9:46 AM, PRESOTTO LUCA <[log in to unmask]<mailto:[log in to unmask]>> wrote: Dear Abhinay, Most likely, You didn't input the age correctly indeed. The age vector should be as long as the number of images you've given in input. (65x2 in this case). BTW, what kind of statistical test are you performing, where you input FDG images in a Group and amyloid ones in another? Are you sure your design is correct? Have a nice day, Luca ________________________________________ Da: SPM (Statistical Parametric Mapping) <[log in to unmask]<mailto:[log in to unmask]>> per conto di Aj <[log in to unmask]<mailto:[log in to unmask]>> Inviato: martedì 15 marzo 2016 23.11 A: [log in to unmask]<mailto:[log in to unmask]> Oggetto: [SPM] Horzcat error in SPM12 Dear All, I am trying to perform one way Anova , with ANCOVA option selected as yes, as I am doing this on the PET data. I have two sets of images for 65 subjects, FDG images, amyloid tracer (florbetapir) images and age of these cases. There are 65 scans in each grp. ,and I am trying to us age as a covariate. There is age for each subject so 65 values in the age column. When I try to run the model by selecting following options Centering : As implied by Ancova Global calculation: Mean Variance : Equal I get a following error about HORZCAT. Does this error mean I didn't include the age data correctly? ------------------------------------------------------------------------ Running job #1 ------------------------------------------------------------------------ Running 'Factorial design specification' Mapping files : ...done Calculating globals : ...done Design configuration : Failed 'Factorial design specification' Error using horzcat Dimensions of matrices being concatenated are not consistent. In file "C:\Analysis\Spm12\config\spm_run_factorial_design.m" (v6219), function "spm_run_factorial_design" at line 906. -- Abhinay D. Joshi Wayne, PA Cellular:-817-995-3962<tel:817-995-3962> Sent from my iPhone Rispetta l'ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail. -- Abhinay D. Joshi Wayne, PA Cellular:-817-995-3962<tel:817-995-3962> Rispetta l'ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail. -- Abhinay D. Joshi Wayne, PA Cellular:-817-995-3962 Rispetta l'ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.