JISCMail - SPM Archives

Hi Marek

In a tutorial style (with code in sup material): Pernet 2014 Misconceptions in the use of the General Linear Model applied to functional MRI: a tutorial for junior neuro-imagers.

http://journal.frontiersin.org/article/10.3389/fnins.2014.00001/full

The best paper ever IMO (but paywalled): Poline and Brett 2012 The general linear model and fMRI: Does love last forever?

http://www.sciencedirect.com/science/article/pii/S1053811912001607

Dr Cyril Pernet,
Senior Academic Fellow, Neuroimaging Sciences
Centre for Clinical Brain Sciences (CCBS)

The University of Edinburgh
Chancellor's Building, Room GU426D
49 Little France Crescent
Edinburgh EH16 4SB
[log in to unmask]
http://www.sbirc.ed.ac.uk/cyril
http://www.ed.ac.uk/edinburgh-imaging

Thank you Torben and guys,

Could you explain me a bit more the formula of efficiency: inv(c’*pinv(X'*X)*c), or probably suggest some reading (I used to have some linear algebra very long time ago, so I hope I would be able to understand, but at the moment I do not...).

What would be the X and c in the formula? What would be the result, and how to understand it?

I'll be grateful for the explanation.

Marek

W dniu 2016-10-26 o 13:42, Torben Lund pisze:

[log in to unmask]" type="cite">Dear List, Marek and Marko

I can see that the condition approach can be quite difficult given that it is a rapid event related design. How about comparing the efficiency inv(c’*pinv(X'*X)*c) of the regressors for the different conditions before and after inclusion of the scan nulling regressors? As the stimulus duration is different for the different conditions they cannot easily be compared with each other. But if the efficiency for one condition is reduced by 90 % and only by 20% for another condition then you have a problem.

Best

Torben

Torben Ellegaard Lund
Associate Professor, PhD
Center of Functionally Integrative Neuroscience (CFIN)
Aarhus University
Aarhus University Hospital
Building 10G, 5th floor, room 31
Noerrebrogade 44
8000 Aarhus C
Denmark
Phone: +45 7846 4380
Fax: +45 7846 4400
http://www.cfin.au.dk

[log in to unmask] (for official stuff)

[log in to unmask] (for faster response)

Den 26. okt. 2016 kl. 11.22 skrev Marko Wilke <[log in to unmask]>:

Marek,

I may have gotten your explanation wrong but I think you are taking a very complicated approach - why not simply do what the reviewer suggested, i.e., count number of scrubbed scans per condition? I think trying to determine whether the impact of scrubbing is different would only be my second step, after ascertaining that an impact is to be expected in the first place...

Or in other words - if all subjects had a comparable number of scans removed, and these were randomly distributed across conditions - why bother determining BOLD signal amplitude etc.?

Cheers
Marko

Marek Wypych wrote:

Dear SPMers

In the review of our paper on children (who move quite a lot, and we
used ART to regress out moved volumes) one of the Reviewers asked: “Did
the authors verify that the number of “rejected” scans was equally
distributed across conditions?” So we want to check it now.

In the experiment we had four conditions: two visual and two auditory.
The experimental design was quite rapid – visual trial lasted about 2
seconds and auditory about 4 seconds. The ITI was randomized and
differed between 4-7 seconds. Thus, after convolution of the design with
the standard HRF (which has max about 12 seconds after the onset), the
conditions in the model partly overlap, and “rejection“ of one volume
may affect few conditions in the same time.

My idea is to read the (absolute) values of predicted BOLD signal in the
moment of volume rejection across all conditions from design matrix in
the estimated SPM.mat (SPM.xX.X) and then to sum it within each condition.
1. First: do you think it is a good idea? Or maybe you have other
suggestions?

2. Assuming this is a reasonable idea, I have the second question. As I
have already mentioned the trials differed in length – thus the
predicted BOLD will have higher amplitude in longer (auditory) trials,
than in shorter (visual) trials. Should I normalize the values by the
length of the trials? As I understand GLM normalizes the condition to
calculate betas (am I right?), so to estimate the influence of motion on
betas I should also normalize it?

I’ll be grateful for your comments and suggestions.

Best regards
Marek

--
____________________________________________________
Prof. Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Universitäts-Kinderklinik
Abt. III (Neuropädiatrie)

Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
University Children's Hospital
Dept. III (Pediatric Neurology)

Hoppe-Seyler-Str. 1
D - 72076 Tübingen, Germany
Tel. +49 7071 29-83416
Fax +49 7071 29-5473
[log in to unmask]

http://www.medizin.uni-tuebingen.de/kinder/epn/
____________________________________________________

-- 
Marek Wypych PhD
Laboratory of Brain Imaging (LOBI)
Neurobiology Center
Nencki Institute of Experimental Biology
Pasteur 3, 02-093 Warsaw, Poland
Tel.: +48 22 5892 550
http://lobi.nencki.gov.pl/

Ta wiadomość została sprawdzona na obecność wirusów przez oprogramowanie antywirusowe Avast.
www.avast.com