Dear John and Philipp,
thanks for your swift and helpful replies.
Best,
Falk
________________________________
Von: SPM (Statistical Parametric Mapping) <[log in to unmask]> im Auftrag von Ashburner, John <[log in to unmask]>
Gesendet: Mittwoch, 26. Februar 2020 11:22:41
An: [log in to unmask]
Betreff: Re: [SPM] AW: [SPM] Reproducing results from "PRoNTo: Pattern Recognition for Neuroimaging Toolbox"
I'd suggest trying the Shoot toolbox, rather than Dartel. It's usage is similar to that of Dartel in that it works on "imported" tissue maps generated by the segmentation. The functionality you need to generate the same representations for machine learning are also available to you. Page 195 of the SPM12 manual gives a limited amount of information about how to use it.
Best regards,
-John
________________________________
From: SPM (Statistical Parametric Mapping) <[log in to unmask]> on behalf of Saemann, Philipp <[log in to unmask]>
Sent: 25 February 2020 15:52
To: [log in to unmask] <[log in to unmask]>
Subject: [SPM] AW: [SPM] Reproducing results from "PRoNTo: Pattern Recognition for Neuroimaging Toolbox"
Hi Falk,
this all sounds as if DARTEL was used after Unified Segmentation.
So best, check the DARTEL Pipeline in the SPM12 Manual and
process your T1-WI in this way, so first segmentation, and then
coregistration between the subjects using DARTEL.
DARTEL usually results in an output at the resolution of (1.5 mm)^3.
So you should get familiar with DARTEL.
What data exactly to feed the pattern recognition with, is a second
question - several ways to go here. My first idea would be to
run a GLM an Regress out General covariates such as ICV, Age,
Age2 and sex and potential interaction Terms and use the residual
maps. But best check this in the original paper you mention.
Best,
Philipp
________________________________
Von: SPM (Statistical Parametric Mapping) <[log in to unmask]> im Auftrag von [log in to unmask] <[log in to unmask]>
Gesendet: Dienstag, 25. Februar 2020 16:46
An: [log in to unmask]
Betreff: [SPM] Reproducing results from "PRoNTo: Pattern Recognition for Neuroimaging Toolbox"
Dear all,
We want to make use of the Pattern Recognition for Neuroimaging Toolbox (PRoNTo) for SPM using structural MRI data. Currently, we don’t have a specific research question in mind, however, we want to see what we may expect from the results using PRoNTo.
A student of us should reproduce the results of the paper [1], specifically the question “Which features lead to the best discrimination between the considered groups?” (p. 11). Unfortunately, we are already stuck at the pre-processing of the data and neither the paper, nor the manual or course material seems helpful in this regard. Maybe someone else can help to set everything up accordingly. In the paper [1] it says:
“Images were segmented into different tissue types via the “new segmentation” algorithm (Ashburner and Friston, 2005) implemented for SPM8. Rigidly aligned grey and white matter maps, down-sampled to 1.5 mm isotropic resolution, were then used to diffeomorphically register all subjects to their common average, using a matching term that assumed a multinomial distribution (Ashburner and Friston, 2009). Registration involved estimating initial velocities, from which the deformations were computed by a geodesic shooting procedure (Ashburner and Friston, 2011).”
The segmentation is clear. However, to what data should the grey and white matter maps be rigidly aligned? How (and probably why) would I downsample the data to 1.5mm iso? Should the native T1 be downsampled or the segmentations? How can the data (and what data? Native T1? The downsampled data? The segmentations?) be diffeomorphically registered to their common average with SPM?
Best regards,
Falk
[1] Schrouff et al. (2013) PRoNTo: Pattern Recognition for Neuroimaging Toolbox
...............................................
[klinikum_logo_schmal_e-mail_blau]
University Clinic for Neurology
Otto-von-Guericke-university Magdeburg
Medical faculty
Leipziger Str. 44
39120 Magdeburg
Phone +49-391-6117-512
[log in to unmask]<mailto:[log in to unmask]>
http://www.kneu.ovgu.de/kneu.html<https://eur01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.kneu.ovgu.de%2Fkneu.html&data=02%7C01%7C%7Ceb425dcaff0049fc0c4e08d7ba0ab1ba%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C637182427402860119&sdata=j0G26FUa5tXl2iVMun9dI0B2X9O86iYTu%2FKQY%2BUSMPA%3D&reserved=0>
|