Re: [Freesurfer] Longitudinal data from different sequences and scanners

[James Gullickson]

External Email - Use Caution

Martin,

Thanks for the feedback. Given our data set (1mm^3 timepoint1 and 0.8mm^3
timepoint 2), what would be the best way to salvage this data and look for
longitudinal changes? Would it be possible to upsample/downsample the
images so that they are the same resolution (i.e. using mriconvert)? Also,
would a normalized measure like ventricle-brain-ratio theoretically be
resistant to these scanner/acquisition induced biases?

Thanks,

James

On Mon, Apr 16, 2018 at 5:15 AM, Martin Reuter 
wrote:

> Hi,
>
> never switch hardware or protocols in a longitudinal study. This is a good
> example for what happens: you will find effects that are caused by
> different acquisition rather than anatomical differences. It will be
> impossible to disentangle true change from scanner/acquisition induced
> changes, especially with only a few time points and no or little subjects
> scanned on both scanners or with both protocols.
>
> Also 5mm resolution is too low for freesurfer and anything can come out of
> that. It won’t be reliable.
>
> Best, Martin
>
>
> On 6. Apr 2018, at 21:21, Matthew Grecsek  wrote:
>
> Here are some statistics I generated for one of my subjects processed
> through the longitudinal stream with T0 as the initial scan and T1 18 mos
> later. (not sure if the inline tables will be properly formatted so I
> attached text files of them).
>
> Unfortunately the only initial scan I had was a 5mm resolution for T0
> versus 1mm for T1. I understand FS recommends a resolution not exceeding
> 1.5mm, but we gave it a try anyway to see if there was anything useful. My
> expectation was that the stats would be off by a consistent ratio due to
> the different resolutions, however I was surprised by the variability.
>
> In particular, as James found, for some ROIs there are net increases in
> cortical thickness and brain volume over time.
>
> Is this simply a factor that the algorithms are confused by the different
> image resolutions and therefore no possible longitudinal study can reliably
> be presumed in this circumstance?
>
> Should we expect similar anomalies in cross-sectional studies, such as if
> my subjects have 1mm resolutions and a collaborating institution has 0.8mm
> subjects?
>
> Cheers,
>
> -Matt
>
> Aseg Stats
> Measure:volume T0 T1 Base T0.long.base T1.long.base
> Left-Lateral-Ventricle 9,455.2   13,085.8
> 12,344.2   10,928.1   13,268.0
> Left-Inf-Lat-Vent   61.8 174.1
> 171.7   90.3 303.3
> Left-Cerebellum-White-Matter   26,892.5   18,563.1
>   17,066.4   23,905.6   17,597.8
> Left-Cerebellum-Cortex   57,390.8   66,458.7
> 64,631.3   60,374.1   65,898.6
> Left-Thalamus-Proper   10,757.3 9,094.2
> 9,918.5   10,350.8 9,588.3
> Left-Caudate 3,619.4 3,673.8
> 3,608.8 3,588.7 3,872.7
> Left-Putamen 5,463.7 5,439.2
> 5,171.0 5,602.0 5,721.7
> Left-Pallidum 2,380.9 2,208.6
> 1,897.1 2,270.8 2,147.9
> 3rd-Ventricle 1,181.8 1,174.6
> 1,328.7 1,432.2 1,416.9
> 4th-Ventricle 1,209.1 1,602.0
> 1,547.7 1,322.4 1,981.1
> Brain-Stem   25,154.2   26,141.3   
> 25,890.7
>   25,988.7   25,954.6
> Left-Hippocampus 4,205.3 4,335.6
> 4,378.0 4,478.2 4,407.2
> Left-Amygdala 1,488.1 1,725.9
> 1,588.6 1,531.3 1,638.0
> CSF 1,440.9 1,321.0 1,551.0
> 1,910.1 1,394.7
> Left-Accumbens-area 295.0 273.2
> 327.7 274.9 380.0
> Left-VentralDC 5,086.1 4,935.5
> 5,429.6 5,106.9 5,041.7
> Left-vessel- 16.6
> 7.7- 48.5
> Left-choroid-plexus 221.6 419.4
> 271.7 429.5 734.2
> Right-Lateral-Ventricle 6,465.1 9,581.5
> 8,695.5 6,893.5 9,713.9
> Right-Inf-Lat-Vent 360.0 333.7
> 408.5 400.5 518.0
> 

Re: [Freesurfer] Longitudinal data from different sequences and scanners

[James Gullickson]

All,

Just wanted to follow up on this message. Does anyone have recommendations
for the best way to compare this data longitudinally?

Thanks very much,

James

On Thu, Mar 29, 2018 at 1:19 PM, James Gullickson  wrote:

> All,
>
> Our lab is working with longitudinal (two time points) data from a cohort
> of around 130 individuals. We are attempting to make comparisons between
> time points 1 and 2 on measures like total brain volume, ventricular
> volume, ventricle-brain-ratio, and cortical thickness. However, we have run
> into an issue--we are seeing net increases in cortical thickness and total
> brain volume over time (which seems biologically implausible, given our age
> range of ~30-60 yrs old). We think it may have something to do with the
> fact that the T1w data from each time point was acquired with slightly
> different parameters and on different scanners, possibly leading to a
> rounding error in quantification of volumes/thickness. Timepoint 1 data are
> 1.0mm isotropic and were acquired on a 3T Siemens Tim Trio with 12ch
> headcoil. Timepoint 2 data are 0.8mm isotropic and were acquired on the
> same scanner, but which went through an upgrade to a Prisma Fit between
> timepoints, using a 32ch coil.
>
> Currently we have been comparing data from the cross sectional stream. We
> would like to use the longitudinal stream if that would improve results,
> but we saw this post that cautioned against it: https://www.mail-archive.c
> om/freesurfer@nmr.mgh.harvard.edu/msg52992.html
>
> What would be your recommendation for comparing this data longitudinally?
> Any thoughts on why we are seeing net increases in volume/thickness, and
> how to avoid that? One idea we had was perhaps degrading each image by
> rigid co-registration and then bringing each image into the halfway space
> between the two (as FSL's SIENA does).
>
> Thanks,
>
> James
>
>
___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.

[Freesurfer] Longitudinal data from different sequences and scanners

[James Gullickson]

 All,

Our lab is working with longitudinal (two time points) data from a cohort
of around 130 individuals. We are attempting to make comparisons between
time points 1 and 2 on measures like total brain volume, ventricular
volume, ventricle-brain-ratio, and cortical thickness. However, we have run
into an issue--we are seeing net increases in cortical thickness and total
brain volume over time (which seems biologically implausible, given our age
range of ~30-60 yrs old). We think it may have something to do with the
fact that the T1w data from each time point was acquired with slightly
different parameters and on different scanners, possibly leading to a
rounding error in quantification of volumes/thickness. Timepoint 1 data are
1.0mm isotropic and were acquired on a 3T Siemens Tim Trio with 12ch
headcoil. Timepoint 2 data are 0.8mm isotropic and were acquired on the
same scanner, but which went through an upgrade to a Prisma Fit between
timepoints, using a 32ch coil.

Currently we have been comparing data from the cross sectional stream. We
would like to use the longitudinal stream if that would improve results,
but we saw this post that cautioned against it: https://www.mail-archive.c
om/freesurfer@nmr.mgh.harvard.edu/msg52992.html

What would be your recommendation for comparing this data longitudinally?
Any thoughts on why we are seeing net increases in volume/thickness, and
how to avoid that? One idea we had was perhaps degrading each image by
rigid co-registration and then bringing each image into the halfway space
between the two (as FSL's SIENA does).

Thanks,

James
___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer


The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.