Dear Evan, The atlas is designed to be symmetric. Is it possible that the asymmetry is in your data? Another FS user (Chris Vriend, CCed) reported a similar effect (you can find his massage in the archive). Doug: any further ideas? Cheers, /E
Juan Eugenio Iglesias Senior research fellow CMIC (UCL), MGH (HMS) and CSAIL (MIT) http://www.jeiglesias.com
From: freesurfer-bounces@nmr.mgh.harvard.edu on behalf of "Gallagher, Evan (NGG)" Evan.Gallagher2@pennmedicine.upenn.edu Reply-To: Freesurfer support list freesurfer@nmr.mgh.harvard.edu Date: Wednesday, September 30, 2020 at 14:18 To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Subject: [Freesurfer] Right-side bias in thalamic segmentation pipeline?
External Email - Use Caution Hi all,
I’ve been working on a project that involves the thalamic segmentation pipeline published by Iglesias et al. in 2018 in NeuroImage (“segmentThalamicNueclei.sh”). While the pipeline itself works great, we have noticed that the right hemisphere VOIs tend to have slightly larger volumes than the left hemisphere VOIs. This has been true in two separate datasets (n=24 and n=25), neither of which has an obvious physiological trait that would lead us to expect these sorts of hemisphere differences.
This is my first time posting a question on this list, so I’m not sure if images are OK. If yes, there are a couple posted at the bottom of this message that should visualize what I’m talking about. If not, I can say that the volume differences are most pronounced in the largest VOIs (PuM, VPL, VLp, etc); in these regions, the right side is ~200-300 cmm larger than the left side. Additionally, the average volumes on both sides are reasonable based on Figure 8 of Iglesias et al (in the sense that our average volumes are well within the violin plots in that figure).
I’m wondering if anyone has encountered this trend before, and/or if there’s a correction I might implement to equalize the volumes. We want to be understand—to the extent possible—the extent to which these differences are driven by biology vs image processing.
Thanks so much, Evan
Images follow:
[cid:image001.png@01D69736.EDCFE020]
[cid:image002.png@01D69736.EDCFE020]
[cid:image003.png@01D69736.EDCFE020]
External Email - Use Caution
Hi all,
First of all, thanks very much for notifying me that this has been asked before. Looking over the previous thread, it seems as though the conclusion was that, because the thalamic atlas was designed to be symmetric, the most likely explanation for the observed asymmetry in our data was the data itself. We have now applied this pipeline to two different clinical populations, one of which includes smokers (described here: https://pubmed.ncbi.nlm.nih.gov/26272810/) and one of which includes patients with epilepsy (unpublished data). I’ve discussed this issue with Jake Dubroff (first author on paper linked above), and he agrees that, at least in terms of underlying biology, there isn’t any obvious reason that smokers would have hemisphere differences in thalamic volume. Nevertheless, we’re going to run an ‘internal control’ of sorts and apply the pipeline to an additional set of ~40 healthy controls, just to see if these biases persist in a population that really shouldn’t have any hemisphere differences. Stay tuned on that.
In the meantime, looking forward to hearing if there are any other potential explanations for this trend. Thanks again for all your help!
Best, Evan
--- Evan Gallagher Neuroscience PhD Candidate Mach Labhttps://www.pennmedicine.org/departments-and-centers/department-of-radiology/radiology-research/labs-and-centers/molecular-imaging/radiopharmaceutical-chemistry-and-biology-lab | University of Pennsylvania Email: evangall@pennmedicine.upenn.edumailto:evangall@pennmedicine.upenn.edu Tel: 781-507-1353
From: freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edu On Behalf Of Iglesias Gonzalez, Juan E. Sent: Wednesday, September 30, 2020 2:35 PM To: Freesurfer support list freesurfer@nmr.mgh.harvard.edu Cc: Greve, Douglas N., Ph.D. DGREVE@mgh.harvard.edu; chrisvriend@gmail.com Subject: [External] Re: [Freesurfer] Right-side bias in thalamic segmentation pipeline?
Dear Evan, The atlas is designed to be symmetric. Is it possible that the asymmetry is in your data? Another FS user (Chris Vriend, CCed) reported a similar effect (you can find his massage in the archive). Doug: any further ideas? Cheers, /E
Juan Eugenio Iglesias Senior research fellow CMIC (UCL), MGH (HMS) and CSAIL (MIT) http://www.jeiglesias.com
From: <freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu> on behalf of "Gallagher, Evan (NGG)" <Evan.Gallagher2@pennmedicine.upenn.edumailto:Evan.Gallagher2@pennmedicine.upenn.edu> Reply-To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu> Date: Wednesday, September 30, 2020 at 14:18 To: "freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu> Subject: [Freesurfer] Right-side bias in thalamic segmentation pipeline?
External Email - Use Caution Hi all,
I’ve been working on a project that involves the thalamic segmentation pipeline published by Iglesias et al. in 2018 in NeuroImage (“segmentThalamicNueclei.sh”). While the pipeline itself works great, we have noticed that the right hemisphere VOIs tend to have slightly larger volumes than the left hemisphere VOIs. This has been true in two separate datasets (n=24 and n=25), neither of which has an obvious physiological trait that would lead us to expect these sorts of hemisphere differences.
This is my first time posting a question on this list, so I’m not sure if images are OK. If yes, there are a couple posted at the bottom of this message that should visualize what I’m talking about. If not, I can say that the volume differences are most pronounced in the largest VOIs (PuM, VPL, VLp, etc); in these regions, the right side is ~200-300 cmm larger than the left side. Additionally, the average volumes on both sides are reasonable based on Figure 8 of Iglesias et al (in the sense that our average volumes are well within the violin plots in that figure).
I’m wondering if anyone has encountered this trend before, and/or if there’s a correction I might implement to equalize the volumes. We want to be understand—to the extent possible—the extent to which these differences are driven by biology vs image processing.
Thanks so much, Evan
Images follow:
[cid:image001.png@01D698E2.5D29F530]
[cid:image002.png@01D698E2.5D29F530]
[cid:image003.png@01D698E2.5D523DB0]
No, my guess is that it is in the data, meaning that it is biological or created by some kind of bias field. You can always left-right reverse the data and run it through the pipeline to see if the asymmetry reverses, though this is still not conclusive.
On 9/30/2020 2:35 PM, Iglesias Gonzalez, Juan E. wrote:
Dear Evan,
The atlas is designed to be symmetric. Is it possible that the asymmetry is in your data? Another FS user (Chris Vriend, CCed) reported a similar effect (you can find his massage in the archive).
Doug: any further ideas?
Cheers,
/E
Juan Eugenio Iglesias
Senior research fellow
CMIC (UCL), MGH (HMS) and CSAIL (MIT)
*From: *freesurfer-bounces@nmr.mgh.harvard.edu on behalf of "Gallagher, Evan (NGG)" Evan.Gallagher2@pennmedicine.upenn.edu *Reply-To: *Freesurfer support list freesurfer@nmr.mgh.harvard.edu *Date: *Wednesday, September 30, 2020 at 14:18 *To: *"freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Subject: *[Freesurfer] Right-side bias in thalamic segmentation pipeline?
* External Email - Use Caution *
Hi all,
I’ve been working on a project that involves the thalamic segmentation pipeline published by Iglesias et al. in 2018 in NeuroImage (“segmentThalamicNueclei.sh”). While the pipeline itself works great, we have noticed that the right hemisphere VOIs tend to have slightly larger volumes than the left hemisphere VOIs. This has been true in two separate datasets (n=24 and n=25), neither of which has an obvious physiological trait that would lead us to expect these sorts of hemisphere differences.
This is my first time posting a question on this list, so I’m not sure if images are OK. If yes, there are a couple posted at the bottom of this message that should visualize what I’m talking about. If not, I can say that the volume differences are most pronounced in the largest VOIs (PuM, VPL, VLp, etc); in these regions, the right side is ~200-300 cmm larger than the left side. Additionally, the average volumes on both sides are reasonable based on Figure 8 of Iglesias et al (in the sense that our average volumes are well within the violin plots in that figure).
I’m wondering if anyone has encountered this trend before, and/or if there’s a correction I might implement to equalize the volumes. We want to be understand—to the extent possible—the extent to which these differences are driven by biology vs image processing.
Thanks so much,
Evan
Images follow:
freesurfer@nmr.mgh.harvard.edu
-
Douglas N. Greve -
Gallagher, Evan (NGG) -
Iglesias Gonzalez, Juan E.