NITRC NITRC Community Forum: open-discussion

RE: LLMs in Neuroimaging Workshop - Boston, Oct 16, 2026

Kanwal Shahbaz — Fri, 04 Sep 2026 17:28:34 GMT

Lookinf forward to it.

Longitudinal intensity calibration for T1 T2 ratio in MRTool

Hannah Hafner — Fri, 04 Sep 2026 17:28:34 GMT

Dear NITRC community,
Our lab is working with the MRTool toolbox in SPM12 to calculate the T1/T2 ratio for a neurodevelopmental sample. MRTool uses linear scaling and nonlinear scaling with histogram matching to calibrate the T1 and T2 images before taking the ratio. Since we are using longitudinal data, we want to use a common intensity calibration for each subject instead of calibrating each subject's individual timepoint separately. Is there a way we could achieve this? For example, we have used SPM to create an average T1 image for each subject, and were wondering if it would be possible to calibrate each individual session T1 and T2 using the average image in MRTool.
Any help would be greatly appreciated,
Hannah

RE: LLMs in Neuroimaging Workshop - Boston, Oct 16, 2026

Eduardo Gonzalez-Toledo — Fri, 04 Sep 2026 12:29:25 GMT

It would be great if this Course can be attended online!!

LLMs in Neuroimaging Workshop - Boston, Oct 16, 2026

David Kennedy — Thu, 03 Sep 2026 12:55:26 GMT

If you are in Boston in October, please note the following: ReproNim is pleased to announce an upcoming workshop: Neuroimaging, Reproducibility and LLMs: Early promises and future questions Date: October 16th 9:30am - 1:00pm EDT (In person only) Location: MIT - Singleton Auditorium, Building 46 (3rd floor, 3002), 43 Vassar St, Cambridge, MA 02139 Registration: Required but free and open to all! Register at: <a href="https://www.eventbrite.com/e/neuroimaging-reproducibility-and-llms-early-promises-and-future-question-tickets-1999618019324?aff=oddtdtcreator&keep_tld=true" target="_new">https://www.eventbrite.com/e/neuroimagin...</a> The objective of this in-person workshop is to introduce the technology and train neuroscience / neuroimaging researchers on the potential of Large Language Models in brain imaging research. To achieve this, we will show and discuss key directions for which LLMs have - and will have - a fundamental impact, review and contrast research project examples, introduce latest techniques, and discuss potential limitations and risks associated with this transformative technology. Speakers will include: JB Poline (McGill), David Kennedy (UMassMed), Brent McPherson (McGill), Michelle Wang (McGill), Yaroslav Halchenko (Dartmouth), Johanna Bayer (McGill), and Satra Ghosh (MIT).

OASIS3 - individual scan acquisition year and country

Daniel Franco — Thu, 09 Jul 2026 13:21:00 GMT

Hi all,
Is there a way to obtain the acquisition year for each scan (or at least the year of baseline per subject, so that I can derive a year of image acquisition)? Similarly, is country of residence available? I'm assuming all participants reside within the United States.
I'm currently needing this data to assign contextual features in my analysis.
Thanks in advance, Daniel

Subject: Fully HIPAA compliant Deployment for OASIS-3 Longitudinal Data Mapping

MohammedYusuf Modan — Mon, 06 Jul 2026 17:09:50 GMT

<div class="paragraph normal ng-star-inserted" data-start-index="702">Hi Saif,</div>
<div class="paragraph normal ng-star-inserted" data-start-index="708">We have the exact technical infrastructure required to instantly accelerate your OASIS-3 neuroimaging research and ADRC Clinical Data workflows.</div>
<div class="paragraph normal ng-star-inserted" data-start-index="852">To ensure your study can efficiently map subject IDs and calculate "days-from-entry" alignments to the corresponding MRI/PET sessions, we can deploy a dedicated, locked-down research "Spoke" tailored specifically for your team.</div>
<div class="paragraph normal ng-star-inserted" data-start-index="1079">Our Research Setup Engine will bypass the burden of building a local data environment so you can focus entirely on selecting the clinical assessments closest to each imaging session. Your deployment will include:</div>
<ul class="ng-star-inserted">
<li class="paragraph list-item normal ng-star-inserted" data-start-index="1291">Secure Python FastAPI Backend: A dedicated, containerized environment to safely and rapidly execute your longitudinal data parsing and mapping scripts.</li>
<li class="paragraph list-item normal ng-star-inserted" data-start-index="1442">The BigQuery "Evidence Lake": A highly secure, centralized data storage pool designed for sub-second data retrieval and structuring complex, multi-year datasets.</li>
<li class="paragraph list-item normal ng-star-inserted" data-start-index="1603">Zero-Trust Processing & Ephemeral Memory: Our architecture ensures that data processing sessions are instantly wiped from local memory once your data is securely routed to the Evidence Lake, maintaining absolute data privacy.</li>
</ul>
<div class="paragraph normal ng-star-inserted" data-start-index="1828">Additionally, should your neuroimaging research ever scale into active clinical application, please note that this same infrastructure natively supports our advanced AI clinical documentation and autonomous compliance monitoring modules, which can be activated on demand.</div>
<div class="paragraph normal ng-star-inserted" data-start-index="2099">Let us know if you would like us to provision this secure environment for your study, and we can immediately initiate the setup.</div>
<div class="paragraph normal ng-star-inserted" data-start-index="2099"> </div>
<div class="paragraph normal ng-star-inserted" data-start-index="2099">Best,</div>
<div class="paragraph normal ng-star-inserted" data-start-index="2099">— Manager, Quantum Security Syndicate-m Ascent Enterprise, LLC Secure Posture Portal: www.quantumaudit.health Direct Support: support@quantumaudit.health</div>

spmT on 3d render visibility

Arshad Zaman — Mon, 06 Jul 2026 17:09:50 GMT

Hi, 
Does anyone know if it is possible to improve spm activations to better be  visible on the 3d render view? while in A+C+S activations apprear fine, on the render brain image the surface activations are bearly visible. I know one can play around with the depth/Azimuth functions for repective cross sections, however those are only really useul if you can also see the surface activations clearly too. is there anyway to bring the sruface activations actually onto the sruface of the 3d rendered brain? I have played around with options, darkest, brightest, etc but no joy. I have seen a few online tutorials, however there too the surface activations are not clearly visible at all.
Any suggesitons will be much appreciated.
Bw,
Arshad

Clinical Toolbox for SPM - Troubleshooting

Kilian Frohlich — Thu, 02 Jul 2026 13:18:53 GMT

Dear Community,
I have been using the clinical toolbox for years. Now, it produced following error the first time:
clinical_mrnormseg ERROR: Dimension mismatch 
 
Anyone familiar with this? 
 
Yours!

RE: Question about clinical labels in OASIS-3

saif_khan — Mon, 29 Jun 2026 15:20:01 GMT

I will download the ADRC Clinical Data spreadsheet and use the subject ID together with the days-from-entry information to match the clinical assessments to the corresponding MRI/PET sessions. Since OASIS-3 is longitudinal, I will select the clinical assessment closest to each imaging session when assigning labels.
Thank you also for pointing me to the Imaging Data Dictionary and the ADRC Clinical Data documentation. This information is very useful for my study.

RE: Question about clinical labels in OASIS-3

wissal ghachem — Mon, 29 Jun 2026 13:57:03 GMT

Hi,
the cognitive/diagnostic status isn't produced by a command-line script — it lives in the clinical data tables you download from the OASIS-3 data browser as spreadsheets (CSV). The download scripts only fetch the imaging/FreeSurfer/PUP files; they don't generate labels.
There are two places to look, both in the ADRC Clinical Data table:
<ol class="[li_&]:mb-0 [li_&]:mt-1 [li_&]:gap-1 [&:not(:last-child)_ul]:pb-1 [&:not(:last-child)_ol]:pb-1 list-decimal flex flex-col gap-1 pl-8 mb-3">
<li class="font-claude-response-body whitespace-normal break-words pl-2">CDR (Clinical Dementia Rating) global score — the quickest "normal vs. impaired" flag:
<ul class="[li_&]:mb-0 [li_&]:mt-1 [li_&]:gap-1 [&:not(:last-child)_ul]:pb-1 [&:not(:last-child)_ol]:pb-1 list-disc flex flex-col gap-1 pl-8 mb-3">
<li class="font-claude-response-body whitespace-normal break-words pl-2">CDR = 0 → cognitively normal</li>
<li class="font-claude-response-body whitespace-normal break-words pl-2">CDR = 0.5 → very mild impairment (commonly treated as MCI / at-risk)</li>
<li class="font-claude-response-body whitespace-normal break-words pl-2">CDR ≥ 1 → dementia (1 mild, 2 moderate, 3 severe)</li>
</ul>
</li>
<li class="font-claude-response-body whitespace-normal break-words pl-2">dx1–dx5 (coded clinician diagnosis) — more specific than CDR because it gives the etiology. <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">dx1</code> is the primary diagnosis, with values like "Cognitively normal", "AD dementia", "vascular dementia", plus contributing factors (mood disorder, vitamin deficiency, etc.). So "at risk of AD" specifically means a CDR > 0 with <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">dx1</code> pointing toward AD dementia / uncertain dementia, rather than just any impairment.</li>
</ol>
How to get it:
<ol class="[li_&]:mb-0 [li_&]:mt-1 [li_&]:gap-1 [&:not(:last-child)_ul]:pb-1 [&:not(:last-child)_ol]:pb-1 list-decimal flex flex-col gap-1 pl-8 mb-3">
<li class="font-claude-response-body whitespace-normal break-words pl-2">Open the OASIS-3 project in the data browser (central.xnat.org).</li>
<li class="font-claude-response-body whitespace-normal break-words pl-2">Load the Subjects and ADRC Clinical Data tables.</li>
<li class="font-claude-response-body whitespace-normal break-words pl-2">Export as a Spreadsheet (CSV). You'll get columns like <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">cdr</code>, <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">dx1</code>–<code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">dx5</code>, <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">mmse</code>, <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">ageAtEntry</code>, plus a session label.</li>
</ol>
One important detail: OASIS-3 is longitudinal, so each clinical row is tagged with the subject ID + "days from entry" (the <code class="bg-text-200/5 border border-0.5 border-border-300 text-danger-000 whitespace-pre-wrap rounded-[0.4rem] px-1 py-px text-[0.9rem]">dXXXX</code> suffix), and a subject's status can change over time. To label a given scan, match by subject ID and pick the clinical assessment closest in days-from-entry to that MR/PET session.
All of this is documented in the OASIS-3 Imaging Data Dictionary (the "ADRC Clinical Data / dx1–dx5" and CDR sections) on oasis-brains.org.
Hope this helps.