Benedikt Sundermann

Poor image quality introduces systematic bias into MRI data Analyses that include low-quality MRI data underestimate cortical thickness and overestimate surface area, according to new work from the ABCD Study.

Big data won't solve anything if it's crappy data. 😒

We need to 1) train people to inspect datasets before using them and 2) ensure that openly shared data is good data.

www.thetransmitter.org/spectrum/poo...

My dissertation / first RR is out now in Imaging Neuroscience! Lots of folks discuss how between-person heterogeneity may limit the utility of group-averages for brain functioning - we empirically tested that idea! doi.org/10.1162/IMAG...

Using Extreme Value Statistics to Reconceptualize Psychopathology as Extreme Deviations From a Normative Reference Model Overview of our approach for extreme value statistics. First, we fit a normative model to imaging phenotypes, before employing a ‘peaks-over-threshold’ approach widely used in meteorology and finance....

👀

Using Extreme Value Statistics to Reconceptualize Psychopathology as Extreme Deviations From a Normative Reference Model
onlinelibrary.wiley.com/doi/full/10....

1000s of studies have been published mapping neuroanatomical changes in psychiatric disorders. Can we ever converge on robust disease phenotypes? Our latest PP led by Trang Cao in a huge collab effort investigates this question...

@nsb-lab.bsky.social

Breath hold fMRI is great for measuring CVR... *except when your participant or patient doesn't quite perform the task the way you'd like. Let's see if we can fix that! (Spoiler, it looks like we can, using the respiratory belt data to predict or fill in missing PETCO2 data!)

fMRI can do more than you think Advances in brain imaging technology provide new and different information about the brain.

ICYMI: Recent technological advances in functional MRI provide a range of new information about brain physiology. Collaboration between engineers and neuroscientists will help maximize these gains, writes Laura Lewis.

www.thetransmitter.org/human-neurot...

Benchmarking methods for mapping functional connectivity in the brain - Nature Methods In this Analysis, Liu et al. benchmark more than 200 pairwise statistics for functional brain connectivity in tasks such as hub mapping, distance relationships, structure–function coupling and behavio...

paper: doi.org/10.1038/s415...
code: github.com/netneurolab/...

@zhenqi.bsky.social will present this work at the @ohbmofficial.bsky.social symposium: "Validating brain connectivity measures" on Friday Jun 27 3.45PM (poster 1386)

3/3 See zenodo.org/records/1570... for detailed results figures for different atlases. We hope that despite all limitations of this approach this work shows how much information about functional networks is hidden in the regions we score every day in #neuroradiology #neuroimaging #CT

A network correspondence toolbox for quantitative evaluation of novel neuroimaging results - Nature Communications Here, the authors present the Network Correspondence Toolbox, which enables researchers to examine and report spatial correspondence between their neuroimaging results and widely used brain atlases.

2/3 This work builts the exceptional network correspondence toolbox doi.org/10.1038/s414... by Ru Kong and ASPECTS masks in MNI space from doi.org/10.1371/jour...

1/3 Ever wondered how the individual cortical #ASPECTS regions routinely assessed in acute #stroke imaging are related to functional #brain #networks? We have now addressed this with an atlas correspondence analysis. See our brief report preprint at doi.org/10.1101/2025...

Mapping cortical Alberta Stroke Program Early CT Score (ASPECTS) regions to functional brain networks https://www.medrxiv.org/content/10.1101/2025.06.20.25330017v1

Generative AI and Vision Language Models in Neuroimaging: Some Advances and Open Questions YouTube video by Paul Thompson

Sorry my math/stats workshop talk was a bit of a disaster as the videos didn't play :) but here it is anyway
www.youtube.com/watch?v=nCfa...

GitHub - BRAIN-TO/bto_mri_protocols_info: This is a repository of wiki, code and containers for post-processing data acquired using the BRAIN-TO imaging protocols on a Siemens Prisma 3 T This is a repository of wiki, code and containers for post-processing data acquired using the BRAIN-TO imaging protocols on a Siemens Prisma 3 T - BRAIN-TO/bto_mri_protocols_info

On 🧠 #WorldBrainDay, let me highlight “Advancing Clinical and Neuroscientific Research Through Accessible and Optimized Protocol Design at 3T” by @srikash.bsky.social and Kâmil Uludağ, PhD (@uhn.ca, @uoft.bsky.social)

🔗 github.com/BRAIN-TO/bto...

#MRI #NeuroSky #OpenScience @mritobi.bsky.social

Keynote series Dr. Lucina Uddin - Mapping the Brain: Why Standardization Matters — OHBM Communications The Network Correspondence Toolbox (NCT) helps neuroscientists compare brain imaging data across different atlases, addressing inconsistencies in how brain networks are labeled. Developed by Dr. Lucin...

Thanks for the great write up @ohbm-com.com! More to come #OHBM2025 @ohbmofficial.bsky.social 🧠 loom.ly/rNZb504

Coming June 2025! We are excited to launch the NIH Brain Development Cohorts (NBDC) Data Hub. This new data ecosystem will host & facilitate access to data from the Adolescent Brain Cognitive Development (ABCD) Study & the HEALthy Brain and Child Development (HBCD) Study.

1/13

Ten principles for reliable, efficient, and adaptable coding in psychology and cognitive neuroscience - Communications Psychology Programming is essential for modern research in neuroscience and psychology, but it can quickly become a source of frustration and error. This Primer introduces ten practical principles guiding resear...

Many processes from industry are not appropriate or overkill for an academic setting, but we can and should be doing better than we currently are.

Our paper tries to give practical advice on how you can improve your scientific workflow, without being a coding guru.

www.nature.com/articles/s44...

My editorial on how journals can earn trust.

We often use journal names as proxies for quality. This is bad bc it’s not valid. But it could be. Editors could make journal name a valid signal. And we could place value on journals that show us how they do that.

journals.sagepub.com/doi/full/10....

A network correspondence toolbox for quantitative evaluation of novel neuroimaging results - Nature Communications Here, the authors present the Network Correspondence Toolbox, which enables researchers to examine and report spatial correspondence between their neuroimaging results and widely used brain atlases.

It's finally here! Use the Network Correspondence Toolbox to help contextualize your neuroimaging findings 🧠

Opaque thresholding hides information: context around statistically significant clusters is removed, treats subthreshold regions as having no activity. Transparent thresholding shows information: context around significant clusters is displayed, can observe relative importance of subthreshold regions.

Transparency matters when reporting results in neuroimaging! arxiv.org/abs/2504.07824 This could be a landmark in setting publishing standards for interpretability/reproducibility. Stellar work by @afni-pt.bsky.social @fmri-today.bsky.social @brightmg.bsky.social @gangchen6.bsky.social et al.

The mind craves binaries: good or bad, true or false, on or off. It’s tidy. It’s comforting. But the world rarely plays along. Reality tends to unfold in gradients, not in absolutes. And so does statistical evidence. Data analysis doesn’t speak in black and white, but in shades of uncertainty.

We tell our trainees to really LOOK at their data, every step of the way. This paper takes that wisdom even further, so future external readers really see our neuroimaging results in all their complex glory 👏 Thanks @afni-pt.bsky.social for spearheading this amazing team effort!

[AFNI Academy] Go Figure: Transparency in images preserves context and clarifies interpretation YouTube video by AFNI Bootcamp

A brief video about the topic is now available here:
www.youtube.com/watch?v=YpWl...
It covers a few (but not all!) of the examples from the "Go Figure" draft.

The result of a large (42 authors!) collaboration:
"Go Figure: Transparency in neuroscience images preserves context and clarifies interpretation"
arxiv.org/abs/2504.07824
TL;DR: The FMRI world can (and should) improve results interpretation and reproducibility *today*, via transparent thresholding.

New neurointerventional preprint from our Department. #stroke #evt #thrombectomy

Some examples of what VoxelPrompt can do!

VoxelPrompt: A Vision-Language Agent for Grounded Medical Image Analysis We present VoxelPrompt, an agent-driven vision-language framework that tackles diverse radiological tasks through joint modeling of natural language, image volumes, and analytical metrics. VoxelPrompt...

Recently thought about a model which takes a delineation instruction / segmentation protocol as input and provides the respective segmentation(s). Today I read VoxelPrompt. Super interesting work by Andrew Hoopes, Victor Ion Butoi, John Guttag and @adriandalca.bsky.social.

Is anyone aware of a well-curated list of organisations and foundations worldwide which contribute to research infrastructure by providing databases, repositories, standards etc. who can directly receive donations via their websites? Thinking of @cos.io as an example…. Please continue!

Why is the "T1 DARK RIM" a big deal in #MultipleSclerosis #MRI?

🧵1/6

Students’ perspective on new teaching concepts for medical studies: case- and competency-based learning in radiology - Insights into Imaging Objective This study aimed to evaluate medical students’ perception of a new radiology teaching format for abdominal diagnostics. The format transitioned traditional lectures and seminars to a case- a...

insightsimaging.springeropen.com/articles/10....

Thanks to everyone's feedback, we have updated our calculator to optimize sample size N & scan time T for fMRI studies: leonoqr.github.io/ORSP_Calcula...

The first new feature is that users can explore how different N & T leads to different accuracy, e.g., N=1000 & T=30min => 81% max accuracy. 🧵