Sign in

Logan Dowdle

@dowdlelt.bsky.social
143 followers 279 following 16 posts

Assistant Professor in Cognitive Neuroscience @ Maastricht University. Magnets on my mind. Previously postdoc at CMRR, Minnesota

PostsRepliesMedia
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
It was really just an experiment based around "Can this tool be faster?", "Can topup play nicely with 3dNwarpapply?", "If my eyes see the brain move, then why didn't moco!?" So - if you have data you are struggling with, alignment, motion, etc - I'm interested.
010
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
I view this as a look at what is possible - and extend truly heartfelt gratitude to the developers of AFNI, ANTS, SPM, FSL, NORDIC, GLMdenoise/GLMsingle, medic/warpkit and all others. This grew out of frustrations in working with large data and trying to glue tools together.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
I've tested on Linux CPU (16 threads/8 cores), Mac M3 - but primarily on a 5070Ti 16Gb. GPU benefits are CUDA only right now. The code tries to adapt to the GPU you provide (so even 8Gb vram should work?) but does expect enough normal ram to hold what you are throwing at it.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
And so much more. I've tested it, I've tried to validate each stage, and its very much in-progress, but it works right now on what I've thrown it at (3mm to 0.35, 1 session to 10, 3T to 7T, one task to 21). I could say a lot more - but I think its time someone else breaks it!
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
ffs_nordic reproduces NORDIC at 2x the speed - and if you save out the number of components removed, you can update your statistical maps to use z-stats, adjusted for the per-voxel DoF loss with ffs_util_updatedof.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
ffs_autoproc (inspired by afni_proc.py) can take a BIDS + FreeSurfer & SUMA folder and create a full bash script to process the data. And, just to be clear, this all works on the CPU as well - pytorch will adapt. Choose is you want simple affine - or nonlinear xrun/xfmap/xses!
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
I was also thrilled to see warpkit (MEDIC) out there in the wild - FFS supports 5D warp files and uses the GPU to apply them - so its very very speedy, even on huge datasets (tested with whole brain 1.25mm multiecho BOLD data). I wrap warpkit for this, for ease.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
If you like MELODIC ICA - ffs_ica reproduces it roughly 10x faster. And I've always liked topup - ffs_blipflip matches it in 30s instead of 15 minutes. Each of these I've validated against the reference code.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
3dREML-style ARMA(1,1) modeling 5x faster ffs_reml) with scaling and smoothing as part of the load (saving space). Want GLMsingle type approach? then look to ffs_hrfopt/ffs_denoise/ffs_ridge. And use events.tsv files directly - no need for TR-locking or identical durations.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
AFNI style motion correction (ffs_moco), affine alignment (MUCH faster), nonlinear transforms in 3 flavors (ANTS SyN, AFNI qwarp, Optic Flow based), nonlinear motion correction in 2 minutes (!) instead of 4 hours (ffs_locomoco), slice timing (ffs_slicetime), etc etc
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
So, the idea is a full workflow - BIDS nii data to a final timeseries with a single interpolation step. That is - full alignment with slice time correction all in one giant interpolation with ffs_nwarp - a drop in, fast, replacement and extension of 3dNwarpApply.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
It could be that the subject got out of the scanner, or that I have multiple field maps. The shim changed - life happens. I want to be able to iterate quickly, solve the problem and then...all those transforms must be applied in a single step to reduce blurring.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
I just really want to take data - a lot of data, and make sure a voxel is a voxel is a voxel. So that might mean nonlinear motion correction. It likely means across session (or runs), nonlinear warps. Maybe nonlinear to anat.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
It revolves around a few ideas - the first is a GPU focus, because when you spend all your time doing linear algebra and interpolation this starts to make sense. The next also speed related - uses zstd natively, instead of gzip - thats another 2x speed increase, easy.
100
Logan Dowdle @dowdlelt.bsky.social · 03/08/2026
I want to talk about something I have been working on, and want to be upfront that its agent assisted coding - basically, an opinionated toolbox for intensive ("hardcore") or layer fMRI projects - ones that demand precision: fastfuncstuff here: github.com/dowdlelt/fas...
github.com
GitHub - dowdlelt/fastfuncstuff: A repository of GPU accelerated fMRI tools based on AFNI, FSL and others along with fMRI simulation code. A bit a mess, but we are having fun.
A repository of GPU accelerated fMRI tools based on AFNI, FSL and others along with fMRI simulation code. A bit a mess, but we are having fun. - dowdlelt/fastfuncstuff
110
Reposted by Logan Dowdle
Blaise Frederick @blaisefrederick.bsky.social · 02/07/2026
Dr. Tianye Zhai’s careful and exhaustive new paper evaluating the effects of numerous fMRI denoising pipeline components, alone and in combination, just came out in HBM - check it out! onlinelibrary.wiley.com/doi/10.1002/...
onlinelibrary.wiley.com
Benchmarking fMRI Denoising Pipelines
This paper benchmarked fMRI denoising pipelines on task- and resting-state, single- and multi-echo fMRI datasets, in a modern 1-step regression framework, with pre-whitening and time-shifted physiolo...
1118
Logan Dowdle @dowdlelt.bsky.social · 11/06/2026
Looking forward to this!
010
Reposted by Logan Dowdle
Dan Handwerker @danielhandwerker.bsky.social · 11/06/2026
About to start my travel to #OHBM2026 Excited to help organize a full day ed course on multi-echo fMRI. Schedule is here: me-ica.github.io/ohbm-2026-mu... @enekourunuela.com @drjuliamoser.bsky.social @dowdlelt.bsky.social @ohbmofficial.bsky.social
me-ica.github.io
192
Reposted by Logan Dowdle
Faruk Gulban @ofgulban.bsky.social · 30/03/2026
If you are setting up the meso-veins protocol for the first time at your 7T (doi.org/10.1126/scia...), you should watch this video where I am providing the most up to date pointers. As well as the tips on quality controlling your images: youtu.be/57URvSjUYRY
033
Reposted by Logan Dowdle
Dan Handwerker @danielhandwerker.bsky.social · 30/03/2026
How do we define "good" fMRI data? Especially with resting state, there are circularity risks if we evaluate data quality as showing the networks we expect to see. Javier Gonzalez-Castillo (& me & others) developed pBOLD, a new metric that uses multi-echo info. www.biorxiv.org/content/10.6... 1/8
biorxiv.org
13419
Reposted by Logan Dowdle
layerfmri.bsky.social @layerfmri.bsky.social · 29/12/2025
New layer-fMRI manuscript finding laminar response differences across orientation-tuned surround suppression. doi.org/10.64898/202... By Emerson et al
091
Reposted by Logan Dowdle
Ravi Menon @northernthrux.bsky.social · 13/11/2025
14 months after submission, our article “Stimulus-modulated approach to steady state (SASS): a flexible paradigm for event-related fMRI" is now out in @natmethods.nature.com . You can read it here rdcu.be/ePJo6 It is the first first author paper from my student @renilmathew.bsky.social 👏🏽👏🏽👏🏽 …1/N
rdcu.be
Stimulus-modulated approach to steady state (SASS): a flexible paradigm for event-related fMRI
Nature Methods - Stimulus-modulated approach to steady state (SASS) is an acquisition scheme for event-related fMRI that generates data with high temporal signal-to-noise ratios interspaced with...
45820
Reposted by Logan Dowdle
Paul Taylor @afni-pt.bsky.social · 17/11/2025
New AFNI Academy playlist! This tutorial presents afni_proc.py's quality control HTML for single subject FMRI. The APQC HTML has systematic views of data and useful derived quantities. Users can instantly rate, comment and query the fully processed subject data. www.youtube.com/watch?v=hD9z...
0126
Reposted by Logan Dowdle
Gang Chen @gangchen6.bsky.social · 20/09/2025
Data only shows associations. Turning those into claims about mechanism or causation? That requires a Rosetta Stone of prior knowledge + theory. Resting-state fMRI is purely observational; correlation is its currency. From this, plenty of "theoretical toys" about brain function can be built...
12712
Reposted by Logan Dowdle
Dan Handwerker @danielhandwerker.bsky.social · 06/08/2025
Nice work by @yoichimiyawaki.bsky.social showing the fMRI hemodynamic response contains enough info to decode stimulus type within 2sec. Happy to have helped with this project. www.biorxiv.org/content/10.1...
biorxiv.org
Rapid decoding of neural information representation from ultra-fast functional magnetic resonance imaging signals
High spatio-temporal resolution is crucial for neuroimaging techniques to improve our understanding of human brain function. While the fMRI signal is slow and shows a spread in latencies over space, t...
0276
Reposted by Logan Dowdle
Mark Lescroart @neuromdl.bsky.social · 12/06/2025
🚨Paper alert!🚨 TL;DR first: We used a pre-trained deep neural network to model fMRI data and to generate images predicted to elicit a large response for each many different parts of the brain. We aggregate these into an awesome interactive brain viewer: piecesofmind.psyc.unr.edu/activation_m...
piecesofmind.psyc.unr.edu
Cortex Feature Visualization
2116
Reposted by Logan Dowdle
Matthias Nau @matthiasnau.bsky.social · 02/12/2024
#Neuroimaging crowd, hear me out! 👀 Did you ever want to add #EyeTracking to your #fMRI study but found it too much hassle? Got existing data you’d love to add eye tracking to? Consider trying out *MR-based eye tracking* (i.e. inferring gaze from eye voxels)! Here is a 🧵 with a few options!👇 1/9
827295
Reposted by Logan Dowdle
layerfmri.bsky.social @layerfmri.bsky.social · 27/11/2024
A starter pack of layer-fMRI folks here: go.bsky.app/TUWFNzg
12110
Reposted by Logan Dowdle
Eneko Uruñuela @enekourunuela.com · 27/11/2024
For all those researchers wanting to do hemodynamic deconvolution with my #neuroimaging pySPFM package: Just updated the documentation to include two clear examples of how to call the pySPFM package and what outputs you can expect. Should help you get the most out of this tool.
261
Reposted by Logan Dowdle
Faruk Gulban @ofgulban.bsky.social · 14/11/2024
While fine-tuning 3D EPI parameters, we captured an impressively clear view of the human LGN! If you look closely, you can almost make out faint layers—wish they were more distinct though 🫤
072
Reposted by Logan Dowdle
layerfmri.bsky.social @layerfmri.bsky.social · 25/11/2024
New layer-fMRI review on the methodological challenges and prospects of laminar fMRI. By Yuhui Chai et al., doi.org/10.1093/psyr...
0103
Reposted by Logan Dowdle
Jaan Aru @jaanaru.bsky.social · 20/11/2024
The craziest paper I have ever done is this thought experiment with Albert Gidon and Matt Larkum. In the first journal, reviewer 1 recommended that we should not try to publish this; reviewer 2 called it "wacky". Thanks for the motivation: A sequel is coming up! journals.plos.org/plosbiology/...
journals.plos.org
Does brain activity cause consciousness? A thought experiment
The authors of this Essay examine whether action potentials cause consciousness in a three-step thought experiment that assumes technology is advanced enough to fully manipulate our brains.
911925
Reposted by Logan Dowdle
Micah G. Allen @micahgallen.com · 19/11/2024
"A neural basis for distinguishing imagination from reality" This new preprint from @nadinedijkstra.bsky.social @peterkok.bsky.social & @smfleming.bsky.social is simply *stunning*. Beautifully elegant design, cognitive modelling, and brain imaging. Total package. osf.io/preprints/ps... 🧠📈 🧪
617148
Reposted by Logan Dowdle
Eneko Uruñuela @enekourunuela.com · 13/09/2023
Friendly reminder for everyone who has just joined #neuroskyence : There is a #neuroimaging feed you can follow too! And publishing there is as easy as using the hashtag. Just like it and pin it 👇 bsky.app/profile/did:...
047