Vince Calhoun @vcalhoun.bsky.social

Multimodal brain imaging & genetics, signal & image processing, general all around geek, husband, father of three, lover of purple. @[email protected]

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@KordingLab @ylecun IMO engineers can be especially powerful in research. Deeply theoretical (cf. signal/image processing, control, and information theory) while bringing a bias toward pragmatic experiments that fail quickly, quantify precision/output, and narrow the solution space.

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@bttyeo @tianchuzeng @kkli20111 @ZShaoshi @ten_photos Suggestion for slightly more precise (but still promotional) title: “Statistical tests often ignore cross-validation dependence in biomedical ML.” or, if you want to go further out on the ledge: "“Almost Everything You Thought Was Significant in Biomedical ML May Be Wrong". ;-)

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@bttyeo @tianchuzeng @kkli20111 @ZShaoshi @ten_photos in case helpful, we harp a lot on various types of bias/circularity (can sometimes be tricky to avoid) in this paper : https://t.co/lW2Ch6s8yL

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

nice to meet up with @PTenigma & Jess Turner at SOBP this year!

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

The next edition of the IEEE Machine Learning for Signal Processing (MLSP) Workshop will take place in Atlanta, USA, from September 28 to October 1, 2026! 🤩

Take a look at the website for further information and deadlines: https://t.co/iYOnuOYo57

See you soon Atlanta! 🚀

[PET Connectivity @ConnectivityPet]

The Molecular Connectivity Working Group (MCWG) invites you to our upcoming Symposium “Molecular connectivity: Best practices for data analysis” in Bordeaux (France) on June 19th, 8:30 am - 1:00 pm CEST, in the same city and right after the #OHBM2026 Annual Meeting!

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@micahgallen @LeahBanellis No matter how many reviews, the Thalarnus finds a way. ;-)

[Youngjo Song @ys1j13]

Moreover, using only phase information from thirteen identified flows, a linear classifier decoded 23 tasks at 89.9% accuracy, outperforming network activation amplitude. This suggests the brain's key control variable is when dynamics align, not where or how much it activates.

[Youngjo Song @ys1j13]

This addresses the stability–flexibility problem: how does flexible task activation arise from stable FC architecture? Assuming stable flows, our framework reconstructs 23 task activation maps (mean R = 0.73) while preserving FC topology across all task states.

[Youngjo Song @ys1j13]

Flows are identified by modeling coherence among intrinsic networks, referred to as intrinsic network flows (INFs). Remarkably, the flow sequence of the dominant flow (INF 1) — the one contributing most to dynamics — closely resembles the joint embedding of FC gradients 1 and 2.

[Youngjo Song @ys1j13]

Happy to share my first work as a postdoc!🎉 We show diverse cognition can emerge by retiming stable intrinsic flows. Such flows reconstruct FC, whole-brain gradients, and even 23 task activations via phase alone, unifying resting & task dynamics. #neuroscience #CenterTrends

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@PTenigma Don’t worry, we can repair it at a distance via entanglement, though observing whether it worked will break it again.

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@KordingLab @TroyASmithPhD Fair, though that reads pretty tame compared to a lot of abstracts. The paper itself is much more qualified.

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@KordingLab @TroyASmithPhD That would be a very different paper. ;) I read this more as modest effects + lots of caveats, not ‘street layout explains everything'.

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@KordingLab @TroyASmithPhD Like saying ‘the layout of a gym can’t affect fitness’…until you realize it changes how often and how hard people use it. I haven't read in detail, but given how strongly navigation demands are linked to hippocampus, it doesn’t seem that far-fetched.

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@KordingLab @TroyASmithPhD Right, like saying diet can’t affect weight because metabolism does a million other things. ;-)

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@KordingLab @TroyASmithPhD I’d frame it less as ‘obvious confounds’ and more as ‘important residual confounds.’ They did include area-level SES and several other adjustments. The tougher issue is selection and behavior, which is harder to nail down.

[Vince Calhoun @vcalhoun.bsky.social @vdcalhoun]

@AndrewZalesky @shansiddiqi Totally agree with open sharing, but what's allowed depends on the IRB, some take the view that *anything* at the subject level comes with risk of re-identification. But even if not allowed, one can consider a federated analysis vault (https://t.co/KSrAnvZCXY)