Thanks that's helpful, and yeah sounds simpler than a modularity algorithm.
04.02.2026 00:06
👍 0
🔁 0
💬 0
📌 0
Thanks that's helpful, and yeah sounds simpler than a modularity algorithm.
Important work! Curious how you all define “precision”? (without having combed the methods). Trying to build a mental model around minimum viable criteria for PNM.
Nice work! Especially reassuring to see null findings with some disorders.
When a brain researcher solved a logistical problem by going rogue, the idea proved remarkably infectious.
By @lyrebard.bsky.social
#neuroskyence
www.thetransmitter.org/brain-imagin...
One major benefit resulting from all the LNM studies, regardless of how the dust settles: a massive collection of lesions have been compiled. That can be a powerful reference set for future studies testing how clustered a new set of lesions are.
For existing LNM studies, preprocessing choices for the normative connectome play a big role in determining how the LNM maps look. Global signal regression is going to change the degree distribution, and we know GSR is not a no-brainer.
Maybe a set of lesions converge on how they perturb the low-dimensional functional state space (h/t D Jones). We talk about this in our recent atrophy-FC paper: www.nature.com/articles/s41.... I think structure-function methods like ours can help link disparate lesions to common cognitive outcomes.
The core question motivating LNM studies is: how do disparate lesions converge on a common syndrome? Put another way: what's the structure-function-cognition mapping? I think patient fMRI is crucial here!
To strengthen the "chance" case, the simulated null lesions should have the same spatial autocorrelation as the true lesions.
For future "LNM 2.0" studies, a good research question may simply be: does a set of lesions cluster on some spatial feature - connectivity, gradient, gene expression - more than expected by chance?
If they did, you'd get different LNM connectivity maps back for different syndromes. Instead, the lesions are actually uniformly distributed across the connectome, which results in you getting the functional connectome degree map back as the LNM map.
The LNM method in a nutshell is: do the lesions for a syndrome cluster on some spatial feature of the healthy functional connectome? The takeaway from this study: no, the lesions don't cluster.
This is strong and careful work. I like how they boiled LNM down to its essence: LNM = sum(M x C). They clearly thought deeply about the method.
Been pondering the lesion network mapping study all day. Here's my $0.02 [🧵]:
More than 200 published studies and at least seven ongoing clinical trials rely on potentially faulty brain network maps, according to a study published yesterday.
By @avaskham.bsky.social
#neuroskyence
www.thetransmitter.org/brain-imagin...
I am recruiting a Postdoc to join my lab at UMN. If you or someone you know is interested in studying individual differences in brain and cognitive aging, check out the listing and my website in my bio and apply!
I appreciate RTs to help get the word out as well :)
Paradigm shift away from regionalization, finally?
Where to next?
- Deploy this biomarker as a real world test (radiata.ai)
- Develop non-invasive neurostimulation therapy for functional connectivity imbalances
- Apply eigenmode analysis to develop new fMRI biomarkers
- @lollopasquini.bsky.social's curiosity led to us looking into gradients, which led to the idea of gradient imbalance and hypo/hyper connectivity in dementia. Just followed the thread.
- Eigenmode analysis finally made sense after reading Strogatz’s “Nonlinear dynamics and chaos” and taking a long hike to the beach at Point Reyes.
Things that happened along the way:
- Back in 2013, Helen Zhou’s Brain paper about convergent and divergent functional connectivity in AD/FTD got lodged in my mind and never left.
- The idea for structure-function mapping in AD and FTD came at OHBM 2019 in Rome. Idea to paper took a long time.
Sincere thanks to the participants, outstanding colleagues at the @ucsfmac.bsky.social, and to the Tau Consortium for support. 🙏
Key finding 5: Sensory-association imbalance is a promising cognitive biomarker for prognosis/monitoring because 1) higher imbalance at baseline predicts accelerated cognitive decline and 2) functional biomarkers will likely show more dynamic response to treatment.
Key finding 4: Structure and function biomarker scores both contribute to cognitive impairment.
Key finding 3: Eigenmode analysis reveals reductions in gradient amplitude and phase, which we call collapse. Those disruptions that add up to observed FC differences.
Key finding 2: Hypo and hyperconnectivity appear as two sides of the same coin. Different atrophy patterns perturb specific functional gradients, in which anticorrelated network pairs are embedded.
Key finding 1: Sensory-association functional connectivity imbalance (SAI) appears in all syndromes. As atrophy increases, sensory connectivity weakens and association connectivity gets stronger. Did not expect this.
We mapped structure-function relationships in Alzheimer’s disease and frontotemporal dementia. This was a good cohort because the atrophy patterns collectively cover almost the entire brain.
Excited to share our new work in @natcomms.nature.com: Functional network collapse in neurodegenerative disease
How does functional connectivity change across dementia types and stages?
www.nature.com/articles/s41...
❤️ Bluesky fMRI people! 3-day #fMRI course live online Jan 7-9, 2026.
#SPM, #ICA, GLM, connectivity, mediation, MRI physics, #DataScience with @vcalhoun.bsky.social and Kent Kiehl.
We love talking methods & connecting with colleagues! Come join us!
Register here:
sites.google.com/dartmouth.ed...