Harry H Behjat
@aitchbi.bsky.social
imaging neuroscience • data science on graphs • brain structure-function-pathology interplay • spatial methods • MRI PET • Alzheimer’s disease
🌐 https://linktr.ee/aitchbi
🌐 https://linktr.ee/aitchbi
[🧵9] fourth, we asked, does FC also have prognostic value, that is, does individualised FC explain individual follow-up tau-PET patterns? it does.
October 7, 2025 at 8:59 AM
[🧵9] fourth, we asked, does FC also have prognostic value, that is, does individualised FC explain individual follow-up tau-PET patterns? it does.
[🧵8] third, we asked, is the superior explanatory power of FC over canonical PET patterns specific to tau and
absent for e.g. B-amyloid pathology? it is.
absent for e.g. B-amyloid pathology? it is.
October 7, 2025 at 8:59 AM
[🧵8] third, we asked, is the superior explanatory power of FC over canonical PET patterns specific to tau and
absent for e.g. B-amyloid pathology? it is.
absent for e.g. B-amyloid pathology? it is.
[🧵7] second, as a largely overlooked dimension, we asked, does FC actually perform better in explaining individual tau-PET patterns than just using canonical patterns of tau-PET? it does.
October 7, 2025 at 8:59 AM
[🧵7] second, as a largely overlooked dimension, we asked, does FC actually perform better in explaining individual tau-PET patterns than just using canonical patterns of tau-PET? it does.
[🧵6] first, we asked, does patient-specific FC help better to better explain the variance in individual tau-PET patterns than group-level FC does? it does.
October 7, 2025 at 8:59 AM
[🧵6] first, we asked, does patient-specific FC help better to better explain the variance in individual tau-PET patterns than group-level FC does? it does.
[🧵5] in a large sample of deeply phenotyped patients across the AD continuum from @biofinder.bsky.social, we rigorously validated, at the individual level, the association between patterns of tau-PET and functional connectivity, on four fronts.
October 7, 2025 at 8:59 AM
[🧵5] in a large sample of deeply phenotyped patients across the AD continuum from @biofinder.bsky.social, we rigorously validated, at the individual level, the association between patterns of tau-PET and functional connectivity, on four fronts.
does brain connectivity drive spread of pathological proteins in Alzheimer’s disease?
✨ preprint: www.biorxiv.org/content/10.1...
if the question intrigues you, please read on 🧵⤵️
✨ preprint: www.biorxiv.org/content/10.1...
if the question intrigues you, please read on 🧵⤵️
October 7, 2025 at 8:59 AM
does brain connectivity drive spread of pathological proteins in Alzheimer’s disease?
✨ preprint: www.biorxiv.org/content/10.1...
if the question intrigues you, please read on 🧵⤵️
✨ preprint: www.biorxiv.org/content/10.1...
if the question intrigues you, please read on 🧵⤵️
In Alzheimer's disease, spatial patterns of tau pathology manifest notable heterogeneity, one form of which is hemispheric asymmetry. But what drives this asymmetry?
@teanijarv.bsky.social investigates 🔎
@teanijarv.bsky.social investigates 🔎
April 22, 2025 at 7:08 PM
In Alzheimer's disease, spatial patterns of tau pathology manifest notable heterogeneity, one form of which is hemispheric asymmetry. But what drives this asymmetry?
@teanijarv.bsky.social investigates 🔎
@teanijarv.bsky.social investigates 🔎
Big shout out to Filip & Bratislav + co-authors for this very useful & important work!
An example (@biofinder.bsky.social data) showing subjects' FC "strength" can to some degree explain pathological tau protein patterns better than group-FC but not their mere "degree".
An excellent null model.
An example (@biofinder.bsky.social data) showing subjects' FC "strength" can to some degree explain pathological tau protein patterns better than group-FC but not their mere "degree".
An excellent null model.
February 20, 2025 at 8:41 AM
Big shout out to Filip & Bratislav + co-authors for this very useful & important work!
An example (@biofinder.bsky.social data) showing subjects' FC "strength" can to some degree explain pathological tau protein patterns better than group-FC but not their mere "degree".
An excellent null model.
An example (@biofinder.bsky.social data) showing subjects' FC "strength" can to some degree explain pathological tau protein patterns better than group-FC but not their mere "degree".
An excellent null model.
For data defined on top of a graph/network (e.g. annotations on brain graphs), it is efficient to decompose them with filters that adapt to the spectral characteristics of the data at hand | w. @dimitrivdv.bsky.social @rikossenkoppele.bsky.social + C-F Westin
February 19, 2025 at 11:16 AM
For data defined on top of a graph/network (e.g. annotations on brain graphs), it is efficient to decompose them with filters that adapt to the spectral characteristics of the data at hand | w. @dimitrivdv.bsky.social @rikossenkoppele.bsky.social + C-F Westin
I have found this very helpful #academicsky
December 16, 2024 at 11:08 AM
I have found this very helpful #academicsky
NIBS indeed has potential in AD, especially when we get over using one-size-fit-all implementations since macro-scale functional/structural networks, and their disruption patterns, entail notable idiosyncrasies.
#alzsky #neuroskyence #medsky
academic.oup.com/brain/advanc...
#alzsky #neuroskyence #medsky
academic.oup.com/brain/advanc...
November 25, 2024 at 9:57 AM
NIBS indeed has potential in AD, especially when we get over using one-size-fit-all implementations since macro-scale functional/structural networks, and their disruption patterns, entail notable idiosyncrasies.
#alzsky #neuroskyence #medsky
academic.oup.com/brain/advanc...
#alzsky #neuroskyence #medsky
academic.oup.com/brain/advanc...
Beyond cortical geometry: brain dynamics shaped by rare long-range connections
www.biorxiv.org/content/10.1...
#neuroscience #neuroimaging
www.biorxiv.org/content/10.1...
#neuroscience #neuroimaging
November 19, 2024 at 7:21 AM
Beyond cortical geometry: brain dynamics shaped by rare long-range connections
www.biorxiv.org/content/10.1...
#neuroscience #neuroimaging
www.biorxiv.org/content/10.1...
#neuroscience #neuroimaging
Cortical maps manifest hemispheric asymmetry in various forms and scales.
Our attempt to unveil them through 🧠 eigenmodes: doi.org/10.1101/2024.1…
Work led by Alicia Milloz✨#neuroskyencec#compneuroskyk#neurosciencece
Our attempt to unveil them through 🧠 eigenmodes: doi.org/10.1101/2024.1…
Work led by Alicia Milloz✨#neuroskyencec#compneuroskyk#neurosciencece
November 1, 2024 at 11:27 AM
Cortical maps manifest hemispheric asymmetry in various forms and scales.
Our attempt to unveil them through 🧠 eigenmodes: doi.org/10.1101/2024.1…
Work led by Alicia Milloz✨#neuroskyencec#compneuroskyk#neurosciencece
Our attempt to unveil them through 🧠 eigenmodes: doi.org/10.1101/2024.1…
Work led by Alicia Milloz✨#neuroskyencec#compneuroskyk#neurosciencece
A beautifully-illustrated read, taking you through neat microstructure modeling techniques used to extract exquisite tissue properties from seemingly-minimally-featured diffusion MRI signals!
Paper: www.sciencedirect.com/science/arti...
#neuroskyence #neuroimaging
Paper: www.sciencedirect.com/science/arti...
#neuroskyence #neuroimaging
October 18, 2023 at 2:11 PM
A beautifully-illustrated read, taking you through neat microstructure modeling techniques used to extract exquisite tissue properties from seemingly-minimally-featured diffusion MRI signals!
Paper: www.sciencedirect.com/science/arti...
#neuroskyence #neuroimaging
Paper: www.sciencedirect.com/science/arti...
#neuroskyence #neuroimaging
🔟 [🧵 end] We have also recently shown applicability of learned graphs within the context of 🧠 fingerprinting.
Paper: doi.org/10.1101/2023...
We look forward to further extending the learning method & its applications, &/or to see others do so! :)
#neuroskyence #compneurosky #neuroimaging #EEG
Paper: doi.org/10.1101/2023...
We look forward to further extending the learning method & its applications, &/or to see others do so! :)
#neuroskyence #compneurosky #neuroimaging #EEG
October 14, 2023 at 11:31 AM
🔟 [🧵 end] We have also recently shown applicability of learned graphs within the context of 🧠 fingerprinting.
Paper: doi.org/10.1101/2023...
We look forward to further extending the learning method & its applications, &/or to see others do so! :)
#neuroskyence #compneurosky #neuroimaging #EEG
Paper: doi.org/10.1101/2023...
We look forward to further extending the learning method & its applications, &/or to see others do so! :)
#neuroskyence #compneurosky #neuroimaging #EEG
9️⃣ The proposed method outperforms several alternative methods in motor imagery decoding.
October 14, 2023 at 10:55 AM
9️⃣ The proposed method outperforms several alternative methods in motor imagery decoding.
8️⃣ Rather than directly using GFT coefficients, we find an optimal linear combination of them based on their dynamics during the course of a given motor imagery trial.
This is done using the Fukunaga-Koontz transform (FKT), similar to CSP-based methods in EEG.
This is done using the Fukunaga-Koontz transform (FKT), similar to CSP-based methods in EEG.
October 14, 2023 at 10:54 AM
8️⃣ Rather than directly using GFT coefficients, we find an optimal linear combination of them based on their dynamics during the course of a given motor imagery trial.
This is done using the Fukunaga-Koontz transform (FKT), similar to CSP-based methods in EEG.
This is done using the Fukunaga-Koontz transform (FKT), similar to CSP-based methods in EEG.
7️⃣ Within the terminology of graph signal processing, we denote the resulting decomposition coefficients (of an EEG map) as the graph Fourier transform (GFT) of the map.
We show the applicability of these coefficients within the context of motor imagery decoding.
We show the applicability of these coefficients within the context of motor imagery decoding.
October 14, 2023 at 10:54 AM
7️⃣ Within the terminology of graph signal processing, we denote the resulting decomposition coefficients (of an EEG map) as the graph Fourier transform (GFT) of the map.
We show the applicability of these coefficients within the context of motor imagery decoding.
We show the applicability of these coefficients within the context of motor imagery decoding.
6️⃣ Eigenmaps of learned graphs provide a better decomposition of EEG maps into components associated with a wide range of spatial frequencies.
October 14, 2023 at 10:52 AM
6️⃣ Eigenmaps of learned graphs provide a better decomposition of EEG maps into components associated with a wide range of spatial frequencies.
5️⃣ We learned subject-specific graphs, leading to subject-specific eigenmaps; 20 selected eigenmaps are shown across 3 subjects.
Subject-specific eigenmaps provide an ON basis, which can be used for subject-tailored feature extraction from EEG maps.
Subject-specific eigenmaps provide an ON basis, which can be used for subject-tailored feature extraction from EEG maps.
October 14, 2023 at 10:52 AM
5️⃣ We learned subject-specific graphs, leading to subject-specific eigenmaps; 20 selected eigenmaps are shown across 3 subjects.
Subject-specific eigenmaps provide an ON basis, which can be used for subject-tailored feature extraction from EEG maps.
Subject-specific eigenmaps provide an ON basis, which can be used for subject-tailored feature extraction from EEG maps.
4️⃣ Difference in spatial topography of eigenmaps is reflected in their associated eigenvalues, but can also be quantified by computing zero-crossings of the maps.
Eigenmaps of learned graphs provide a better discretization of spectral range in relation to spatial frequency.
Eigenmaps of learned graphs provide a better discretization of spectral range in relation to spatial frequency.
October 14, 2023 at 10:50 AM
4️⃣ Difference in spatial topography of eigenmaps is reflected in their associated eigenvalues, but can also be quantified by computing zero-crossings of the maps.
Eigenmaps of learned graphs provide a better discretization of spectral range in relation to spatial frequency.
Eigenmaps of learned graphs provide a better discretization of spectral range in relation to spatial frequency.
3️⃣ The spatial topography of eigenmaps of learned graphs notably differ from that of a correlation graph.
October 14, 2023 at 10:49 AM
3️⃣ The spatial topography of eigenmaps of learned graphs notably differ from that of a correlation graph.
2️⃣ The spectra of learned graphs notably differ from that of correlation graphs.
On the right, several eigenmaps of a learned graph are shown, spanning the spectrum of the graph.
On the right, several eigenmaps of a learned graph are shown, spanning the spectrum of the graph.
October 14, 2023 at 10:47 AM
2️⃣ The spectra of learned graphs notably differ from that of correlation graphs.
On the right, several eigenmaps of a learned graph are shown, spanning the spectrum of the graph.
On the right, several eigenmaps of a learned graph are shown, spanning the spectrum of the graph.
1️⃣ The learning scheme allows inferring a graph structure on which an ensemble training set of EEG maps are observed as smooth spatial functions.
Notably, the scheme allows inferring sparse graphs; two methods used, one (log-penalized) resulting in greater sparsity.
Notably, the scheme allows inferring sparse graphs; two methods used, one (log-penalized) resulting in greater sparsity.
October 14, 2023 at 10:46 AM
1️⃣ The learning scheme allows inferring a graph structure on which an ensemble training set of EEG maps are observed as smooth spatial functions.
Notably, the scheme allows inferring sparse graphs; two methods used, one (log-penalized) resulting in greater sparsity.
Notably, the scheme allows inferring sparse graphs; two methods used, one (log-penalized) resulting in greater sparsity.
Happy to share our work on 🧠 graph structure learning from EEG ✨ wherein we use eigenmodes of learned graphs to decompose EEG maps & extract features.
Paper: doi.org/10.1016/j.bs...
#neuroskyence #compneurosky #neuroimaging #EEG
🧵 ⤵️
Paper: doi.org/10.1016/j.bs...
#neuroskyence #compneurosky #neuroimaging #EEG
🧵 ⤵️
October 14, 2023 at 10:45 AM
Happy to share our work on 🧠 graph structure learning from EEG ✨ wherein we use eigenmodes of learned graphs to decompose EEG maps & extract features.
Paper: doi.org/10.1016/j.bs...
#neuroskyence #compneurosky #neuroimaging #EEG
🧵 ⤵️
Paper: doi.org/10.1016/j.bs...
#neuroskyence #compneurosky #neuroimaging #EEG
🧵 ⤵️