Alina Studenova

Alina Studenova - in-consciousness-we-trust Recently, I have witnessed a discussion during one of the talks in our department. The issue was about the study of motor imagery. The experiment went as follows: participants imagined a movement; the...

I read another book. “In Consciousness We Trust: The Cognitive Neuroscience of Subjective Experience” by Hakwan Lau. Finally, the book about consciousness (=subjective experience) that I can comprehend. Read my opinion here
www.alinastudenova.com/home/blog/bo...

Such a cool and rigorous paper by @studenova.bsky.social !
bsky.app/profile/stud...

Another nice example of merging empirical data analysis with computational modelling to gain insights into brain biophysics.

I thank the wonderful co-authors.😍
Felix Ströckens, Luke J. Edwards @lukejoeledwards.bsky.social, Anna-Lena Stroh @lenastroh.bsky.social, Saskia Helbling @saskiahelbling.bsky.social, Burkhard Maess, Kerrin J. Pine, Katrin Amunts, Evgeniya Kirilina, Nikolaus Weiskopf, Arno Villringer, Vadim Nikulin

So now I'm also interested in between- vs. within-variability in structure and function. What are the factors that influence each kind of variability? (6/6)

We speculate that positive correlations across regions are driven by the total number of neurons and the negative correlation across participants—by circuits involving inhibitory interneurons. We did some simulations with TVB to support this idea. (5/6)

Meanwhile, across participants, alpha power correlated *negatively* with myelin and iron estimates from a mid-cortical component. (4/6)

Across regions, using data from different cohorts and different modalities, we found that alpha power correlated positively with the thickness of layer IV and with myelin and iron estimates from the mid-cortical component. (3/6)

Structure was obtained from the post-mortem brain and high-resolution MRI (in-vivo histology). Dynamics were obtained from EEG and MEG.
We looked for relations from two angles: across regions and across participants. (2/6)

How dynamics arise from the structure is my biggest interest. In this study, we started with a small step and asked how structure constrains dynamics. Spoiler: would that it were so simple… (1/6)

Proprioceptive neurons are found in layer 2/3 of the mouse S1. They are involved in movement execution. This paper shows that without those neurons, the movement will be executed anyway. But the trajectory becomes more stereotyped and less variable. Cool!🤩 Noise in behavioral estimates is not noise.

This paper suggests that in visual statistical learning, associations are better captured when predictability is high in both directions. Dual Factor Heuristic best explains learning effects. Interesting!👍 If a meeting is always on Monday, do Mondays predict meetings or do meetings predict Mondays?

underappreciated concept 🙂

for more dipoles: neighboring dipoles have different orientations & can have different stimulus preferences. so even though topographies look similar, subtle differences remain. maybe that's why it is possible to decode information from relatively few sensors?

The angle of the dipole obviously matters. If active dipoles in two conditions are slightly misaligned, a difference can be observed. However, this difference is not driven by differences in amplitudes, but only by differences in orientations. Here, I simulated such a scenario.
#brainmovies

Processing speed, attention, and working memory correlate with both P300 and intelligence, and therefore may explain the association. I wish more studies would control for domain-general capacities.

A meta-analysis showing a small but significant association between P300 amplitude and latency and general cognitive abilities (intelligence). The larger and earlier the P300, the more intelligent the person. The elephant in the room, which the authors also acknowledge, is domain-general capacities.

In y/n detection tasks, the internal signal and noise distributions have unequal variance. Instead of using d', we should be using dₐ. dₐ is usually estimated from confidence ratings, but this paper shows (in visual tasks) that it's possible to get dₐ from reaction times. Good news!👏 Will surely try

Woop woop 🎉 Excited to share that our paper has been accepted in Imaging Neuroscience!

Can ventral striatal reward signals — originating from deep within the brain — be reconstructed from scalp EEG?

We tried to answer this question using deep learning.

More soon 👀 Stay tuned.

With each of repetition runs,
Here comes magnetic resonance.
#brainrhymes

And after that the spins relax:
Projections onto xy-plain
Are lost, while z-component part
Is gained successfully again.

From that the relaxation times
Of vectors lost and vectors gained,
(Which correspondingly T2
And T1-time) are ascertained.

The water proton spins align
With the external magnet field.
Magnetization ramping up
Such an alignment will yield.

The RF pulse then flips the spins
To certain number of degrees,
The frequency is set as such,
They're flipping more or less with ease.

Explainable AI techniques have been proposed as tools to "debug" machine learning models. In a popular example, authors discovered a photographer's watermark when applying XAI in an ima... Explainable AI techniques have been proposed as tools to "debug" machine learning models. In a popular example, authors discovered a photographer's watermark when applying XAI in an image classificat...

XAI techniques have been proposed as tools to "debug" machine learning models by revealing what they have learned.

Challenging this narrative, we find that XAI attributions are strongly dominated by salient features independent of the role of these features for a model.

See: tinyurl.com/48m4axnv

V. interesting!⚡
Quote "RSA is not an accurate, direct measure of CNS-generated cardiac vagal activity."
RSA - Respiratory Sinus Arrhythmia, the rhythmic change in heart rate during breathing.

LLMs have cognitive biases too, such as confirmation bias, anchoring bias and negativity bias. This survey provides a nice overview. The most interesting and informative part is section with examples.😄

Cool! If this cloud look like a brain, than I'm in motor cortex😄

I still entertain this possibility. P300 is quite likely a composite evoked response with multiple source locations, and possibly it is generated via several mechanisms. The authors actually also admit this possibility. Great!👏

Nice to see a study addressing mechanisms of evoked response generation! This study obtained that P300 and alpha desynchronization stem from different sources. In the past, I claimed that P300 is partially generated by alpha amplitude asymmetry or baseline-shift mechanism, i.e., they are related.

If oscillations are modulated in amplitude from trial to trial, and those oscillations have a non-zero mean, they "create" an evoked response. It is a baseline-shift mechanism for ERP/ERF generation. Here is a simulated example of how alpha oscillations may generate readiness potential.
#brainmovies

Activity-dependent and injury-induced myelination is carried out via different routes. In healthy zebrafish, activity-related plasticity was linked to recently proliferated oligodendrocyte precursor cells (OPCs). But during demyelination, OPCs without prior proliferation were also recruited. Cool!👍

Energy pathways from whole-brain transcriptomics have certain spatial distributions. 4 out of 5 maps look similar, while pentose phosphate pathway looks different. Interesting that energy maps only partially agree with PET data. Nice!👏 I wonder how those maps correlate with vascular architecture.