List of potential layer-fMRI application studies

Recently, my adviser Peter Bandettini asked me to compile a list of all potential applications for layer-fMRI. It was a fun exercise to think about it. Below, you find a list of all potential applications that I could think of. 

1.) Suggestions from Klaas Enno Stephan regarding computational Psychiatry

Klaas Enno Stephan wrote a nice review article about potential applications of layer-fMRI. He argues that most psychiatric diseases are network diseases and that they can be better understood with layer-fMRI. He specifically focuses on:

  1. Autism
  2. Schizophrenia
  3. Psychosomatics

He suggests the application of functional tasks of Mismatch negativity. this task does not require training and works even with patients in a mild coma.

2.) Suggestions from Bob Turner

Bob Turner suggested a list of the following fMRI tasks for layer-dependent applications:

Slide from Bob Turners talk, March 28th 2017, Glasgow
  1. motor imaginary vs. actual motion
  2. visual imaginary vs. actual vision
  3. auditory imaginary vs. actual hearing
  4. cross modal vs. unimodal stimulations
  5. top-down vs. bottom-up attentional modulation
  6. self motion vs. other-motion in mirror system brain areas
  7. touch experienced vs. touch observed

3.) Suggestions summarized by Lawrence and colleagues

  1. Characterizing fMRI responses and receptive field characteristics across cortical depths
  2. Forward, backward, and lateral message-passing within the neocortex
  3. Visual mental prediction
  4. Working memory and mental imagery
  5. Selective attention
  6. Visual saliency
  7. Multisensory integration: (V1/A1/S1)
  8. Consciousness
  9. Hallucinations and delusions: testing different hypotheses of their underlying mechanisms

4.) Excitation and Inhibition

Excitatory and inhibitory connections are differently distributed across cortical depth. Thus, layer-fMRI can address questions about the underlying circuits engaged:

  • In multiple studies it is not clear whether the fMRI signal increases due to increased excitatory input or due to reduced inhibitory input. The different layer-dependent signatures for those two cases can be investigated with layer-fMRI.
  • The phenomenon of negative BOLD can be explained by multiple mechanisms including: surround suppression, inter-hemispheric inhibition, blood stealing, venous back-pressure, mismatch of CBV and CBF, etc. The different layer-dependent fMRI signatures of those effects could shed light on which of the potential mechanisms is responsible.

5.) Inter-participant differences

There is a huge variability of data quality and layer-fMRI responses across participants. And I could not find any literature that addresses this. I think the inter-individual differences of layer-responses should be investigated more carefully in the future for multiple reasons:

  1. The gyrification pattern variability across people is in the range of 5-10 cm.  This means that the sub-millimeter structures of layers fMRI responses need to be conducted in the individual persons’ brain anyway. This makes inter-participant investigations very easy.
  2. From an experimentator perspective it is very frustrating to have a drop-out rate of 20%-30% participants (common on my experiment). When my experiment doesn’t work in 3 out of 10 participants, and I have no clear explanation for it, reviewers will give me a hard time about it. Thus, it would be important to understand why some people are bad layer-fMRI participants and pre-screen them accordingly.
  3. In one isolated case of a participant who was a violin player, I found that she could perform the same tapping-movement task as control participants without engaging the deeper output layers of the primary motor cortex. It would be interesting to investigate these effects more systematically.

6.) Investigating the effect of neuro-modulators and receptors

Different layers in different brain areas have a very different expression of synaptic receptors.

  • Thus, a disturbance of the neuro-receptors e.g. in chronic stress, or with drug abuse would be expected to have a clear layer-signature in layer-fMRI too.
  • The unique layer-dependent receptor fingerprint across brain areas should be manifested in layer-fMRI as well.

7.) One-shot learning

What is so peculiar about the picture below? Did you immediately see the cigar sticking out of the brick wall? Some people see it immediately while others need several minutes. And once you see it, it cannot be unseen: from this moment on, your brain is different from another person’s brain. Now the visual cortex also receives feedback input from higher-order cognitive brain areas into its upper cortical layers. The middle cortical layers, however, continue to receive feedforward input from the eyes, irrespective of whether one sees the cigar or not. Seeing the cigar, it is the feedback into the visual cortex that distinguishes our brain state from that of people who don’t. These kind of processes could be investigated with layer fMRI.

Cigar sticking out of a brick wall. See solution with highlighted cigar at the bottom of this post.

8.) Development

The cortical layers play a crucial role in maturation and brain development. For example, the primary motor cortex of fetal brains has a layer 4. However, the primary motor cortex of healthy adult brains lack layer 4. Only for few specific movement disorders, layer 4 of the primary motor cortex does not completely disappear across the maturation. Layer-fMRI could shed some new light on the function of layer 4 in the primary motor cortex across these developmental processes.

9.) Layer-fMRI in animal models

As in other fields of fMRI, animal models could help to serve as a ground truth for subsequent human studies. Animal studies are valuable for layer-fMRI with respect to multiple aspects:

  1. Future animal studies could investigate the effect of directional effectivity my means of invasive modulation of neural connections. E.g, To see how the layer-dependent activity changes if you cut the connections from areas that are higher in the hierarchy or if you cut neural connections to areas lower in the hierarchy.
  2. Future animal studies could investigate the layer-dependent effect of layer-dependent neuro-receptors. E.g. pharmacological interventions can block receptors of specific layers compared to other layers.
  3. The luxury of higher data quality and acquiring segmented EPI allows much higher resolutions. Thus, laminar fMRI can investigate the biological limits of layer-fMRI.

10.) Exploratory studies: describing what we see

Layer-fMRI is a rather new method and provides data that could not be investigated until now. This means that layer-fMRI is in a phase of its history where is makes sense to conduct many exploratory studies too. E.g. it would be interesting to repeat any fMRI task that has been conduced in the last 20 and describe the layer-dependent result for it. Maybe some results will be surprising.

11.) Layer Atlas

Cyto- and Myeloarchitectonic layer distribution have been tightly coupled big brain area mapping efforts throughout the history: Brodman and the Vogts distinguished up to 200 cortical brain areas based on specific layer distributions of cells neuronal cell bodies and myelin. The Brodman areas are largely defined by their cytoarchitectonic layer features. After that, multiple big atlases mapped multiple layer-dependent features, such as layer-dependent receptor expression (Amunts and Zilles) and MRI contrasts, such as T1 and T2* profiles.

Now that fMRI is capable to achieve layer fMRI resolutions too, I believe it’s time to generate an layer-fMRI atlas of layer-fMRI responses across all brain areas. I would hypothesize that every brain area has a different inherent layer-dependent fMRI signature.

In the same fashion as pervious studies made atlases about layer-dependent distributions of T1 and cytoarchitecture in ex-vivo samples, layer fMRI can create atlases of functional layer-features. E.g. layer hubness, layer-dependent vascular density, etc.

This enables new questions to be adressed. E.g. organizations across the cortical hierarchies (see anti correlation of hierarchy level and hub depth).

12.) Characterizing brain areas based on their feed-forward and feed-back driven layer-sigantures

It appears that some brain areas have most of its synchronous resting-state fluctuations in the superficial layers. The deeper layer, on the other hand seem to be less synchronous. In other brain areas this this is the other way around and the deeper layers are the once that are most synchronous. Based on the corresponding layer-dependent hubness, areas can be classified as feed-forward-driven or feedback-driven.

Screenshot 2018-09-04 15.46.24.png
example of classifying brain areas based on their layer-dependent resting-state fluctuation amplitude.

13.) Investigation of resting-state layer-fMRI to determine inter-area hierarchies.

Layer-dependent resting-state correlation profiles with respect to any given reference brain area can be strongest in the feed-forward input layers or in the feed-back input layers. This gives some indication of the area is above of below in the hierarchy.

Example of how the resting-state connectivity can be explored. Here I picked any seed voxel and asked: which other brain area is receiving input from here? In this example the seed region goes along the visual hierarchy and thus less and less areas as higher in the hierarchy.

14.) Determining a directional connectome

With standard resting-state fMRI it is possible to determine all functional connection between brain areas. With layer fMRI, it is possible to add an additional dimension to this. Focusing on the off-diagonal entries (e.g. layer 4 and layer 3 edges), it becomes possible to provide information about the directionality of the functional connections.

Example of layer-dependent connectome analysis. Instead of just investigating the functional connectivity strength of all brain areas. Layer-fMRI can investigate the the directionality of these connections. This might be valuable information of graph-theoretical studies.

These cross-correlation matrices are inspired by the work from Polimeni et al. 2010.

15.) Comparison of layer-fMRI and electrophysiology

It is believed that different frequency bands of electrical activity has a different effect on the fMRI signal.

  • 0-8Hz has not effect on the fMRI signal
  • 8Hz-14Hz modulates the BOLD magnitude
  • 14Hz-20Hz has no effect on the fMRI signal
  • 20Hz-38Hz modulates the shape of the BOLD time course
  • 38Hz-50Hz has no effect on the BOLD signal
  • 50Hz-250Hz modulated the BOLD magnitude.
  • 250Hz-900Hz has no effect on the BOLD signal
  • 900Hz-8000Hz has a large effect on the negative BOLD signal.

The group around Nikos Logothetis has investigated that these bands are differently manifested across cortical layers (e.g. here).

Also, Scheeringa et al. investigated inter-trial modulations of layer-dependent BOLD and compared it with simultaneously acquired EEG. They found that GE-BOLD modulations of the superficial layers match the modulations of the alpha band.

Future investigations of layer-fMRI and electrophysiology will provide a better understanding of these relationships during task and resting-state.



Solution of highlighted cigar referring to one-shot learning.




Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s