CoChat

Mapping Cortical Laminar Structure in the 3D BigBrain

Open access

The DOI matches its Crossref record; 47 of the 47 references it cites were found in a registry or an index.

Abstract

Histological sections offer high spatial resolution to examine laminar architecture of the human cerebral cortex; however, they are restricted by being 2D, hence only regions with sufficiently optimal cutting planes can be analyzed. Conversely, noninvasive neuroimaging approaches are whole brain but have relatively low resolution. Consequently, correct 3D cross-cortical patterns of laminar architecture have never been mapped in histological sections. We developed an automated technique to identify and analyze laminar structure within the high-resolution 3D histological BigBrain. We extracted white matter and pial surfaces, from which we derived histologically verified surfaces at the layer I/II boundary and within layer IV. Layer IV depth was strongly predicted by cortical curvature but varied between areas. This fully automated 3D laminar analysis is an important requirement for bridging high-resolution 2D cytoarchitecture and in vivo 3D neuroimaging. It lays the foundation for in-depth, whole-brain analyses of cortical layering.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1093/cercor/bhy074

Cites

47 of 47 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. Automatically identified cortical layers on the BigBrain displayed on three orthogonal planes. The brain was sectioned coronally and reconstructed to create a 3D isotropic volume at 20 μm. Cortical intensity profiles (zoomed insert), perpendicular to the cortex, were extracted at all vertices on the surface and used to identify continuous cortical layers. On the gray-scale histological images, minimum intensity pixels are white, maximum are black.
Figure 1. Automatically identified cortical layers on the BigBrain displayed on three orthogonal planes. The brain was sectioned coronally and reconstructed to create a 3D isotropic volume at 20 μm. Cortical intensity profiles (zoomed insert), perpendicular to the cortex, were extracted at all vertices on the surface and used to identify continuous cortical layers. On the gray-scale histological images, minimum intensity pixels are white, maximum are black.CC BY · PubMed Central
Figure 2. Investigation of density profiles. (a) Cellular density profiles were created from areal measurements of thickness and neuronal density of each cortical layer reported by von Economo & Koskinas (von Economo and Koskinas 1925). The first letter of each areal code corresponds to the lobe, the second letter indicates successive lobar measurements e.g., OA, occipital area A, TB, temporal area B, etc. All von Economo profiles exhibited a large positive gradient between layer I, which has the lowest neuronal density, and layer II. Most profiles also exhibited a consistent mid-profile peak corresponding to high neuronal density in layer IV. A notable exception included FA, the agranular motor cortex, so called because no granular layer is visible under a microscope. (b) von Economo density profiles were compared with intensity profiles from corresponding areas in the BigBrain. Shown here are von Economo areas OA (extrastriatal occipital cortex) and FD (granular anterior frontal cortex). The layer I/II boundary and mid-cortical layer IV peak is consistently present in both von Economo and BigBrain profiles.
Figure 2. Investigation of density profiles. (a) Cellular density profiles were created from areal measurements of thickness and neuronal density of each cortical layer reported by von Economo & Koskinas (von Economo and Koskinas 1925). The first letter of each areal code corresponds to the lobe, the second letter indicates successive lobar measurements e.g., OA, occipital area A, TB, temporal area B, etc. All von Economo profiles exhibited a large positive gradient between layer I, which has the lowest neuronal density, and layer II. Most profiles also exhibited a consistent mid-profile peak corresponding to high neuronal density in layer IV. A notable exception included FA, the agranular motor cortex, so called because no granular layer is visible under a microscope. (b) von Economo density profiles were compared with intensity profiles from corresponding areas in the BigBrain. Shown here are von Economo areas OA (extrastriatal occipital cortex) and FD (granular anterior frontal cortex). The layer I/II boundary and mid-cortical layer IV peak is consistently present in both von Economo and BigBrain profiles.CC BY · PubMed Central
Figure 3. Identification of cortical layers. (a) Automatically detected layers from cortical profiles between pial and white surfaces. (b) An initial estimate for layer I/II boundary position was placed at the layer I/gray boundary of the tissue classification, made on 200 μm isotropic voxels (yellow/gray voxels). Surfaces were adjusted to the nearest maximum gradient, with two different smoothing steps to produce a smooth surface following the visible boundary between these layers: intensity values were smoothed across the surface and vertices were moved to the position on the profile nearest the smoothed value, and geometric mesh smoothing removed high curvature kinks by averaging neighboring coordinates—as shown by the arrow and the green dotted line. (c) The initial position of layer IV was placed at the first major peak after an upwards inflection in the intensity profile, which corresponds to layer II/III. Surfaces were adjusted to the nearest profile maximum, with two iterative smoothing steps: equivolumetric depth values (spectral colored 200 μm voxels) were smoothed across the surface to reposition smooth but inaccurately placed regions (purple dashed line) and geometric mesh smoothing was again used to remove high curvature kinks. On the gray-scale histological images, minimum intensity pixels are white, maximum are black.
Figure 3. Identification of cortical layers. (a) Automatically detected layers from cortical profiles between pial and white surfaces. (b) An initial estimate for layer I/II boundary position was placed at the layer I/gray boundary of the tissue classification, made on 200 μm isotropic voxels (yellow/gray voxels). Surfaces were adjusted to the nearest maximum gradient, with two different smoothing steps to produce a smooth surface following the visible boundary between these layers: intensity values were smoothed across the surface and vertices were moved to the position on the profile nearest the smoothed value, and geometric mesh smoothing removed high curvature kinks by averaging neighboring coordinates—as shown by the arrow and the green dotted line. (c) The initial position of layer IV was placed at the first major peak after an upwards inflection in the intensity profile, which corresponds to layer II/III. Surfaces were adjusted to the nearest profile maximum, with two iterative smoothing steps: equivolumetric depth values (spectral colored 200 μm voxels) were smoothed across the surface to reposition smooth but inaccurately placed regions (purple dashed line) and geometric mesh smoothing was again used to remove high curvature kinks. On the gray-scale histological images, minimum intensity pixels are white, maximum are black.CC BY · PubMed Central
Figure 4. Verification of cortical layers. (a) Distance of layer IV vertices from layer IV boundaries manually delineated in 37 areas on 5 μm histological sections. Most points lie directly within or close to layer IV and there is no apparent systematic bias in the small number of points lying outside of these bounds. (b) Verification that local, morphologically determined variability in automatically identified layer IV depth followed that of the manually delineated layer IV bounds, shown here for six samples on a single coronal section: 2807. The percentage depth of the midpoint of the manually delineated layers between the white and layer I/II boundary and the percentage depth of the automated layer IV between automated white and layer I/II surfaces was compared for each sample region. The mean Pearson’s correlation coefficient for this section was r = 0.80 (across all sections r = 0.72). (c) Visual validation that layer I/II and layer IV closely follow manually defined boundaries for section 2807 at 20 μm. The automatically identified surfaces have the following colors: white–red, layer IV - burgundy, layer I/II - green, pial–blue. The white–blue overlays indicate distance below the manually defined layer IV/V boundary, with the lower bound at the white surface. The white–red overlays indicate distance above layer III/IV boundary, with the upper bound at the layer I/II boundary. The automatically identified layer IV surface consistently lies within or close to layer IV. On the gray-scale histological images, minimum intensity pixels are white, maximum are black.
Figure 4. Verification of cortical layers. (a) Distance of layer IV vertices from layer IV boundaries manually delineated in 37 areas on 5 μm histological sections. Most points lie directly within or close to layer IV and there is no apparent systematic bias in the small number of points lying outside of these bounds. (b) Verification that local, morphologically determined variability in automatically identified layer IV depth followed that of the manually delineated layer IV bounds, shown here for six samples on a single coronal section: 2807. The percentage depth of the midpoint of the manually delineated layers between the white and layer I/II boundary and the percentage depth of the automated layer IV between automated white and layer I/II surfaces was compared for each sample region. The mean Pearson’s correlation coefficient for this section was r = 0.80 (across all sections r = 0.72). (c) Visual validation that layer I/II and layer IV closely follow manually defined boundaries for section 2807 at 20 μm. The automatically identified surfaces have the following colors: white–red, layer IV - burgundy, layer I/II - green, pial–blue. The white–blue overlays indicate distance below the manually defined layer IV/V boundary, with the lower bound at the white surface. The white–red overlays indicate distance above layer III/IV boundary, with the upper bound at the layer I/II boundary. The automatically identified layer IV surface consistently lies within or close to layer IV. On the gray-scale histological images, minimum intensity pixels are white, maximum are black.CC BY · PubMed Central
Figure 5. Layer depth and morphology. (a) Layer IV percentage depth was strongly predicted by mean curvature (mm−1) measured on a mid-surface between layer I/II and white surfaces, being deeper in negatively curved fundi and more superficial in positively curved crowns. The relationship between layer IV depth and curvature was best predicted by a cubic regression. This relationship is consistent with predictions from the equivolumetric model. (b) Relative depth of automatically identified layer IV displayed on a spatially smoothed white surface. Consistent with histological studies (Bok 1929), layer IV was located deeper in sulcal fundi and more superficial in gyral crowns. Upper layers (II & III) were therefore relatively thinner in gyral crowns and thicker in sulcal fundi, while the opposite was true for lower cortical layers (V–VI). (c) Layer IV equivolumetric depth showed almost no correlation with curvature, lending support to the use of this model in modeling layers in MRI. (d) Equivolumetric depth of layer IV on a spatially smoothed white surface. Across cortical areas, there remained a variability in layer IV equivolumetric depth. Layer IV was relatively deeper in the calcarine sulcus (V1) and parts of the medial prefrontal cortex. Due to the presence of border effects and allocortex, areas directly adjacent to the medial wall (black) should be interpreted with caution.
Figure 5. Layer depth and morphology. (a) Layer IV percentage depth was strongly predicted by mean curvature (mm−1) measured on a mid-surface between layer I/II and white surfaces, being deeper in negatively curved fundi and more superficial in positively curved crowns. The relationship between layer IV depth and curvature was best predicted by a cubic regression. This relationship is consistent with predictions from the equivolumetric model. (b) Relative depth of automatically identified layer IV displayed on a spatially smoothed white surface. Consistent with histological studies (Bok 1929), layer IV was located deeper in sulcal fundi and more superficial in gyral crowns. Upper layers (II & III) were therefore relatively thinner in gyral crowns and thicker in sulcal fundi, while the opposite was true for lower cortical layers (V–VI). (c) Layer IV equivolumetric depth showed almost no correlation with curvature, lending support to the use of this model in modeling layers in MRI. (d) Equivolumetric depth of layer IV on a spatially smoothed white surface. Across cortical areas, there remained a variability in layer IV equivolumetric depth. Layer IV was relatively deeper in the calcarine sulcus (V1) and parts of the medial prefrontal cortex. Due to the presence of border effects and allocortex, areas directly adjacent to the medial wall (black) should be interpreted with caution.CC BY · PubMed Central
Figure 6. Measurement errors in 2D histology. (a) Errors are introduced due to the angle α between the cortical profile vector and the original 2D coronal section. (b) Estimated 2D thickness measurement error of a 3 mm cortex due to the angle between the coronal plane and cortical profiles. Red areas indicate errors of 1 mm or greater. Thickness measurements made at vertices colored blue and purple would be little affected by oblique slicing. Occipital (top left) and frontal (top right) views showed many profiles approaching 90° such that a single cortical column would be distributed across hundreds of coronal sections making laminar analysis impossible in 2D. (c) Histogram showing the distribution of angles across vertices. At angles above 18°, which made up approximately 69% of vertices, there was a 5% overestimation in any 2D thickness measurement. (d) Calculation of the error in thickness measurement associated with a given profile section angle.
Figure 6. Measurement errors in 2D histology. (a) Errors are introduced due to the angle α between the cortical profile vector and the original 2D coronal section. (b) Estimated 2D thickness measurement error of a 3 mm cortex due to the angle between the coronal plane and cortical profiles. Red areas indicate errors of 1 mm or greater. Thickness measurements made at vertices colored blue and purple would be little affected by oblique slicing. Occipital (top left) and frontal (top right) views showed many profiles approaching 90° such that a single cortical column would be distributed across hundreds of coronal sections making laminar analysis impossible in 2D. (c) Histogram showing the distribution of angles across vertices. At angles above 18°, which made up approximately 69% of vertices, there was a 5% overestimation in any 2D thickness measurement. (d) Calculation of the error in thickness measurement associated with a given profile section angle.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Cited by

See all 85

Metadata

Authors
Wagstyl, K., Lepage, C., Bludau, S., Zilles, K., Fletcher, P. C., Amunts, K., & Evans, A. C.
Affiliations
  • Montreal Neurological Institute and Hospital (CA)
  • Forschungszentrum Jülich (DE)
  • University of Cambridge (GB)
Corresponding author
Wagstyl, Konrad
Venue
Cerebral Cortex
Published
2018-04-18
Volume
28
Issue
7
Pages
2551-2562
ISSN
  • 1047-3211 (ISSN-L)
  • 1460-2199
OpenAlex
W2801394998
Cited by
85
References
47
Citations by year
2026: 5 · 2025: 9 · 2024: 11 · 2023: 11 · 2022: 9 · 2021: 10 · 2020: 18 · 2019: 10 · 2018: 2
FWCI
5.00
Type
journal article
Language
en
Open access
Locations
  • Cerebral Cortex (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • UCL Discovery (University College London) (submittedVersion)
  • JuSER (Forschungszentrum Jülich) (submittedVersion, other-oa)
  • RWTH Publications (RWTH Aachen) (submittedVersion, other-oa)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • Apollo (cc-by)
Funders
  • European Union’s Horizon 2020 Framework Programme for Research and Innovation (720270)
  • Parts of this research received funding from Neuroscience in Psychiatry Network, a strategic award by the Wellcome Trust (095 844/Z/11/Z)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Medical Image Segmentation Techniques
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Advanced Neuroimaging Techniques and Applications
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
Keywords
BigBrain, human cerebral cortex, cortical layers, cortical curvature, pial surface, 3D histology, cytoarchitecture
MeSH terms
Brain Mapping, Cerebral Cortex, Humans, Magnetic Resonance Imaging, Anisotropy, Imaging, Three-Dimensional

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.