R. Austin Benn’s research while affiliated with Université Paris Cité and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (15)


Pipeline to construct the sensory integration model
a Mean time series was computed separately for V1, S1, and A1 based on Glasser’s MMP parcellations. b A non-negative linear model was used to generate sensory components within each vertex by using primary sensory time series as predictors. The lower Venn diagram provides a schematic of the different components within the above equation. c Ratios of the variance explained by primary sensory predictors were ranked and rescaled to a range from 0 to 1, representing one dimension of the sensory integration model, and were named magnitude (r). For each vertex, three sensory parameters (βV, βS, and βA) were converted into an angle (θ) using hue transformation, representing the other dimension that indicated the proportional contributions of different sensory modalities.
Relationship between sensory parameters and sensory integration model
a Surface mapping of group-averaged sensory parameters under movie-watching state. The top row is the visual parameter (βV), the middle is the somatosensory parameter (βS), and the bottom is the auditory parameter (βA). b Ternary plots depict the relationship between the proportions of sensory parameters and the sensory integration model under movie-watching state, whereas surface plots display the corresponding spatial locations on the cortical surface. The color representation in the top row is determined by the group-level sensory magnitude. The color in the middle row is derived by applying the group-level sensory angle as hue. The color in the bottom row combines the group-level angle as hue with the group-level magnitude as saturation. c Projection of group-level sensory magnitudes and sensory angles under movie-watching state onto a polar coordinate system. The color scheme is identical to that used in the bottom row of panel b.
Sensory magnitude and cortical hierarchy
a Left column shows surface mappings of sensory magnitude under movie-watching condition and the principal connectome gradient under resting-state condition⁸. The right density illustrates the relationship between sensory magnitude and principal gradient values. b The boxplot illustrates the network-wise distribution of the magnitude values based on a 7-network parcellation²⁴. For the boxplot, the middle line represents the median, while the box’s upper and lower limits correspond to the 75th and 25th percentiles, respectively. The whiskers extend up to 1.5 times the interquartile range from the upper and lower box limits. Note the striking topographical consistency between the magnitude map and the principal gradient map, despite methodological difference, indicating converging evidence.
Between-session correlations in magnitude and angle
The size of the diamond within the red circle represents the between-session correlation of angles, with larger diamonds meaning higher correlations. The color of the circle represents the between-session correlation of magnitude, the redder the color, the higher the correlation.
Altering spatial patterns of the sensory integration model along sensory streams
The top row is the visual stream from lower- to higher-level regions, as shown on the cortical surfaces. The middle and bottom rows are somatosensory and auditory streams respectively. BA, Brodmann area; STSd, dorsal superior temporal sulcus; STSv, ventral superior temporal sulcus.

+3

A function-based mapping of sensory integration along the cortical hierarchy
  • Article
  • Full-text available

November 2024

·

93 Reads

·

1 Citation

Communications Biology

·

R. Austin Benn

·

Robert Scholz

·

[...]

·

Sensory information mainly travels along a hierarchy spanning unimodal to transmodal regions, forming multisensory integrative representations crucial for higher-order cognitive functions. Here, we develop an fMRI based two-dimensional framework to characterize sensory integration based on the anchoring role of the primary cortex in the organization of sensory processing. Sensory magnitude captures the percentage of variance explained by three primary sensory signals and decreases as the hierarchy ascends, exhibiting strong similarity to the known hierarchy and high stability across different conditions. Sensory angle converts associations with three primary sensory signals to an angle representing the proportional contributions of different sensory modalities. This dimension identifies differences between brain states and emphasizes how sensory integration changes flexibly in response to varying cognitive demands. Furthermore, meta-analytic functional decoding with our model highlights the close relationship between cognitive functions and sensory integration, showing its potential for future research of human cognition through sensory information processing.

Download

Figure 1: Overview of the prediction model and scores used for the assessment of model performance. A Functional and structural measures used as predictors in our models, including resting-state functional connectivity PCA (RS-PCA) and ICA (RS-
Figure 2: A Discriminability of task contrast maps, quantifying how well individual differences are captured. B Accuracy of predicted task contrast maps. This score quantifies how well the overall group-level task distribution can be captured, but does not differentiate between capturing the group-level distribution or individual differences. Scores are shown for the task contrast with the highest test-retest discriminability for each of the seven tasks assessed for HCP. This includes the following contrasts: Language Story-Math, Motor Cue-Avg, Emotion Faces-Shapes, Working Memory 2Back-0Back, Social Theory of Mind-Random, Relational Rel-Match and Gambling Reward-Punish. Only the task names are shown in the figure for brevity. For a more complete description of the tasks see Barch (2013). Horizontal lines indicate both the test retest baseline (aimed at measuring the stability of task contrast maps in individuals across sessions, providing a noise ceiling for each task contrast for both discriminability and accuracy) and the group-average baseline (i.e. the scores that would be obtained if always only the groupaverage task map was was predicted, and by construction equaling zero in the case of discriminability). The apparent overlap in both baselines in panel B evidences that the mean accuracy -in contrast to the discriminability score -does not allow to distinguish whether or not the models are capturing individual differences.
Figure 3: Resting-State PCA-based prediction of full task maps using a linear model. A The model takes group-reference aligned individual resting-state PCA component loadings as input. B The coefficients for these loadings are estimated in a task-specific manner. Individual component loadings are aligned to group-level component loadings prior to model fit, ensuring that they match across subjects. Most predictive components for both the language and motor task are annotated (below the plot) by the highest correlating reference maps at the group-level, highlighting possible relationships to functional gradients (Gradients 3, 5 and 10), functional networks (Visual, Language, Default and Frontoparietal), cell types (Excitatory Type EX4) and microstructure (T1w/T2w). C The upper row contains the predicted map for a sample contrast (Language Story-Math) for a sample subject with intermediate individual discriminability based on the model (rank 34 out of 51 subjects, with lower rank indicating higher
Figure 4: Microstructure and morphology based prediction using per-parcel models. A Discriminability for each of the per-parcel models for the seven main task contrasts. B Discriminability scores from A plotted on the cortical surface for three task contrasts, showing that highest discriminability can be found in occipital regions. C Weights for the linear model predicting task maps, averaged across the five most discriminable per-parcel models, highlighting cortical thickness and curvature as predictive features.
Individual brain activity patterns during task are predicted by distinct resting-state networks that may reflect local neurobiological features

November 2024

·

132 Reads

Understanding how individual cortical features shape functional brain organization offers a promising framework for examining the principles of cognitive specialization in the human brain. This study explores the relationship between various cortical characteristics—i.e resting-state functional connectivity, structural connectivity, microstructure, morphology, and geometry—and the layout of task-specific functional activations. We employ linear models to predict the functional layout of the cortex at the individual level from each of these feature modalities. Our findings demonstrate that resting-state component loadings predict individual task activations, consistently across hemispheres and independent datasets. Whereas the first few components provide a common space for functional activations across tasks, predictive higher-order component loadings demonstrated task-specificity. Cortical microstructure/morphology was notably predictive of activation strength in the occipital cortex, highlighting its relevance for cortical functional specialization. By relating resting state components to a set of reference maps of cortical organization, we identify associations that suggest possible neurobiological underpinnings of specific cognitive functions. The remaining feature modalities were only predictive of group-level functional activations. These results advance our understanding of how distinct cortical features may contribute to functional specialization, guiding future inquiry into the organization of cognitive functions on the cortex.


Comparative neuroimaging of the carnivoran brain: Neocortical sulcal anatomy

October 2024

·

35 Reads

Carnivorans are an important study object for comparative neuroscience, as they exhibit a wide range of behaviours, ecological adaptations, and social structures. Previous studies have mainly examined relative brain size, but a comprehensive understanding of brain diversity requires the investigation of other aspects of their neuroanatomy. Here, we obtained primarily post-mortem brain scans from eighteen species of the order Carnivora, reconstructed their cortical surfaces, and examined neocortical sulcal anatomy to establish a framework for systematic inter-species comparisons. We observed distinct regional variations in sulcal anatomy, potentially related to the species’ behaviour and ecology. Arctoidea species with pronounced forepaw dexterity exhibited complex sulcal configurations in the presumed somatosensory cortex but low sulcal complexity in the presumed visual and auditory occipitotemporal cortex. Canidae had the largest number of unique major sulci with a unique sulcus in the occipital cortex and highly social canids featuring an additional frontal cortex sulcus. We also observed differentially complex occipito-temporal sulcal patterns in Felidae and Canidae, indicative of changes in auditory and visual areas that may be related to foraging strategies and social behaviour. In conclusion, this study presents an inventory of the sulcal anatomy of a number of rarely studied carnivoran brains and establishes a framework and novel avenues for further investigations employing a variety of neuroimaging modalities to reveal more about carnivoran brain diversity.


Comparative neuroimaging of the carnivoran brain: Neocortical sulcal anatomy

October 2024

·

18 Reads

Carnivorans are an important study object for comparative neuroscience, as they exhibit a wide range of behaviours, ecological adaptations, and social structures. Previous studies have mainly examined relative brain size, but a comprehensive understanding of brain diversity requires the investigation of other aspects of their neuroanatomy. Here, we obtained primarily post-mortem brain scans from eighteen species of the order Carnivora, reconstructed their cortical surfaces, and examined neocortical sulcal anatomy to establish a framework for systematic inter-species comparisons. We observed distinct regional variations in sulcal anatomy, potentially related to the species’ behaviour and ecology. Arctoidea species with pronounced forepaw dexterity exhibited complex sulcal configurations in the presumed somatosensory cortex but low sulcal complexity in the presumed visual and auditory occipitotemporal cortex. Canidae had the largest number of unique major sulci with a unique sulcus in the occipital cortex and highly social canids featuring an additional frontal cortex sulcus. We also observed differentially complex occipito-temporal sulcal patterns in Felidae and Canidae, indicative of changes in auditory and visual areas that may be related to foraging strategies and social behaviour. In conclusion, this study presents an inventory of the sulcal anatomy of a number of rarely studied carnivoran brains and establishes a framework and novel avenues for further investigations employing a variety of neuroimaging modalities to reveal more about carnivoran brain diversity.


Comparative neuroimaging of the carnivoran brain: Neocortical sulcal anatomy

July 2024

·

281 Reads

·

1 Citation

Carnivorans are an important study object for comparative neuroscience, as they exhibit a wide range of behaviours, ecological adaptations, and social structures. Previous studies have mainly examined relative brain size, but a comprehensive understanding of brain diversity requires the investigation of other aspects of their neuroanatomy. Here, we obtained primarily post-mortem brain scans from eighteen species of the order Carnivora, reconstructed their cortical surfaces, and examined neocortical sulcal anatomy to establish a framework for systematic inter-species comparisons. We observed distinct regional variations in sulcal anatomy, potentially related to the species' behaviour and ecology. Arctoidea species with pronounced forepaw dexterity exhibited complex sulcal configurations in the presumed somatosensory cortex but low sulcal complexity in the presumed visual and auditory occipitotemporal cortex. Canidae had the largest number of unique major sulci with a unique sulcus in the occipital cortex and highly social canids featuring an additional frontal cortex sulcus. We also observed differentially complex occipito-temporal sulcal patterns in Felidae and Canidae, indicative of changes in auditory and visual areas that may be related to foraging strategies and social behaviour. In conclusion, this study presents an inventory of the sulcal anatomy of a number of rarely studied carnivoran brains and establishes a framework and novel avenues for further investigations employing a variety of neuroimaging modalities to reveal more about carnivoran brain diversity.


Figure 2. A: Parcellated time series (Schaefer atlas, 1000 parcels, 7 Yeo networks 31 ) of each subject were correlated to produce a 1000 x 1000 connectivity matrix. Principal component analysis (PCA) was applied to the thresholded matrix to extract 200 gradients. B: Variance explained by 200 gradients, mean across subjects ± 1 s.d.
Figure 3. Illustration of neighborhood dispersion. A: In a multidimensional gradient embedding, for a given region (red) K closest neighbors are identified (blue). These regions are shown within the black circle. B: Gradient (neighborhood) dispersion of a given region is the mean distance between said region and its K closest neighbors. The same operation is done for every region (N regions = 1000).
Figure 4. A: The types of predictors tested in this work (left to right): connectivity matrices (vectorized), macroscale cortical gradients, neighborhood, and centroid dispersion. B: All four types of features are concatenated together (2) and decomposed using group PCA (2) (each feature group is decomposed separately). The resulting dataset, along with covariates, was divided into the train and holdout set; 10-fold cross-validation (CV) was used to assess the performance of L2-regularized logistic regression on the PCA dataset (3). C: Permutation component importance was computed for each component using the holdout set (1). For each feature type, component importance was imverse transformed to obtain feature importance (2).
Figure 5. A. Permutation importance across feature types. B: Accuracy and F1 score across 13 classifiers (mean cross CV folds; the classifiers performing worse than the dummy classifier were excluded) fit on 936 best features from connectivity, the principal gradient, all gradients, neighborhood dispersion and the 28 values of centroid dispersion (light red). P-values
Demographic statistics of the sample. *These datasets are part of a larger dataset, SPRBS-1600. FD: framewise displacement.
A Comparative Machine Learning Study of Connectivity-Based Biomarkers of Schizophrenia

January 2024

·

253 Reads

·

1 Citation

Functional connectivity holds promise as a biomarker of psychiatric disorders. Yet, its high dimensionality, combined with small sample sizes in clinical research, increases the risk of overfitting when the aim is prediction. Recently, low-dimensional representations of the connectome such as macroscale cortical gradients and gradient dispersion have been proposed, with studies noting consistent gradient and dispersion differences in psychiatric conditions. However, it is unknown which of these derived measures has the highest predictive capacity and how they compare to raw connectivity. Our study evaluates which connectome features — functional connectivity, gradients, or gradient dispersion — best identify schizophrenia. Figure 1 summarizes this work. Surprisingly, our findings indicate that functional connectivity outperforms its low-dimensional derivatives such as cortical gradients and gradient dispersion in identifying schizophrenia. Additionally, we demonstrated that the edges which contribute the most to classification performance are the ones connecting primary sensory regions.



A map of white matter tracts in a lesser ape, the lar gibbon

October 2023

·

197 Reads

·

4 Citations

Brain Structure and Function

The recent development of methods for constructing directly comparable white matter atlases in primate brains from diffusion MRI allows us to probe specializations unique to humans, great apes, and other primate taxa. Here, we constructed the first white matter atlas of a lesser ape using an ex vivo diffusion-weighted scan of a brain from a young adult (5.5 years) male lar gibbon. We find that white matter architecture of the gibbon temporal lobe suggests specializations that are reminiscent of those previously reported for great apes, specifically, the expansion of the arcuate fasciculus and the inferior longitudinal fasciculus in the temporal lobe. Our findings suggest these white matter expansions into the temporal lobe were present in the last common ancestor to hominoids approximately 16 million years ago and were further modified in the great ape and human lineages. White matter atlases provide a useful resource for identifying neuroanatomical differences and similarities between humans and other primate species and provide insight into the evolutionary variation and stasis of brain organization.


Variation in spatial dependencies across the cortical mantle discriminates the functional behaviour of primary and association cortex

September 2023

·

223 Reads

·

13 Citations

Recent theories of cortical organisation suggest features of function emerge from the spatial arrangement of brain regions. For example, association cortex is located furthest from systems involved in action and perception. Association cortex is also ‘interdigitated’ with adjacent regions having different patterns of functional connectivity. It is assumed that topographic properties, such as distance between regions, constrains their functions, however, we lack a formal description of how this occurs. Here we use variograms, a quantification of spatial autocorrelation, to profile how function changes with the distance between cortical regions. We find function changes with distance more gradually within sensory-motor cortex than association cortex. Importantly, systems within the same type of cortex (e.g., fronto-parietal and default mode networks) have similar profiles. Primary and association cortex, therefore, are differentiated by how function changes over space, emphasising the value of topographical features of a region when estimating its contribution to cognition and behaviour.


Figure 3. Variograms calculated for each canonical resting state network (Yeo, Krienen et al., 2011) in humans and in homolog networks in macaques. The middle panel shows the mean variogram (FC dissimilarity by distance along the cortex) calculated across all vertices for each Yeo network in the human Human Connectome Project data; the filled areas are the standard errors of the mean across vertices. The bottom panel shows a similar analysis with fMRI data averaged from 14 awake Macaque monkey as a comparison.
Variation in spatial dependencies across the cortical mantle discriminates the functional behaviour of primary and association cortex

January 2023

·

149 Reads

Recent theories of cortical organisation maintain that important features of brain function emerge through the spatial arrangement of regions of cortex. For example, areas of association cortex are located in regions of cortex furthest from sensory and motor cortex. Association cortex is also ‘interdigitated’ since adjacent regions can have relatively different patterns of functional connectivity. It is assumed that topographic properties such as distance between cortical regions constrain their functions. For example, large distances between association and sensory and motor systems may enable these areas of cortex to maintain differentiable neural patterns, while an interdigitated organisation may enable association cortex to contain many functional systems in a relatively compact space. We currently lack a formal understanding of how spatial organisation impacts brain function, limiting the ability to leverage cortical topography to facilitate better interpretations of a regions function. Here we use variograms, a quantification of spatial autocorrelation, to develop a cortex-wide profile of how functional similarity changes as a function of the distance between regions. We establish that function changes gradually within sensory and motor cortex as the distance between regions increases, while in association cortex function changes rapidly over shorter distances. Subsequent analysis suggests these differential classes of spatial dependency are related to variation in intracortical myelin between sensory motor and association cortex. Our study suggests primary and association cortex are differentiated by the degree to which function varies over space, emphasising the need to formally account for spatial properties when estimating a system’s contribution to cognition and behaviour. Significance statement The spatial arrangements of regions in the human brain are hypothesised to underpin important features of a brain regions function. Currently, however, we lack a formal understanding of how topography shapes brain function, limiting our ability to leverage topographical perspectives to inform better theories of brain function. Here we use a formal mathematical approach to establish that in regions of association cortex function varies across the cortex more rapidly than in sensory and motor cortex, a phenomenon linked to levels of intracortical myelin. This result highlights how topographical features distinguish between cortical regions with different functional profiles and provides a formal account of how spatial differences support different features of brain function.


Citations (9)


... This work may provide more instances of methods to perturbate complex functional imaging data and couple them directly to behavior and external stimuli. In terms of cognitive representation and sensory integration, a recent paper (Wei Wei, et al 2024) demonstrates use of the perturbative approach to describe a 2 dimensional framework of sensory integration along the cortical hierarchy in functional magnetic resonance imaging (fmri) and suggests a role of perturbative approaches in integrating different sensory domains in cognitive processing. Similar approaches for the locus coeruleus specifically are lacking, but recent work demonstrated a representation of learned behavior via network state changes in the sensory thalamus (Masashi Hasegawa, et al 2024), the network driven changes of the locus coeruleus are apparent, and the recent work (Charles Rodenkirch, et al 2018) Locus coeruleus and cortical processing shifts, current findings and future trends. ...

Reference:

Integrating the LC into cortical computation. Current trends, challenges, and future directions.
A function-based mapping of sensory integration along the cortical hierarchy

Communications Biology

... Investigations of regional variations in brain size are still scarce within the carnivoran lineage. First observations indicate cortical expansion and more complex sulcal configurations within the parietal cortex in carnivoran species exhibiting manual object manipulation behaviours-such as raccoons, red pandas, and other species with high forepaw dexterity-compared with species with low forepaw dexterity, such as canids or felids ( Boch, Karadachka, et al., 2024;Manger et al., 2002). However, further research is needed to explore relative lobe expansion and contributing factors in carnivorans. ...

Comparative neuroimaging of the carnivoran brain: Neocortical sulcal anatomy

... The samples were acquired from deceased animals that died of causes unrelated to this research project. Details of sample preparation, scanning procedure, and data preprocessing have been described previously 46,47 . In brief, all brains were extracted and fixed within 24 h after the death of the animal. ...

A map of white matter tracts in a lesser ape, the lar gibbon

Brain Structure and Function

... For instance, sensory and motor functions are arranged as topological maps 5 and processing streams 6,7 . However, an increasing number of studies 2,[8][9][10][11][12] have highlighted the possibility that geometric constraints may provide a set of general principles which could explain the rich neural dynamics that are observed across the cerebral cortex and would therefore help constrain how these neural processes support different cognitive functions. ...

Variation in spatial dependencies across the cortical mantle discriminates the functional behaviour of primary and association cortex

... Neurodesk is also impactful as an educational tool in workshops and courses. The platform was first conceptualized during a 'hackathon' 25 , an event where people with diverse skill sets collaborated on projects and developed research skills. Variability in analysis environments across attendees' computers presents a hurdle for neuroimaging training workshops such as this. ...

Brainhack: Developing a culture of open, inclusive, community-driven neuroscience
  • Citing Article
  • June 2021

Neuron

... However, published adaptations so far seem to still be limited to cortices less gyrified then cetaceans and pinnipeds (Roumazeilles et al., 2022). Thus, currently, FS and FSL-derived methods are not able to process all the diversity of brain morphologies present in nature. ...

Cortical Morphology and White Matter Tractography of Three Phylogenetically Distant Primates: Evidence for a Simian Elaboration

Cerebral Cortex

... Some researchers have addressed this issue by leveraging human datasets for studies involving animals. For example, Wang et al. performed a pretraining process using the human brain dataset and then performed transfer training on a small amount of chimpanzee brain data to achieve segmentation of the chimpanzee brain and skull [20]. Similarly, Wang et al. trained a ResTLU-Net network using a human brain dataset and then applied migration learning to a small Macaque dataset to extract Macaque brain tissue [21]. ...

U-Net Model for Brain Extraction: Trained on Humans for Transfer to Non-human Primates

NeuroImage

... Various methods have been applied to reveal functional connectivity, including model-free methods such as sparse dictionary learning (sDL) 5,7,8 and independent component analysis (ICA), 9 as well as seed-based methods 10 and functional connectome-based cross-subject analysis. 11,12 As pigs require anesthesia during MRI scanning to reduce motion artifacts, task-based fMRI meets many challenges. Existing resting-state fMRI (rs-fMRI) methods mostly use the spatial correlation between a known atlas and generated activation maps 5 to evaluate recovery of brain functions after TBI, whereas temporal correlation (e.g., the FSLNets 9 ) has been used for internetwork comparisons within a group. ...

A Pig White Matter Atlas and Common Connectivity Space Provide a Roadmap for the Introduction of a New Animal Model in Translational Neuroscience

... Finally, detailed comparative anatomy of human and swine brain is lacking, and the electrodes placement may not have captured the activity of the primary visual cortex. Yet, published anatomical studies support the positioning adopted in the present study [23,24]. ...

A Pig White Matter Atlas and Common Connectivity Space Provide a Roadmap for the Introduction of a New Animal Model in Translational Neuroscience