CoChat
Sign Up
Cited by 8Advanced Neuroimaging Techniques and ApplicationsFunctional Brain Connectivity StudiesComplex Network Analysis Techniques

The role of spatial embedding in mouse brain networks constructed from diffusion tractography and tracer injections

Open access
Trinkle, Scott, Foxley, Sean, Wildenberg, Gregg, Kasthuri, Narayanan, La Rivière, Patrick·2021·NeuroImage
DOI 10.1016/j.neuroimage.2021.118576PMC PMC8611903
Open this paper in CoChat — free accountOpen PDF

Abstract

Diffusion MRI tractography is the only noninvasive method to measure the structural connectome in humans. However, recent validation studies have revealed limitations of modern tractography approaches, which lead to significant mistracking caused in part by local uncertainties in fiber orientations that accumulate to produce larger errors for longer streamlines. Characterizing the role of this length bias in tractography is complicated by the true underlying contribution of spatial embedding to brain topology. In this work, we compare graphs constructed with ex vivo tractography data in mice and neural tracer data from the Allen Mouse Brain Connectivity Atlas to random geometric surrogate graphs which preserve the low-order distance effects from each modality in order to quantify the role of geometry in various network properties. We find that geometry plays a substantially larger role in determining the topology of graphs produced by tractography than graphs produced by tracers. Tractography underestimates weights at long distances compared to neural tracers, which leads tractography to place network hubs close to the geometric center of the brain, as do corresponding tractography-derived random geometric surrogates, while tracer graphs place hubs further into peripheral areas of the cortex. We also explore the role of spatial embedding in modular structure, network efficiency and other topological measures in both modalities. Throughout, we compare the use of two different tractography streamline node assignment strategies and find that the overall differences between tractography approaches are small relative to the differences between tractography- and tracer-derived graphs. These analyses help quantify geometric biases inherent to tractography and promote the use of geometric benchmarking in future tractography validation efforts.

AI snapshot

Written from the paper’s full text.

TL;DR

This paper benchmarks mouse brain graphs from ex vivo diffusion MRI tractography against Allen tracer-derived graphs using random geometric surrogate graphs. It finds that spatial embedding/geometry explains substantially more of tractography network topology than tracer network topology, largely because tractography underestimates long-range connectivity.

Key findings

  • Geometry plays a substantially larger role in tractography-derived graphs than in tracer-derived graphs. Tractography underestimates relative long-range connectivity by as much as two orders of magnitude, with statistically significant differences from tracers for fiber distances around 17–22 mm.
  • Tracer graphs place hubs peripherally: 17% of total tracer hub strength was in the isocortex, while no isocortex nodes in either tractography graph were identified as hubs; only 11 individual structures were co-identified as hubs between the tracer and both tractography methods. Tractography hubs were closer to the geometric center of the brain.
  • Raw modularity Q was comparable between tracer and tractography graphs, but the tracer graph showed substantially higher Wrand-normalized modularity. Tractography Q values normalized by geometric surrogates were significantly closer to 1 than the tracer graph, indicating more geometric determination of modular structure.
  • Endpoint tractography matched tracer global efficiency across thresholds, while dense tractography significantly underestimated global efficiency at all densities. Neither tractography method accurately predicted per-node local efficiency.

Methods

  • Compared ex vivo diffusion MRI tractography in five adult mice (9.4 T, 125 μm isotropic, b-value = 3000 s/mm², 30 directions) with the Allen Mouse Brain Connectivity Atlas tracer model from Knox et al. (428 anterograde fluorescent viral tracer experiments; 286 of 291 gray matter regions) and independent cortical tracer data from Gămănuţ et al.
  • Constructed tractography graphs using two streamline node assignment strategies: “endpoint” and “dense,” with weights normalized by node-pair volumes and hemispheric symmetry enforced.
  • Built random geometric surrogate graphs preserving node-strength distribution and low-order weight-distance relationships, plus random surrogates preserving the strength sequence, then compared modularity, hub organization, participation coefficient, eigenvector centrality, and network efficiency.

Limitations

  • Stated: the analysis assumes tracer data as ground truth, but tracer graphs are inherently directed while tractography is undirected, requiring symmetry enforcement that may alter network properties; the Knox et al. model is itself a computational estimate that may carry biases; and the Gămănuţ et al. validation data cover only select ipsilateral intra-cortical edges.
  • Stated: tractography used a single, relatively low b-value and did not use advanced streamline quantitation such as SIFT2.
  • Evident from text: the diffusion MRI sample is small (five mice), and the paper notes that human tractography conclusions cannot be verified with ground-truth imaging, so caution is warranted when generalizing.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.neuroimage.2021.118576

Cites

56 of 56 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

20 of 33 cited sources checked · 9 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Modern tractography approaches still produce high numbers of both false positive and false negative tracts.

    “Recent efforts to validate tractography have uncovered several limitations, and even modern approaches still produce high numbers of both false positive and false negative tracts (Aydogan et al., 2018; Maier-Hein et al., 2017; Schilling et al., 2019; Thomas et al., 2014).”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites Aydogan et al., 2018 source
    • cites Maier-Hein et al., 2017 source

      The Results section, “Tractograms contained more invalid than valid bundles” (page unavailable), reports many false-positive bundles: 88 on average. But “Tractograms contained most of the ground truth bundles” reports that an average of 21 of 25 true bundles were identified. Although their spatial extent was often poorly recovered, the paper does not establish high numbers of false-negative tracts. (LLM, conf=0.87)

      Experimental — accuracy not yet measured
    • cites Schilling et al., 2019 source
  • Premisecitation

    A specific limitation inherent to probabilistic streamline tractography is the fiber-length bias: local uncertainties in the fiber orientation distribution at each step in the tracking process accumulate to produce global errors for longer streamlines, leading to an underestimation of connectivity weights at greater distances.

    “A specific limitation inherent to probabilistic streamline tractography is the fiber-length bias: local uncertainties in the fiber orientation distribution at each step in the tracking process accumulate to produce global errors for longer streamlines, leading to an underestimation of connectivity weights at greater distances.”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites Aydogan et al., 2018 source

      Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites Maier-Hein et al., 2017 source

      Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites Schilling et al., 2019 source

      Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
  • Premisecitation

    There is a true falloff of both structural and functional connectivity with distance in a number of mammalian species.

    “true falloff of both structural and functional connectivity with distance has been shown with high-resolution invasive imaging as well as tractography in a number of mammalian species (Horvát et al., 2016; Markov et al., 2013; Perinelli et al., 2019; Roberts et al., 2016; Rubinov et al., 2015; Supekar et al., 2009).”
    Quote in the paperSubstantial claim2 of 3 sources checked
    • cites Horvát et al., 2016 source

      The Results (‘Functional Layout in Terms of Spatial Clustering of Cortical Areas’; page unavailable) report that connection probability and strength decline with distance in mouse and macaque, using tract-tracing data. The paper does not show a corresponding falloff in functional connectivity; ‘functional layout’ refers here to the organization of the traced cortical network. The Discussion (page unavailable) describes limitations of tractography rather than presenting it as a method used to establish these findings. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites Markov et al., 2013 source

      The Abstract and Introduction report that structural corticocortical connection weights and connection probability fall with distance in macaques, using retrograde-tracing data. Our copy does not show a comparable falloff in functional connectivity, demonstrate it across multiple mammalian species, or report tractography evidence. (LLM, conf=0.89)

      Experimental — accuracy not yet measured
    • cites Roberts et al., 2016 source
  • Premisecitation

    The brain is a spatially embedded network with properties partially inherited from geometry.

    “the brain is itself a spatially embedded network with properties partially inherited from geometry: there is a metabolic wiring cost in the establishment of long-range connections”
    Quote in the paperSubstantial claim2 of 3 sources checked
    • cites Horvát et al., 2016 source
    • cites Markov et al., 2013 source
    • cites Roberts et al., 2016 source
  • Premisecitation

    Neural tracer data is the gold standard for mesoscale connectivity studies in a number of model organisms.

    “Neural tracer data is the gold standard for mesoscale connectivity studies in a number of model organisms (Bota et al., 2015; Girard et al., 2020; Jbabdi et al., 2015; Markov et al., 2014; 2013; Oh et al., 2014).”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites Bota et al., 2015 source

      The Abstract says animal histological pathway-tracing data are needed to validate diffusion tractography, and Materials and Methods describes using tracing reports for a rat cortical connectome. But the paper studies regional macroconnections, while the Discussion describes a neuron-type mesoconnectome as future work; it does not establish tracer data as the gold standard for mesoscale studies across multiple model organisms. Page numbers are unavailable in our copy. (LLM, conf=0.85)

      Experimental — accuracy not yet measured
    • cites Girard et al., 2020 source
    • cites Jbabdi et al., 2015 source
  • Classified as probably not a substantial claim. 8
  • May combine two claims. 18
  • The quote only partly states this claim. 1

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Recent tractography validation reviews suggest that the future of tractography connectomics hinges on incorporation of more advanced anatomical and microstructural priors to address geometric and other biases and make streamlines more quantitative and physically meaningful.

    “Even with these relatively simple post-processing approaches, the results of this study serve to echo recent tractography validation reviews that suggest that the future of tractography connectomics hinges on the incorporation of more advanced anatomical and microstructural priors to tractography pipelines in order to address geometric and other biases and make streamlines more quantitative and”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites Maier-Hein et al., 2017 source

      The Discussion (page not provided) calls for more reliable anatomical priors and advanced diffusion microstructure modeling, and says streamline filtering may reduce tractography biases. It does not say these changes will make individual streamlines more quantitative or physically meaningful; it also cautions that such developments will not resolve tractography’s fundamental ambiguities. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites Schilling et al., 2019 source

      The Discussion (‘Advancements needed’; page unavailable) recommends anatomical priors and microstructural information to improve tracking, and the Introduction identifies geometric and other sources of bias. Our copy does not establish the further claim that these additions would make streamlines more quantitative and physically meaningful. (LLM, conf=0.82)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Girard et al. (2020) rigorously benchmarked 15 tractography algorithms and regularization approaches against tracer data in macaque cortex, but did not present analysis of downstream network measures.

    “A recent study by Girard et al. (2020) rigorously benchmarked 15 tractography algorithms and a number of regularization approaches such as the “anatomically constrained tractography” framework (Smith et al., 2012) and the SIFT2 post-processing algorithm (Smith et al., 2015) against tracer data in the macaque cortex, though it did not present analysis of downstream network measures.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Girard et al. (2020) source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Opting against SIFT2 provides a more direct comparison between our results and similar studies that used Allen tracer data without reference to spatial embedding.

    “This also provides a more direct comparison between our results and those of similar studies that have used the Allen tracer data to benchmark tractography performance without reference to the role of spatial embedding (Aydogan et al., 2018; Calabrese et al., 2015; Chen et al., 2015).”
    Quote in the paperNot a substantial claim3 of 3 sources checked
    • cites Aydogan et al., 2018 source
    • cites Calabrese et al., 2015 source
    • cites Chen et al., 2015 source

      The Introduction and Methods §2.1 identify the Allen Mouse Brain Connectivity Atlas tracer data as the benchmark for evaluating DTI tractography. But Results §3.1 examines how streamline length affects accuracy, and §3.3 compares performance across anatomical locations. The description of the study as proceeding without reference to spatial factors is therefore too broad; the paper does not address the citing authors’ decision against SIFT2. (LLM, conf=0.82)

      Experimental — accuracy not yet measured

  • The quote only partly states this claim.

    Findingown data

    Both tractography methods show values significantly closer to 1 than the tracer graph does for empirical Q normalized by geometric surrogate Q.

    “Both tractography methods show values significantly closer to 1 than the tracer graph does.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

  • May combine two claims.

    Findingown data

    Both tractography methods show only weak correlation with the participation coefficient values from the corresponding tracer graph, but significantly stronger correlations with the values from their own geometric surrogates.

    “Both tractography methods show only weak correlation with the values from the corresponding tracer graph, but significantly stronger correlations with the values from their own geometric surrogates.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

  • May combine two claims.

    Findingown data

    Endpoint tractography graphs place their hub nodes more geometrically central overall than the tracer graph, with distances from the center of mass comparable to their geometric surrogates.

    “Endpoint tractography graphs also place their hub nodes more geometrically central overall than the tracer graph, with distances from the center of mass comparable to their geometric surrogates.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

  • May combine two claims.

    Findingown data

    The differences in standard deviation between tractography and tracers were only statistically significant between around 3-15 mm for endpoint node assignment and for a smaller range around 4-6 mm for dense node assignment.

    “though these differences were only statistically significant between around 3–15 mm for endpoint node assignment and for a smaller range around 4–6 mm for dense node assignment.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    17% of the total tracer hub strength was located along the isocortex, while no isocortex nodes in either of the tractography graphs were identified as hubs.

    “17% of the total tracer hub strength was located along the isocortex, while no isocortex nodes in either of the tractography graphs were identified as hubs.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Dense tractography places its hubs closer to the center of the brain than the tracer graph does, and also places its hubs slightly more central than its own geometric surrogates.

    “dense tractography not only places its hubs closer to the center of the brain than the tracer graph does, but also places its hubs slightly more central than its own geometric surrogates.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

  • May combine two claims.

    Findingown data

    Dense node assignment led to higher Spearman correlations than endpoints for nearly all connections to the pons, and for connections between the hypothalamus, midbrain, and medulla.

    “Dense node assignment led to higher Spearman correlations than endpoints for nearly all connections to the pons, and for connections between the hypothalamus, midbrain, and medulla.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Empirical tractography graphs place the strongest nodes deeper towards the center of the brain compared to the tracer graph, with hubs organized into a ball-like structure comparable to their corresponding geometric surrogates.

    “Empirical tractography graphs, however, place the strongest nodes deeper towards the center of the brain compared to the tracer graph, with hubs organized into a ball-like structure comparable to their corresponding geometric surrogates.”
    • Also stated in: § Abstract · plain
      “tractography to place network hubs close to the geometric center of the brain, as do corresponding tractography-derived random geometric surrogates”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

  • May combine two claims.

    Findingown data

    The mouse brain network as measured with tracer data places its hub nodes further towards the periphery of the brain than would be predicted by geometry alone, with the strongest nodes located across the isocortex, medulla, and inferior hypothalamus.

    “we find the expected result that the mouse brain network as measured with tracer data places its hub nodes further towards the periphery of the brain than would be predicted by geometry alone, with the strongest nodes located across the isocortex, medulla, and inferior hypothalamus.”
    • Also stated in: § Abstract · plain
      “tracer graphs place hubs further into peripheral areas of the cortex”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

  • May combine two claims.

    Findingown data

    All tractography methods showed relatively high correlations in ipsilateral intra-division connectivity and weaker contralateral connectivity to homologous regions.

    “All tractography methods showed relatively high correlations in ipsilateral intra-division connectivity (diagonals in Fig. 1e–f), and weaker contralateral connectivity to homologous regions”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Tractography graphs underestimated hub strength in the medulla and overestimated hub strength in the midbrain and thalamus, particularly for the dense node assignment strategy.

    “Tractography graphs likewise underestimated hub strength in the medulla and overestimated hub strength in the midbrain and thalamus, particularly for the dense node assignment strategy.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Diffusion MRI tractography is currently the only noninvasive method for mapping the human structural connectome.

    “Diffusion MRI tractography is the only noninvasive method to measure the structural connectome in humans.”
    • Also stated in: § Introduction · plain
      “Diffusion MRI tractography is currently the only noninvasive method for mapping the human structural connectome (Yeh et al., 2020)”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Yeh et al., 2020

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationown data

    Modular structure in the mouse structural brain network is both underestimated overall and more geometrically determined in tractography relative to neural tracer imaging.

    “Overall, these results suggest that modular structure in the mouse structural brain network is both underestimated overall and more geometrically determined in tractography relative to neural tracer imaging.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.
    • The authors state this as an assumption: Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.
    • The authors state this as an assumption: The random surrogates Wrand represent the null hypothesis that topological properties of brain networks arise solely from the particular distribution of node strengths and locations.

  • May combine two claims.

    Interpretationreasoning

    This conflation biases the estimated modular structure and the architecture of hub subnetworks.

    “Tractography graphs underestimate connectivity weights at long distances, leading to a conflation of topological and geometric centrality that biases the estimated modular structure and the architecture of hub subnetworks.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The tracer data is a ground truth representation of the underlying mesoscale mouse brain network architecture.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Empirical human tractography graphs exhibit a more peripheral hub network structure than predicted by their corresponding geometric surrogate graphs.

    “For example, Roberts et al. (2016) demonstrated that empirical human tractography graphs exhibit a more peripheral hub network structure than predicted by their corresponding geometric surrogate graphs, similar to our tracer results in Fig. 6 and in contrast to what we observed with tractography networks.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Roberts et al. (2016) source

  • May combine two claims.

    Premisenone given

    The Knox et al. model has been shown to outperform the homogeneous model originally published in Oh et al. (2014) and produces a graph that is nearly 100% fully connected at the scale of the parcellation used in this study.

    “The Knox et al. model has been shown to outperform the homogeneous model originally published in Oh et al. (2014) and produces a graph that is nearly 100% fully connected at the scale of the parcellation used in this study.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Oh et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    The Knox et al. model provides estimates of connectivity at the scale of 100 μm isotropic voxels in the Allen Mouse Brain Common Coordinate Framework (CCF v3), and is derived from 428 whole-brain anterograde fluorescent viral tracer experiments in wild-type C57BL/6J mice.

    “The Knox et al. model provides estimates of connectivity at the scale of 100 μm isotropic voxels in the Allen Mouse Brain Common Coordinate Framework (CCF v3), and is derived from 428 whole-brain anterograde fluorescent viral tracer experiments in wild-type C57BL/6J mice collected for the Allen Mouse Brain Connectivity Atlas (Oh et al., 2014).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites Knox et al.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Oh et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The human brain places its strongest nodes further into geometrically peripheral regions than would be expected by weight-distance effects alone.

    “Previous work with human tractography datasets (Roberts et al., 2016) has revealed that the human brain places its strongest nodes further into geometrically peripheral regions than would be expected by weight-distance effects alone.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Roberts et al., 2016 source

  • May combine two claims.

    Study design and conductreasoning

    The physical interpretation of edge weights differs between modalities, so direct comparison of the edge-weight means between modalities is not meaningful.

    “Note that the physical interpretation of edge weights differs between modalities, so direct comparison of the edge-weight means between modalities is not meaningful.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Data from the Allen Institute Mouse Brain Connectivity Atlas form the most comprehensive whole-brain mesoscale brain network in a mammalian species.

    “Data from the Allen Institute Mouse Brain Connectivity atlas (Oh et al., 2014) form the most comprehensive whole-brain mesoscale brain network in a mammalian species”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Oh et al., 2014

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Study design and conductown data

    Transformed weights show effectively zero correlation with fiber distance using both Pearson and Spearman correlation coefficients, indicating that low-order distance relationships have been effectively removed prior to randomization for the construction of the geometric surrogate graphs.

    “Transformed weights (Figure S2c) show effectively zero correlation with fiber distance using both Pearson and Spearman correlation coefficients, indicating that low-order distance relationships have been effectively removed prior to randomization for the construction of the geometric surrogate graphs.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Enforcing constraints on streamline termination penalizes unrealistic fiber geometries or lower-confidence diffusion data but does not yield a physical sense of the actual origin or termination points of underlying neuronal fibers.

    “When streamlines are made to terminate under reasonably enforced constraints such as on fODF magnitude or streamline curvature, this enforces a penalty on unrealistic fiber geometries or the use of lower-confidence diffusion data, but does not yield a physical sense of the actual origin or termination points of the underlying neuronal fibers.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Normalizing streamline counts by volumes of each node-pair better matches the normalized connection density metric used in the Knox et al. model.

    “For the data used in this study, we took the approach of seeding only from the gray matter to avoid known tract-length connectivity biases (Girard et al., 2014; Jeurissen et al., 2019), and normalized the streamline counts by the volumes of each node-pair in order to better match the normalized connection density metric used in the Knox et al. model.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The tracer data is a ground truth representation of the underlying mesoscale mouse brain network architecture.
26 of 26

  • The authors state this as an assumption.

    Network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.

    “We make the assumption that network properties that are preserved in the geometric surrogate graphs represent those that have been inherited from the spatial embedding of the brain.”

  • The authors state this as an assumption.

    Differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.

    “Likewise, we assume that differences in network properties between empirical and geometric surrogate graphs represent the extent to which those properties arise from other, non-geometric factors.”

  • The authors state this as an assumption.

    The geometric surrogates Wgeo represent the null hypothesis that topological properties of brain networks arise from the falloff of edge weights with distance for a given node-strength distribution.

    “The geometric surrogates Wgeo represent the null hypothesis that topological properties of brain networks arise from the falloff of edge weights with distance for a given node-strength distribution”

  • The authors state this as an assumption.

    The random surrogates Wrand represent the null hypothesis that topological properties of brain networks arise solely from the particular distribution of node strengths and locations.

    “while the random surrogates Wrand represent the null hypothesis that topological properties of brain networks arise solely from the particular distribution of node strengths and locations.”

  • The authors state this as an assumption.

    The tracer-derived network is a significantly higher-fidelity representation of the true mesoscale connectome of the mouse brain.

    “taking the tracer-derived network as a significantly higher-fidelity representation of the true mesoscale connectome of the mouse brain.”

  • The authors state this as an assumption.

    The tracer data is a ground truth representation of the underlying mesoscale mouse brain network architecture.

    “The results in this study rely on the assumption of the tracer data as a ground truth representation of the underlying mesoscale mouse brain network architecture.”
6 of 6

Cites

Flagged first, then by citations

3 of the 59 references couldn’t be matched to a known paper.

Research works printed but not linked: 1

  1. Tournier, 2010, Improved probabilistic streamlines tractography by 2nd order integration over fibre orientation distributions, Proc. Int. Soc.Magn. Reson. Med., 1670

Cited by

Sorted by citations

Related

Authors
Trinkle, S., Foxley, S., Wildenberg, G., Kasthuri, N., & La Rivière, P.
Affiliations
University of Chicago (US)
Corresponding author
Trinkle, Scott
ORCID
Venue
NeuroImage
Publisher
Elsevier BV
Published
2021-12
Volume
244
Pages
118576
ISSN
1053-8119 (ISSN-L)
DOI
10.1016/j.neuroimage.2021.118576
PMC
PMC8611903
OpenAlex
W3200881363
Cited by
8
References
59
Citations by year
2026: 1 · 2025: 2 · 2023: 2 · 2022: 2 · 2021: 1
FWCI
0.61
Type
journal article
Language
en
Open access
Licence
Locations
  • NeuroImage (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion, cc-by)
  • Knowledge@UChicago (University of Chicago) (submittedVersion)
  • Knowledge@UChicago (University of Chicago) (submittedVersion, cc-by-nc-nd)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • University of Chicago
  • University of Chicago (cc-by)
Funders
National Institutes of Health (F31NS113571, P30CA14599, R01EB026300, S10OD025081, S10RR021039, U01MH109100)
Topics
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Advanced Neuroimaging Techniques and Applications
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Functional Brain Connectivity Studies
  • Physical Sciences › Physics and Astronomy › Statistical and Nonlinear Physics › Complex Network Analysis Techniques
Keywords
spatial embedding, diffusion tractography, mouse brain, random geometric graphs, network hubs, network modularity, network efficiency, Allen Mouse Brain Connectivity Atlas
MeSH terms
Animals, Brain, Cerebral Cortex, Mice, Diffusion Tensor Imaging, Connectome

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

Cited sources are checked by gpt-6-sol.

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