{"id":"e19d29a9-3cbc-4fac-b8da-71cf2d8d0f1f","arxiv_id":"2501.02100","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A large-scale morphometric analysis of 11,850 brain cell reconstructions provides reference values for structural, shape, and topological features across species and cell types, with guidance for diffusion MRI modeling.","lead":"This paper measures the shapes of more than 11,000 brain cells from mice, rats, monkeys, and humans, and publishes reference values for the size, branching, and topology of nine cell types. The goal is to give diffusion MRI researchers a biologically grounded checklist of which cellular features their scans can actually see, improving how brain images are interpreted.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Soma radius threshold is only weakly validated and the paper's own '<20%' uncertainty claim is unsupported; soma-derived reference values may carry unquantified bias.","rationale":"The reader identified the same weakest assumption: the soma boundary definition via the nominal SWC radius, with the acknowledged inaccuracy in Fig. 2 and no quantification of bias. My stress-test agrees and sharpens the consequence: because soma-derived metrics directly determine the soma-restriction diffusion-time thresholds and residence times used in the modeling implications (Section 4.2), a 20% error in soma radius leads to ~40% error in the td threshold (quadratic scaling), and S/Vsoma is even more sensitive. The central claim is that the reference values are reliable enough to be an empirical foundation for dMRI modeling, so this unquantified bias is load-bearing. The paper's own caveat in Table 2 and Section 4.3 that values may be under-estimated by <20% is stated without a validation study or error analysis, so the concern is not resolved. A perturbation analysis or independent soma segmentation would settle it. The verdict should remain CONDITIONAL because this is addressable and does not invalidate the other contributions (reference distributions for branch metrics, shape descriptors, topology, and meshes), but it does mean the soma-related reference values should not be used as-is without added caveats or validation. The reader's verdict of CONDITIONAL is appropriate; no change in verdict is needed, but the condition should explicitly include the soma-boundary validation.","tokens_in":23517,"tokens_out":1977,"duration_ms":17548,"concrete_test":"Take a subset of the analyzed reconstructions (e.g., 20-50 cells spanning each cell type and species) for which the original microscopy images or an independent soma segmentation exists (or can be obtained from NeuroMorpho's associated image data), and compare the SWC-derived soma radius used in this paper against the soma radius/volume obtained from a gold-standard 3D soma surface segmentation. Alternatively, perform a perturbation analysis: re-run the soma metric computation with the soma threshold set to 0.8x and 1.2x the nominal SWC radius, and report the resulting percent change in Rsoma, S/Vsoma, and RM Rsoma for each cell type. If the median shift exceeds the claimed '<20%' uncertainty, or if the soma restriction time thresholds change by more than ~40%, then the reference values for soma-related metrics are not adequately validated for use in dMRI modeling.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The most load-bearing concern is the definition of the soma boundary. All soma metrics (Rsoma, S/Vsoma, RM Rsoma, ηsoma) and all derived soma residence/exchange times depend on assigning nodes within the nominal soma radius (radius of the first SWC node from the root) to the soma (Section 2.2). The authors explicitly acknowledge in Fig. 2 that this soma surface definition is 'slightly inaccurate' and state in Section 4.3 that soma volume is expected to be underestimated by <20%, but no validation of this bound is provided. SWC reconstructions use the root node's diameter as the soma radius, which is a known crude proxy for soma extent: the root node often represents the soma center, and its listed diameter is not a reliable measure of the 3D soma boundary, especially for glial cells with irregular somata. The reported Rsoma values (median ~2.5-4.1 um for glia; ~6.7-8.3 um for neurons) feed directly into the soma restriction analysis (td ≥ R^2/5D, Section 4.2) and the 'RM Rsoma' imaging-relevant radius. If the true soma boundary differs from the nominal radius by as little as 20%, the effective soma radius changes by ~20% and the predicted diffusion-time threshold for soma restriction shifts by ~40% (since td scales as R^2). More importantly, S/Vsoma is very sensitive to the threshold, and S/Vsoma is what determines soma residence times used to argue about exchange (Section 4.2). The paper does not quantify this bias, does not compare against a gold-standard soma segmentation (e.g., from brightfield/confocal images or EM), and does not report sensitivity of the reference values to the threshold choice. This is the weakest point in the central claim that the reference values are reliable enough to inform dMRI modeling. The issue is acknowledged in the manuscript but not resolved, and the '<20%' figure appears to be an assumption rather than a measured validation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a large-scale morphometric analysis of 11,850 three-dimensional SWC neuronal and glial reconstructions from the NeuroMorpho database, spanning four species (mouse, rat, monkey, human) and nine cell types. It computes structural, shape, and topological descriptors, reports reference values (quartiles and distributions) for each cell type and species, and interprets these values in terms of their relevance for diffusion-weighted MRI (dMRI) microstructural modeling. The authors derive example predictions for soma restriction, diffusion-mediated exchange, branch undulation, and permeative exchange times, and provide 50 high-resolution 3D surface meshes compatible with Monte Carlo simulators. The paper's central claim is that these reference values and meshes establish an empirical foundation for gray matter dMRI modeling and identify which neural features are detectable by dMRI.","tokens_in":23869,"tokens_out":3843,"duration_ms":37859,"significance":"The resource aspect of the paper is strong: a curated, quality-filtered set of 11,850 reconstructions with computed morphometric distributions, plus 50 surface meshes for simulation studies, is a valuable contribution that directly addresses a recognized gap in gray matter dMRI modeling. The authors are careful to distinguish measured quantities from literature-derived parameters (e.g., permeability, extracellular volume fraction), and they explicitly acknowledge several limitations, including the absence of spines and the crude soma surface definition. If the soma-segmentation concern is addressed, the paper will provide a useful benchmark dataset and a clear template for propagating morphometric measurements into dMRI model predictions. The topological persistence analysis and the demonstration that glial and neuronal topologies differ in rodents are also of interest. The code and data release plans further strengthen the contribution. However, the reliability of soma-related reference values is currently undermined by the unvalidated soma boundary definition, which propagates into soma size, surface-to-volume ratio, and derived exchange-time predictions.","major_comments":[{"comment":"The definition of the soma boundary is load-bearing for a large fraction of the reported reference values. In Section 2.2, all nodes within the nominal soma radius (radius of the first SWC node from the root) are assigned to the soma, and this threshold directly determines Rsoma, RM Rsoma, S/Vsoma, ηsoma, and the soma-derived residence and exchange times reported in Table 2 and Section 4.2. The paper acknowledges in Figure 2 that this soma surface definition is 'slightly inaccurate' and in Section 4.3 states that soma volume is expected to be underestimated by 'on average <20%', but no validation or sensitivity analysis is provided to support this bound. The nominal radius from the SWC root node is a well-known crude proxy for the true soma boundary, particularly for glial cells with irregular somata. Because the residence-time predictions in Section 4.2 scale with R_soma^2 (soma restriction) and R_soma^3 (soma residence time), an unquantified 20% error in soma radius would translate into substantially larger errors in the predicted diffusion-time thresholds. Please add a quantitative validation of the soma segmentation against an independent method on a subset of reconstructions, or at least a sensitivity analysis in which the threshold is varied and the resulting changes in soma metrics and derived predictions are reported.","section":"Section 2.2, Figure 2, Section 4.3, Table 2"},{"comment":"The illustrative dMRI predictions in Section 4.2 depend directly on the soma radius values. For example, the text states that soma restriction becomes measurable for td ≥ 2.5 ms for water using Rsoma ≈ 5 µm, and td ≥ 5 ms using RM Rsoma ≈ 7 µm. Since the criterion is 5D td ≥ R_soma^2, a 20% underestimation in soma radius would change the predicted td threshold by roughly a factor of 1.44 (i.e., ~40% shift). Similarly, the soma-to-projection exchange estimate uses τ_soma^i = π R_soma^3 / (3 R_branch sqrt(N_proj) D), which is cubic in Rsoma. The paper should propagate the uncertainty from the soma boundary definition into these example calculations so readers can see how robust the qualitative conclusions are.","section":"Section 4.2, 'Impact of soma restriction' and 'Impact of diffusion-mediated exchange between soma and projections'"},{"comment":"The computation of fractional anisotropy (FA) depends on two arbitrary or heuristic choices: the cylinder segment length (set to 10 µm) used to decompose each cell, and the 'adjusted FA' procedure that forces τ1 = τ2 to correct for depth-of-field anisotropy. The segment length is not justified, and no sensitivity analysis is reported for FA with respect to this length. The adjusted FA is a post-hoc correction that assumes the imaging artifact only affects the smallest eigenvalue; this is a strong assumption that is not validated. Since FA and adjusted FA are reported as reference values in Table 4 and used in the orientation-dispersion discussion in Section 4.2, the method-dependence of these values should be quantified or explicitly flagged as heuristic. At minimum, please report the sensitivity of FA to the segment length and justify the τ1 = τ2 adjustment with evidence from the reconstructions or a controlled phantom experiment.","section":"Section 2.3, Table 4, Figure 3"}],"minor_comments":[{"comment":"The number of analyzed reconstructions is inconsistent: the abstract states 11,500, while the introduction and Section 2.1 state 11,850. Please unify these numbers.","section":"Abstract and Section 1"},{"comment":"Table 2 is captioned as 'mean ± s.d.' but the columns clearly report quartiles (Q1, median, Q3) as indicated in the text. Either the caption or the table formatting should be corrected. In addition, the formula for RM Rbranch in Table 1 uses '<R^6_soma>' and '<R^2_soma>' instead of branch radii; this is a typographical error that should read '<R^6_branch>' and '<R^2_branch>'.","section":"Table 2, Table 1"},{"comment":"The FA formula uses τ (with an overbar) without defining it; the text should define it as the mean eigenvalue (τ1+τ2+τ3)/3 for clarity.","section":"Section 2.3"},{"comment":"There is a typo: 'metaboilites' should be 'metabolites'. Also, in the sentence 'these estimates become longer if we consider the effective MR radii RM Rsomaand RM Rbranch', a space is missing between 'RM Rsoma' and 'and'.","section":"Section 4.2"},{"comment":"The term 'overlab' in the definition of the topological distance D should be 'overlap'.","section":"Section 2.4"},{"comment":"Some cell-type/species combinations have very small sample sizes (e.g., N=4, N=6, N=11). While these are indicated in Table 2, the shape-descriptor table (Table 4) does not report sample sizes; adding them (or a reference to Table 2) would help readers judge the reliability of the reported FA and OD values.","section":"Section 3.4, Table 4"},{"comment":"The figure caption mentions 'arrows' in the top right corner, but the arrows are not visible in the figure as reproduced; please ensure the pointing elements are visible or rephrase the caption to describe the limitation directly.","section":"Section 2.2, Figure 2"}],"recommendation":"major_revision","confidential_remarks":"This is a resource-type paper that could be an important reference for the gray matter dMRI community. The central issue is the soma segmentation: it is acknowledged as approximate, and the claimed <20% underestimation is not supported by any validation or sensitivity analysis. This undermines a substantial subset of the reference values (soma size, S/Vsoma, residence and exchange times) and the example predictions in Section 4.2. The authors can address this with a focused validation study on a subset of reconstructions or a sensitivity analysis over the soma threshold. The FA adjustment and the cylinder segment length are secondary but still affect reported shape reference values. If these points are addressed, the paper would be a strong candidate for acceptance. I recommend major revision rather than rejection because the underlying measurements are well-described and the resource is valuable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a genuinely useful reference paper for grey matter dMRI, and the scale is new: ~11,800 reconstructions across nine cell types and several species, with structural, shape, and topological descriptors tied to diffusion modeling. The combination of TREES morphometrics, cylinder-decomposition FA, TMD persistence, and the residence/exchange calculations is not new per se, but nobody has put it together on this scale with an eye to dMRI model design. The reference tables and distributions will get cited.\n\nThe paper is careful in several ways: it states inclusion criteria, removes spines and truncated axons, reports quartiles not just means, and flags known limitations (tissue shrinkage, imaging artifacts, soma surface approximation). The discussion of what dMRI can and cannot see—soma restriction vs. exchange vs. branch undulation—is level-headed and uses the measured distributions to compute thresholds.\n\nThe soft spots are real but not fatal. The biggest is the soma boundary. Soma volume, Rsoma, S/Vsoma, and the derived residence/exchange times all depend on the SWC 'nominal soma radius' threshold. The paper admits the surface is 'slightly inaccurate' and asserts an expected underestimation of <20%, but that figure is not validated, and no sensitivity analysis is provided. For S/Vsoma in particular, a 20% radius error could shift derived exchange times by more than the paper implies. A simple threshold sweep or comparison against a few manually segmented somata would settle this. Second, the code, data, and 50 meshes are promised but not yet available; for a reference-value paper, that matters. Third, the abstract says 11,500 while the text says 11,850 and once 11,800—an editorial fix. Fourth, the 'adjusted FA' that sets tau1=tau2 to correct z-compression is a sensible hack but is presented without validation; a note on how much this changes conclusions would help.\n\nThe paper does not overclaim outrageously; the abstract's 'which cell types may be distinguishable' is backed by the topological distance analysis, and the authors include caveats about model interpretation. The citation pattern looks fair, including prior NeuroMorpho-based analyses, and the claims about novelty are modest.\n\nWho should read it: anyone building or fitting biophysical models of GM diffusion, especially SANDI/NEXI users. It deserves a serious referee. My recommendation: send it to review, but ask the authors to add a sensitivity analysis on the soma threshold and make the data available before publication.","headline":"A genuinely useful morphometric reference for grey matter dMRI modeling, held back by an unvalidated soma segmentation and a promised-but-missing data release, but worth sending to review.","tokens_in":24428,"tokens_out":2603,"would_cite":true,"duration_ms":26297,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Analysis of over 11,500 brain cell reconstructions maps which grey-matter features diffusion MRI can detect.","keywords":["diffusion MRI","grey matter","cell morphometry","biophysical modeling","morphological reference values","neuronal and glial cells","topological data analysis","3D cell meshes"],"falsifier":"Take a sample of the analyzed reconstructions and recompute every soma metric after shifting the soma boundary by plus or minus 20 percent, and where possible compare against high-resolution electron microscopy images of the same cells; if the resulting soma radii, MR radii, or residence times move by more than their reported interquartile ranges, the reference values are too threshold-sensitive to anchor diffusion models.","tokens_in":23362,"feed_emoji":"🧠","tokens_out":10879,"duration_ms":100948,"temperature":0.7,"pith_summary":"Diffusion MRI can sense micrometer-scale structure only indirectly, and grey matter has lacked the ground-truth cell morphologies needed to interpret the signal. This paper tries to close that gap by analyzing more than 11,500 three-dimensional reconstructions of nine brain cell types from mouse/rat, monkey, and human, and publishing reference ranges for the traits that matter: soma size, branch length, surface-to-volume ratios, shape anisotropy, and branching topology. From those numbers it derives time-scale estimates showing which traits a typical clinical diffusion experiment can detect: soma restriction and water exchange between projections and the space around them are measurable, while branch curvature, whole-cell domain restriction, and branching are not. The topological and shape analyses also identify which cell types differ enough to be separable, such as glia versus neurons in rodents, and which look conserved across species, such as microglia. The accompanying 50 high-resolution 3D meshes give modelers concrete cell geometries for simulations, turning the paper into a benchmark resource for grey-matter microstructure imaging.","feed_headline":"11,500 brain cell shapes mapped for sharper grey-matter MRI","feed_subtitle":"Reference values for nine cell types give diffusion MRI the ground truth it lacked.","key_machinery":"The machine that carries the analysis is a cell-by-cell morphometry pipeline applied to skeleton-format reconstructions (SWC files), a standard textual representation of cell nodes, connections, and radii. Each cell is split at the nominal soma radius into a soma and its projections; the soma is turned into a 3D mesh for volume and surface; projections are divided into branches and then into cylindrical subsegments; and the branch radii, lengths, angles, curvature, undulation, and surface-to-volume ratios are summarized as quartiles within each species and cell-type group. Two derived quantities carry much of the argument: the effective MR radii $\\mathrm{RMR}_{\\mathrm{soma}} = (\\langle R_{\\mathrm{soma}}^7\\rangle / \\langle R_{\\mathrm{soma}}^3\\rangle)^{1/4}$ and $\\mathrm{RMR}_{\\mathrm{branch}} = (\\langle R_{\\mathrm{branch}}^6\\rangle / \\langle R_{\\mathrm{branch}}^2\\rangle)^{1/4}$, which set the length scale relevant to diffusion MRI, and the residence and exchange times $\\tau_i = 1/((S/V)\\kappa)$ and $\\tau_{\\mathrm{ex}} = \\tau_i f_{\\mathrm{ec}}$, which convert surface-to-volume ratios and membrane permeability into the time scales that determine whether a feature is detectable. Shape descriptors come from decomposing cells into 10 $\\mu$m cylinders and fitting orientation distributions; topology comes from persistence barcodes of branch paths relative to the soma.","core_discovery":"The paper claims to provide the first systematic empirical reference set for grey-matter diffusion modeling. From 11,850 three-dimensional reconstructions of nine cell types across mouse/rat, monkey, and human cortex, it derives quartile distributions for structural traits (soma radius, branch length, surface-to-volume ratios, curvature and tortuosity), shape traits (fractional anisotropy and orientation dispersion of the arbor), and topological traits (persistence barcodes and pairwise distances between cell types). Its central discovery is a set of time-scale estimates: under typical clinical diffusion times, soma restriction and water exchange between projections and the extracellular space are the features diffusion MRI can detect, while branch curvature, whole-cell domain restriction, and branching order are largely invisible. It also supplies 50 high-resolution 3D surface meshes, one per available cell-type and species combination, intended for Monte Carlo simulation of diffusion signals.","pith_inferences":["Editorial extension: using the provided meshes in Monte Carlo simulators, one could directly estimate whether glial and neuronal morphologies produce distinguishable diffusion-weighted signals at clinically realistic noise levels; the paper furnishes the geometry but does not run those simulations.","Editorial extension: because the reference values come from spine-free reconstructions, modelers could combine the reported branch surface-to-volume ratios with published spine densities to predict how much in-vivo residence-time estimates would shorten; the paper flags the effect but does not fold it into its headline time cutoffs.","Editorial extension: the persistence-barcode distances could be turned into a healthy-tissue baseline, allowing the same topological pipeline to score pathological samples for morphology changes; this is a use the paper's resource enables but does not itself claim.","Editorial extension: the reported sensitivity of soma metrics to the nominal soma boundary suggests a practical calibration study, recomputing all reference values under a range of soma-radius thresholds to identify which descriptors are stable enough for clinical model fitting."],"forward_implications":["If these reference values are right, grey-matter dMRI models should keep a soma compartment: at a soma radius near 5 µm, soma restriction is detectable for water at diffusion times above about 2.5 ms and for metabolites above about 12.5 ms.","Water exchange between projections and the extracellular space is a measurable effect at typical clinical times (derived exchange times roughly 3–30 ms), so exchange-inclusive models are preferable when the diffusion time exceeds a few tens of milliseconds.","Branch curvature, whole-cell domain restriction, and branching are negligible at standard diffusion times, since their effects only appear at hundreds of milliseconds or more, supporting simplified cylinder or stick models for many acquisitions.","Fractional anisotropy and orientation dispersion can separate cells with polarized arbors such as Purkinje and granule cells from most glial cells, making DTI- and NODDI-style contrasts plausible readouts of cortical and cerebellar cytoarchitecture.","In rodents, glial and neuronal morphologies are topologically distant, which supports the prospect of separating glial and neuronal contributions to the dMRI signal through appropriate modeling."],"supporting_citations":[{"why":"Provides the large open-access pool of 3D cellular reconstructions from which the 11,850-cell analyzed dataset was selected.","marker":"[39]"},{"why":"Defines the soma-and-neurite grey-matter diffusion model whose nominal cell geometry the new reference values are intended to constrain and refine.","marker":"[32]"},{"why":"Supplies the effective MR radius formulas and the exchange-inclusive grey-matter model used to convert morphology into diffusion-time predictions.","marker":"[34]"},{"why":"Establishes the cylinder-decomposition and scatter-matrix method used to compute fractional anisotropy and orientation dispersion for each cell.","marker":"[45]"},{"why":"Introduces the topological morphology descriptor and persistence-barcode construction used for cell-type and cross-species topological distances.","marker":"[61]"},{"why":"Gives the residence-time vs. permeability relation used to convert measured surface-to-volume ratios into intracellular residence times.","marker":"[53]"},{"why":"Gives the two-compartment exchange model relating residence times to exchange times at a specified extracellular volume fraction.","marker":"[54]"}],"fun_headline_variants":["11,500 brain cells map what MRI can and cannot see","Cell shape atlas sharpens grey-matter diffusion MRI","Reference cells reveal dMRI's limits in grey matter","11,500 reconstructions demystify brain MRI signals","Grey matter: which cell features affect diffusion MRI?"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported soma measurements assume that the radius attached to the first node of each reconstruction marks the true boundary between the cell body and its branches; if that boundary is wrong, soma volume, surface, MR radius, and all derived exchange times shift with it.","fun_headline_variants_meta":{"raw":{"variants":["11,500 brain cells map what MRI can and cannot see","Cell shape atlas sharpens grey-matter diffusion MRI","Reference cells reveal dMRI's limits in grey matter","11,500 reconstructions demystify brain MRI signals","Grey matter: which cell features affect diffusion MRI?"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000714,"raw_usage":{"total_tokens":3238,"prompt_tokens":1000,"completion_tokens":2238,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":616,"completion_tokens_details":{"reasoning_tokens":2159}},"tokens_in":616,"tokens_out":2238,"duration_ms":15550,"temperature":1.0,"reasoning_tokens":2159,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:14:20.570110+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a sample of the analyzed reconstructions and recompute every soma metric after shifting the soma boundary by plus or minus 20 percent, and where possible compare against high-resolution electron microscopy images of the same cells; if the resulting soma radii, MR radii, or residence times move by more than their reported interquartile ranges, the reference values are too threshold-sensitive to anchor diffusion models.","supporting_citations":[{"cited_title":"Sandi: a compartment-based model for non-invasive apparent soma and neurite imaging by diffusion mri","cited_arxiv_id":null,"evidence_quote":"Defines the soma-and-neurite grey-matter diffusion model whose nominal cell geometry the new reference values are intended to constrain and refine."},{"cited_title":"Neuromorpho","cited_arxiv_id":null,"evidence_quote":"Provides the large open-access pool of 3D cellular reconstructions from which the 11,850-cell analyzed dataset was selected."},{"cited_title":"Diffusion time dependence, power-law scaling, and exchange in gray matter","cited_arxiv_id":null,"evidence_quote":"Supplies the effective MR radius formulas and the exchange-inclusive grey-matter model used to convert morphology into diffusion-time predictions."},{"cited_title":"Using diffusion anisotropy to characterize neuronal mor- phology in gray matter: the orientation distribution of axons and dendrites in the neuromorpho","cited_arxiv_id":null,"evidence_quote":"Establishes the cylinder-decomposition and scatter-matrix method used to compute fractional anisotropy and orientation dispersion for each cell."},{"cited_title":"A topological representation of branching neuronal morphologies","cited_arxiv_id":null,"evidence_quote":"Introduces the topological morphology descriptor and persistence-barcode construction used for cell-type and cross-species topological distances."},{"cited_title":"Machine learning based compartment models with permeability for white matter microstructure imaging","cited_arxiv_id":null,"evidence_quote":"Gives the residence-time vs. permeability relation used to convert measured surface-to-volume ratios into intracellular residence times."},{"cited_title":"Monte carlo study of a two-compartment exchange model of diffusion.NMR in Biomedicine, 23(7):711–724, 2010","cited_arxiv_id":null,"evidence_quote":"Gives the two-compartment exchange model relating residence times to exchange times at a specified extracellular volume fraction."}],"review_version":1}