henhaobio
Sign in

bioRxivNeurosciencePreprint

The modular nature of long-association pathways in the human brain

Chiara MaffeiiD, Marina CélestineiD, Ting GongiD, Robert J. JonesiD, Evan Dann, Julia F. Lehman, Jean C. AugustinackiD, Hui WangiD, Suzanne N. HaberiD, Anastasia YendikiiD

PDFDOI
In-depth full-text digest available · Jump to digest ↓

Abstract

Human brain functions are facilitated by large-scale networks, yet the organization of the long-range association pathways that form the wiring of these networks remains poorly understood. The advent of diffusion MRI (dMRI) tractography enabled, for the first time, the non-invasive mapping of these pathways, albeit at low resolutions. This contributed to a simplistic view of these pathways in the neuroimaging field, where they are conceptualized as monolithic entities, each linking a pair of distant cortical regions. This contrasts with classical anatomic tracer studies showing that each of these pathways is a network in itself, serving as a conduit for multiple fiber systems including short- and long-range projections. While higher-resolution dMRI and optical microscopy techniques for imaging the wiring of the human brain have emerged in recent years, our simplistic view of long association pathways from low-resolution tractography has not yet evolved. Here we present the first proof-of-principle study showing that, with these techniques, we can now replicate the modular nature of human brain pathways, as it has been known from tracer studies in animals. We demonstrate this on the most dorsal branch of the superior longitudinal fasciculus, the proposed backbone of the dorsal attention network. We show that the long-range, frontoparietal connections that we see with dMRI tractography at lower resolution (1.5 mm) are in fact composed of multiple, layered fiber systems of varying lengths that can be reconstructed with dMRI tractography at higher resolution (<750 m). We validate this result across species, modalities, and scales by combining in vivo and ex vivo diffusion MRI with anatomic tracing in macaque and polarization-sensitive optical coherence tomography in human. These findings underline that advancing from macroscopic to mesoscopic imaging has the potential to transform our understanding of the wiring of the human brain, and allow the study of human brain networks in health and disease at a level of detail heretofore only possible in animal models.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (16 Oct 2026) · CC BY 4.0

Diffusion MRI tractography below 750 µm resolves the dorsal superior longitudinal fasciculus into layered fiber systems of varying length, matching macaque tracing; 1.5 mm tractography misses this.

Human tractography has treated long association pathways as single bundles joining two distant cortical regions, whereas tracer studies in animals show each carries several short- and long-range fiber systems, and human reports on the dorsal branch of the superior longitudinal fasciculus (SLF-I) conflict. This work tests whether mesoscale ex vivo diffusion MRI (dMRI) and polarization-sensitive optical coherence tomography (PS-OCT) recover the tracer-like organization in humans.

Key results

  1. ShownAt 1.5 mm in vivo resolution, SLF-I connectograms contain long direct links from frontal regions (area 10, medial and lateral prefrontal cortex) to parietal cortex that are absent at higher resolution, where most long frontoparietal links run in the cingulum bundle (CB) instead.

    Connectograms from macaque ex vivo dMRI (700 and 500 µm), human ex vivo hemispheres (750 µm), one in vivo gSlider-SMS subject (760 µm) and 10 MGH-USC HCP subjects (1.5 mm). At higher resolution the SLF-I held shorter links, with the highest streamline counts between premotor/motor and parietal regions; the 1.5 mm profile appeared in all but one HCP subject. · A fundamentally different organization in…; Connectivity matrices of SLF-I and neighboring…

  2. SuggestedIn two ex vivo human hemispheres, the SLF-I rostral to area 6 is made mostly of short relay fibers between neighbouring dorsomedial prefrontal areas, while longer area 6M-to-parietal streamlines form layers of progressively shorter bundles.

    Manual dissections at 750 µm: rostral to 6M, mainly 6M-8B, 8B-9M, intra-9M and 9M-10 links, with only a few longer parietal-to-8B or caudal 9M streamlines beneath them and above the CB. Block dMRI at 250 µm and PS-OCT at 10 µm showed no long direct connections under the short fibers. · High-resolution ex vivo dMRI reveals the modular…; PS-OCT in human tissue reveals predominantly…

  3. ShownTracer injections in macaques place the longest direct SLF-I connections between parietal areas PE/PGM and pre-supplementary motor areas 6M/6D; injections in 9M or 8B yield only short SLF-I fibers, and no case showed direct PGM/PE-to-8B or 9M fibers.

    7 male macaques, dextran-amine anterograde/bidirectional tracers (Cases 1-7). Cases 1-3 (9M, 8B) terminated mainly in 9L, 6M, 6D and 24, with most fibers entering the CB; Cases 4-5 (rostral 6M) gave sparse 9M/6D terminations; Case 6 (6D) reached 8B/9M anteriorly and PGM/PE posteriorly; Case 7 (PE) ended in caudal 6M. · Anatomic tracing in NHP shows multiple fiber…; Anatomic Tracer Data in NHPs: acquisition and…

  4. SuggestedEx vivo macaque tractography (one brain at 700 µm, one at 500 µm) reproduces the modular SLF-I pattern seen with tracers but adds connections the tracers do not show.

    Most streamlines ran PE/PGM to 6M in layers, with short 6M-8B/9M streamlines rostrally; ROIs at the Case 1, 6 and 7 injection sites gave matching routes (9M-6M; 6D-9M and 6D-PE; PE-caudal 6M). Longer PGM/PE-to-8B (both brains) and 9M (700 µm brain) streamlines and lateral projections to middle and inferior frontal gyri lacked tracer support. · Ex vivo tractography in NHPs replicates the…

  5. SuggestedAgainst binarized tracer connectivity, macaque ex vivo tractography matches tracers for parietal regions but misses 9M-to-area-6 links, while human tractography adds false positive connections, extending to parietal and many frontal targets at 1.5 mm in vivo.

    Connectograms aggregated over all 7 tracer cases vs tractography (M3, H1, H5) for prefrontal, pre-supplementary motor and parietal regions: human 750 µm showed false positives from 9M/8B to cingulate and motor cortex, 1.5 mm also to parietal regions; parietal seeds gave false negative anterior cingulate links ex vivo and many false positive frontal links in vivo. · High-resolution tractography in NHPs and humans…

Methods, data and reproducibility

Fiber orientations from constrained spherical deconvolution (MRtrix3) on in vivo and ex vivo dMRI from 1.5 mm to 250 µm, plus PS-OCT orientation maps at 10 µm tracked with an in-house algorithm, fed probabilistic connectograms and manual virtual dissections (deterministic in humans, probabilistic in macaques) of the SLF-I and cingulum bundle under a shared parcellation, compared with tracer injections in 7 macaques.

  • M1: anterograde/bidirectional tracer injections (lucifer yellow, fluororuby or fluorescein dextran amine) in dorsomedial cortex · macaque · n = 7 male animals
  • M2: ex vivo whole-brain dMRI at 700 µm, 4.7T Bruker, 515 diffusion-weighted volumes · macaque (M2+M3: 2 Macaca mulatta, 2 Macaca fascicularis) · n = 2 animals
  • M3: ex vivo whole-brain dMRI at 500 µm, 4.7T Bruker, plus multi-slice multi-echo images · macaque · n = 2 animals
  • H1: ex vivo dMRI of formalin-fixed left hemispheres at 750 µm, 3T Siemens Trio, DW-SSFP sequence · human · n = 2 hemispheres (neurologically normal donors)
  • H2: ex vivo dMRI of dorsomedial frontal tissue blocks at 250-300 µm, 9.4T Bruker · human · n = 8 tissue blocks (S3); 3 and 4 blocks per specimen (S18)
  • H3: serial-sectioning PS-OCT (1300 nm) of sub-blocks, 3-10 µm in-plane, downsampled to 10 µm · human · n = 10 blocks
  • H4: in vivo dMRI with gSlider-SMS at 760 µm on MGH Connectom 1.0, 2664 DW volumes over nine two-hour sessions · human · n = 1 subject (male, 32 years)
  • H5: MGH-USC Human Connectome Project in vivo dMRI at 1.5 mm, 552 DW volumes over four shells · human · n = 10 subjects (7 males; age 20 to 59 years)
Methods & tools
MRtrix3, FSL (eddy correction, DTIFIT), FreeSurfer v.7.5.0, ANTs, Trackvis, Neurolucida (MBF Bioscience), IMOD, Fiji, NMT v2.0 macaque template, Paxinos atlas (in-house volumetric version), custom serial-sectioning PS-OCT system (1300 nm), in-house multi-scale tractography algorithm
Reagents & environment
FreeSurfer v.7.5.0 and the NMT v2.0 template are versioned; versions of MRtrix3, FSL, ANTs, Trackvis and the in-house tractography code are not given.
Reproducibility gaps
No code or data links. No release or link is given for the in-house multi-scale tractography algorithm or the C++ PS-OCT acquisition software; in vivo data are called publicly available with no accession; block counts differ (8 in S3, 3 and 4 per specimen in S18).
Limitations (6)
  • Ex vivo macaque tractography still departs from the tracer data: it misses 9M-to-area-6 streamlines, adds longer PGM/PE-to-8B streamlines and lateral frontal projections, and places the SLF-I higher in the superior frontal gyrus than the tracer fibers. · stated by the authors
  • Human tractography disagrees with aggregated tracer connectivity at both resolutions tested: at 750 µm it adds cingulate and motor connections and misses anterior cingulate links from parietal regions; at 1.5 mm it adds parietal and many frontal connections. · stated by the authors
  • SLF-I fibers lie more dorsally in the macaque superior frontal gyrus than in humans, which the authors tentatively attribute to greater human cortical gyrification. · stated by the authors
  • Only the SLF-I was examined; the authors note that whole-brain mapping is needed to learn whether other major fiber systems share this organization. · stated by the authors
  • Human false positives and negatives are scored against binarized macaque tracer connectivity, which assumes cross-species homology; agreement is judged visually, and the reported significant drop in long-to-short connection ratio at 750 µm comes with no statistic in the text. · noted in this summary
  • The finest-scale evidence rests on few samples: two ex vivo human hemispheres, one 760 µm in vivo subject, one macaque brain per resolution for dissection and a single parietal tracer injection (Case 7); one author drew all virtual dissections manually. · noted in this summary

Read this if you build or interpret association-bundle tractography and want evidence on how voxel size changes the anatomy you recover, checked against macaque tracers and human PS-OCT.

Citations
1

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISYeh '20Maffei '26Rokem '17Liang '22Wang '21Kai '19
68% semantic matchNeuroscience
Fang-Cheng F. Yeh·NeuroImage·Sep 1, 2020· 239 citations