[Paper Review] Parameterizable Consensus Connectomes from the Human Connectome Project: The Budapest Reference Connectome Server v3.0

Balázs Szalkai, Csaba Kerepesi+2 more|Repository of the Academy's Library (Library of the Hungarian Academy of Sciences)|Feb 15, 2016

Functional Brain Connectivity Studies4 references4 citations

TL;DR

The Budapest Reference Connectome Server v3.0 generates parameterizable consensus connectomes from diffusion MRI data of 477 subjects in the Human Connectome Project, identifying edges present in at least k subjects (default k=209). It enables customizable filtering by sex, edge confidence, weight threshold, and tractography parameters, producing robust, downloadable graphs in CSV and GraphML formats for neuroimaging research and error correction.

ABSTRACT

Connections of the living human brain, on a macroscopic scale, can be mapped by a diffusion MR imaging based workflow. Since the same anatomic regions can be corresponded between distinct brains, one can compare the presence or the absence of the edges, connecting the very same two anatomic regions, among multiple cortices. Previously, we have constructed the consensus braingraphs on 1015 vertices first in five, then in 96 subjects in the Budapest Reference Connectome Server v1.0 and v2.0, respectively. Here we report the construction of the version 3.0 of the server, generating the common edges of the connectomes of variously parameterizable subsets of the 1015-vertex connectomes of 477 subjects of the Human Connectome Project's 500-subject release. The consensus connectomes are downloadable in csv and GraphML formats, and they are also visualized on the server's page. The consensus connectomes of the server can be considered as the "average, healthy" human connectome since all of their connections are present in at least $k$ subjects, where the default value of $k=209$, but it can also be modified freely at the web server. The webserver is available at \url{http://connectome.pitgroup.org}.

Motivation & Objective

• To provide a standardized, reference human connectome based on frequently occurring connections across healthy subjects.
• To enable researchers to generate consensus connectomes by filtering for edge presence across a customizable number of subjects (k).
• To support error correction in individual connectomes by identifying connections consistently present across multiple subjects.
• To offer a web-based platform for interactive visualization and programmatic download of consensus graphs in multiple formats.
• To enhance reproducibility and comparability in brain network analysis by providing a parameterized, scalable reference connectome.

Proposed method

• Construction of 1015-vertex connectomes using deterministic streamline tractography with 20k, 200k, and 1M fibers per subject.
• Computation of edge confidence as the percentage of subjects in which each edge appears.
• Application of user-defined thresholds for minimum edge confidence (1–100%), minimum edge weight (mean or median), and weight function (fiber count, length, FA, or electrical connectivity).
• Selection of consensus edges present in at least k subjects, with k configurable by the user (default k=209).
• Integration of pre-computed edge weights and confidences into a web server for real-time graph generation.
• Visualization via a modified WebGL Brain Viewer enabling rotation, magnification, and node labeling with anatomical names.

Experimental results

Research questions

• RQ1 What is the structure of a consensus human connectome that reflects connections present in a majority of healthy subjects?
• RQ2 How does varying the minimum number of subjects (k) required for edge inclusion affect the topology and robustness of the consensus connectome?
• RQ3 To what extent can consensus connectomes be used to correct spurious connections in individual subject connectomes?
• RQ4 How do different edge weight functions (e.g., fiber count, fractional anisotropy) influence the selection of consensus edges?
• RQ5 Can sex-based filtering of subjects reveal differences in structural connectivity patterns in the healthy human brain?

Key findings

• The Budapest Reference Connectome Server v3.0 generates consensus connectomes from 477 subjects of the Human Connectome Project’s 500-subject release.
• The default consensus graph, named 'Budapest Reference Connectome v3.0', is available for direct download without parameter adjustment.
• Users can customize the consensus graph by setting k (minimum number of subjects for edge inclusion), edge confidence threshold (1–100%), and edge weight thresholds.
• The server supports three tractography fiber launch counts (20k, 200k, 1M), with results dependent on this parameter.
• Edge weights are computed using four functions: fiber count, fiber length, fractional anisotropy, or electrical connectivity, with median or mean aggregation.
• The web interface enables interactive 3D visualization of the consensus graph, with node labels and spatial clustering for clarity, and supports export in CSV and GraphML formats.