This site uses cookies, so that our service can work better. I understand

Project 5: Hypothesis­-driven white matter tractography from T1­-weighted MRI images

Authors: Anastasia Osoianu / Charl Linssen / Katja Heuer / Roberto Toro



Polarized light imaging (PLI) as well as the tractography of high angular resolution diffusion weighted imaging (DWI) data reveal a gross white matter (WM) geometry of striking regularity [1, 3]. This makes the neuroanatomist wonder whether it would be possible to generate a connectome based exclusively on a small set of hypotheses:

• more than 90% of white matter connections are cortico­cortical

• the density of fibres is homogeneous throughout the white matter

• fibres are oriented perpendicular to gyral crowns and parallel to sulcal fundi

• fibres are sticky, which makes them aggregate in bundles of similar orientation.

How much of a real brain connectome would be recovered by such a simple and reductionistic model?


We propose to approach this question in the simple case of a 2D coronal slice. We use methods inspired by swarm intelligence [2], and simulate a dynamical system where particles are instantiated at the gray matter/white matter boundary. The particles are allowed to propagate within the white matter mask, following simple rules that reflect the above list of hypotheses. After reaching a steady state, the white matter orientation distribution is reconstructed on a voxel-by-voxel basis, for each individual voxel based on the statistics of the particle paths crossing it. This is roughly the opposite of streamline tractography, where paths are sampled based on a distribution map. By implementing derived measures such as an anisotropy index, the resulting map can be quantitatively compared to empirical data. We are especially interested in multimodal effects that occur e.g. when two fibre bundles cross. We will validate our model on coronal slices of the vervet monkey, as an excellent empirical dataset is available [3]. Proof­ of­ concept code is available via GitHub [4].

A list of 1-5 key papers/materials summarising the subject:

[1] Frontier Research Topic "Wiring Principles of Cerebral Cortex" —

[2] Craig Reynolds' "Boids" —

[3] Vervet monkey brain slice scanned using Polarized Light Imaging —


A list of requirements for taking part in the project (education level / English level / programming language required):

- familiarity with Python, Javascript or similar programming language

- affinity for simulating and analysing dynamical systems

A maximal number of participants on the project:


Skills and competences you can learn during the project:

You will reflect on the organisational principles of white matter network connectivity across spatial scales, and formulate hypotheses about it. The simulation will be used to test as well as generate these hypotheses. The outcome of the simulation is continuously compared to that derived from PLI images.

Next to collective brainstorming, you can focus on any of two main themes in the project:

1. metrics and quantification, e.g. downloading and processing the PLI images; design the anatomical WM mask; computing derived measures such as anisotropy indices; comparison (e.g. Kullbeck-Leibler divergence) with the map generated by the simulation;

2. development of the simulation method: what rules do particles propagate under? parameter optimisation using smart search (particle swarm optimisation on the parameters? meta-metaheuristic!), generate appropriate network graph (e.g. having small-world properties) that goes into the particle simulation as a boundary condition

Is there a plan for extending this work to a paper in case the results are promising?