Increasing the realism of in silico cortex substrates for diffusion MRI simulation

1609 

Abstract Submission 

Anas Bachiri¹, Alexis Brullé¹, Ivy Uszynski¹, Cyril Poupon¹

¹BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA, Gif-sur-Yvette, France

Anas Bachiri  
BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA
Gif-sur-Yvette, France

Alexis Brullé  
BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA
Gif-sur-Yvette, France

Ivy Uszynski  
BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA
Gif-sur-Yvette, France

Cyril Poupon  
BAOBAB, NeuroSpin, Paris-Saclay University, CNRS, CEA
Gif-sur-Yvette, France

Numerical simulations have provided an interesting approach to studying and validating brain microstructure models¹. For this purpose, many methods to generate virtual brain cells have been proposed. However, these methods target only the generation of substrates with specific brain cells of a given type. For example, generating a virtual scene populated only by axons ^2,3 or neurons of multiple types⁴. To the best of our knowledge, no method was proposed to synthesize virtual microstructure scenes combining neurons with white matter cells such as axons. This work aims to close the modeling gap of virtual brain microstructure by extending the MEDUSA framework⁵ to allow the generation of hybrid virtual substrates. The presented method allows the synthesis of volume of interests (VOIs) combining neurons with axons with similar volume fractions to values reported on human brain gray matter microstructure⁶.

Dataset: The NEUROMORPHO dataset^7,8 has been used to reconstruct virtual neurons of the human brain in the ".SWC" format.
Cell Reconstruction: The reconstruction method creates spheres from the provided cell ".SWC" file which are then interpolated to represent synapses and to fill the gaps between different connected parts of the neuron. Thus providing the same sphere-atom representation used in MEDUSA. The reconstruction code is available at the repository⁹.
Generation of neuron populations: Four different substrate types with VOI of 100x100x100µm³ were considered. For each substrate type, ten different samples have been generated. The first substrate type a) comprises axons with an 85% volume fraction (VF). Second type b) contains a population of axons with 44% VF and a population of pyramidal neurons with 6% VF. The substrate type c) is similar to b) with a more dense population of neurons of 12% VF. Finally, the type d) has only a neuron population with 15% VF. The VF values in substrates b) and c) are comparable to those reported from human brain samples scanned with electron microscopy⁶. The MEDUSA framework was used to generate the substrates and to perform cell de-overlapping to minimize intersections between different cells.
Diffusion MRI simulations: Diffusion MRI simulations were performed on the generated substrate samples. For each sample type, the diffusion MRI signals were orientation-averaged, and an average signal for all samples of the same type was calculated. The diffusivity used is D₀=3x10^-9 m²/s with 500k walkers initialized in both intra and extracellular space and a time step of 15μs.

Figure 1 illustrates a reconstructed pyramidal neuron cell from the NEUROMORPHO dataset using the reconstruction code⁹, the spheres used to represent the neuron highly overlap to ensure a high quality representation of the cell. Figure 2 upper panel shows four samples representing each substrate type generated in a VOI of 100x100x100μm³. The sparse representation of these substrates with sphere atoms allows to keep the memory required for each sample below 60 Mbytes. Finally, the simulated diffusion MRI signals in the bottom part of Figue 2 show a clear distinction between signal attenuations in the four different substrate types. The diffusion process parameters were chosen such that the diffusion is in the short time limit (t_d < 20 ms) and the cells were considered to be impermeable.

This work demonstrates a proof of concept for synthesizing virtual substrates with hybrid cells mixing neurons and axons, thus increasing the modeling realism of synthetic gray microstructure. Future work directions include large-scale simulation of microstructure and exploring the ability to add more components in the synthetic substrates such as blood vessels and glial cells, which were not included in this study.

Diffusion MRI Modeling and Analysis ¹

Methods Development ²

Cortex

Modeling

MRI

Neuron

Open Data

Open-Source Code

Open-Source Software