Fuzzy Logic Approach to Improving Stable Election Protocol for Clustered Hete...
file_format_presentation
1. A Better File Format for Representing Neuron
Morphology
R. Jarvis
July 13, 2015
2. Introduction
The neural modelling community uses different file
formats, to represent the geometry (and topology) of
a neural tree.
Each file format has offers different strengths and
weaknesses.
The different formats used leads to a lack of standards
and portability in modelling work.
3. Introduction Continued
Why does the neural modelling community choose
different representations of cell form?
Review the strengths and weaknesses of file formats,
such that we can see the limitations (barriers to
universal file format).
4. SWC and Conical Frustum
The traced volume of the cell is output from LM.
At each point a conical frustum is fitted to represent
point of the neuron.
Result is trees of connected conical frustum. [?].
[?]
5. SWC
The C and ID columns can be regarded as a sparse
form of an adjacency matrix.
[?]
6. Why is SWC so popular?
Good compromise between representing geometric
variability and achieving computational tractability.
This system of spatial discretization eases indexing and
recording from ROI.
Also eases splitting the cell trees over multiple CPUs [?].
7. Why SWC is not the desired standard?
The frustum is limited in how well it can represent
variations in surface, and such variations are important
determinants of neuron excitability.
Not suited to representing spines or myline [?].
Within neuromorpho no minimum quality of
reconstruction, or header information [?].
8. MATLAB toolbox Trees
Trees is based on the idea that neuron connectivity
can be represented by two or more binary trees
(graphs).
[?]
9. MATLAB Trees: Benefits and Drawbacks
Benefits
– Matrix representation of connectivity convenient for
morphometric analysis comparison and editing tree
morphology.
Why is it not the standard?
– The type of geometry representation is not a mesh.
– Neither Trees or SWC represent time varying structural
changes.
– Although can output .swc, .mtr extension saves a MATLAB
workspace.
11. Mesh Can Better Represent Variation in Surface
Varying geometry of surface effects electrotonic
compactness
– and firing patter of the neuron [?].
Alters passive resistance to current following synaptic
input.
Surface geometry important to Reaction Diffusion.
[?]
12. STereoLithography (STL) and Polygon surface
solid name
facet normal ni nj nk
outer loop
vertex v1x v1y v1z
vertex v2x v2y v2z
vertex v3x v3y v3z
endloop
endfacet
endsolid name
15. Abaqus and STL Why Are they Used?
Surfaces are a good model for reaction diffusion.
In the case of Abaqus. Dendrite Spine Endoplasmic
Reticulum can be represented, which is important for
reaction diffusion.
The membrane potential can be evaluated on and
between vertices using STEPS.
16. So Why doesn’t everyone Use Abaqus and STL?
Less computationally tractable esp. for network scale
simulations.
Discretization not simple, vertices not clearly in or out
of conical ROI.
Current injection into membrane requires defining
direction of current flow from outside to inside the cell
(unlike NEURON/GENESIS).
Splitting trees across CPUs not simple since mesh
vertices much more interconnected than conical
frustum.
17. Multisplit Trees
To make dynamic simulations computationally
tractable. Neuron trees can be split up and simulated
on different CPUS.
When solving the cable equation dependence
between nodes is important.
[?]
18. Synthetic Morphologies
NeuroMAC output provides pre-synaptic and
post-synaptic coordinates, in addition to cell id’s.
An adj. matrix is lacking, there is no standard for
multidimensional connectivity information.
Given NeuroMAC, NETMORPH, CX3D and Neugen the
future should include dynamic sim. in of time-varying
neuron structure.
– [?] [?] [?]
20. Use Case: New morphology file format
Primary Actors: Neuroscientists
Stakeholders: Neuroscientists investigating any or all of:
– Statistics of the tree structure of neurons
– Electrophysiology in neurons and networks
– Connectivity in forests of neurons.
Scope: Representing the time varying or static
geometry and topology of a single neuron, or a
connected forest of neurons.
Level: Sub function -
21. Use Case: continued
Triggers: Actor has completed nano-meter resolution
neural imaging, image stack is available for dynamic
or structural modelling.
Preconditions: Available polygon mesh nodes and
vertices of manifolds.
Success Guarantees:
– More accurate results than achieved via 1D exclusive use of
the 1D cable equation.
– The solver exploit FEM simulation at the area of the spine.
– simulation executes faster
– Lesser RAM/CPU burden than with a full FEM simulation.
– Dynamic simulations with time varying structural plasticity are
not be made impossible by the file format.
22. Recommendations for New File Format
Should be XML allows nesting of meta data at any
level of resolution.
Also XML documents facilitate the nesting of cell
geometries in a forest.
Should be a data container for two types
representation (mesh and frustum).
Needs to load balance friendly.
Should not make it harder to make a time series
extension.
23. Conclusion
In this presentation, I have:
Described the limitations of existing morphology file
format.
Described the tradeoff between computational
tractability and biophysical/geometric accuracy.
Proposed possible XML tree structures of a NINEML
extension universal morphology file format.
24.
25. Are there any Questions?
Thank-you For your Attention
