Week 1 - introduction Flashcards
What is a model?
Simplified abstraction of data or a system
how can models vary ?
Explanatory power - From descriptive to mechanistic
Spatial scale = Macro to micro
Biological realism - abstract to realistic`
Why build models of the nervous system?
Make assumptions explicit
Make experimental predictions and generate hypothesis
Reduces dimensionality of nervous system data
Understand the structure/function/development/disease
what are nodes and edges in networks?
Nodes are the building blocks of the network
Edges are the relationship between nodes
what is the range of spatial scales that connectomics can be used for?
Micro - single cells
Meso - Local-circuits / processing units
Macro - Large scale grey areas
Describe flywire
The first complete connectome of the entire drosophila brain
5x10^7 synapses described between 139,000 neurons
What is the range of temporal scales that connectomics can study changes across
Fast - intrinsic dynamics (ms) e.g hippocampal cells in vitro, fmri dynamics
Mid - e.g experience dependent plasticity over days
Slow - Ageing and development over years
describe connectomics in structural and functional networks
Connectomics can be used to study structural and functional brain networks
Structural networks shape the functional networks on a fast time scale (seconds/minutes) - this is known as brain dynamics
Function shapes structure on a slow time scale - this is known as plasticity (days/ years)
describe network types:
Binary : simple, either there is a connection between two nodes or there isnt
Weighted : edges have weight meaning some connections can be stronger
Directed: edges have a direction
real brain networks are weighted and directed but human connectome models are usually weighted and undirected. This is because recovering the directionality of connections is quite difficult (it requires causal statistical models or invasive imaging)
descirbe networks as a matrix
Networks are represented as square matrices
Rows and columns correspond to nodes
Entries in the matrix indicate whether theres an edge between the two nodes, and its weighting
what is a network neuroimaging pipeline?
- imaging (e.g diffusion mri, structural or functional mri)
- Parcellation (defining nodes, grey matter is subdivided into smaller areas)
- Connectivity (Defining edges)
- Thresholding (only retaining some connections and setting others to zero)
- Analysis
Defining nodes in human MRI
Nodes are defined using structural MRI. T1w MRI scans allow you to define white matter.
Then segmentation can be applied, which involves subdividing/ labelling different tissues
Then parcellation is applied where the grey matter is divided into subregions. This can be done using anatomical principles (with anatomical atlas), or it can be done randomly. Or it can be done based on homogenous structure/function between regions.
describe the structural connectome to define edges
It is derived from diffusion weighted imaging and tractography
diffusion imaging measures the movement of water molecules within each voxel in the brain. If the voxel has a white matter pathway the water molecules will tend to move in the direction of the white matter pathway
Tractography involves analysing these directions to determine white matter tracts
describe morphometric similarity networks to define edges
Measures features such as grey matter thickness, volume etc, for each region
Then you estimate correlations between these features for pairs of regions to determine whether the regions are similar in their morphometry
This has been proven to show a strong overlap with connectivity measures, suggesting this can be an indirect way of measuring anatomical connectivity.
An advantage of this is that you can use only a T1 weighted scan to do this (you don’t need diffusion imaging as well)
A disadvantage is that it’s more indirect
describe structural covariance networks to define edges
like morphometric similarity except you only measure a single feature e.g grey matter thickness, and compare it across participants .
So the question is, for each two regions, are they similarly thick across all participants
This produces a structural covariance matrix
What is a similarity networks?
Combining SCN, MSNs and other networks to measure similarity between regions. Then use these to definine indirect measures of connectivity between networks
What is the functional connectome?
Combine nodes with functional synchrony to see if the activity between nodes is correlated over time
Commonly used with functional MRI, measuring changes in BOLD signal
what is dynamic vs static functional connectivity?
static methods generate a single network per scanning session
But dynamically transient ‘states’ can be derived as well e.g co-activation patterns (CAPS) or hidden markov models
why is connectivity important?
- Studying relationships between structure and function
- We can learn how structural networks constrain functional networks
- we can study how this relates to grey matter atrophy in neurodegenarative disorders. Studies show that networks with high functional and structural connectivity are more likely to show grey matter atrophy
What is structure function coupling?
Correlating each edge between a structural and functional network to obtain a measure of correlation
You can also see how this correlation varies between brain regions
Describe thresholding to remove weak edges
Most of these networks are dense, meaning that a relationship is estimated between almost every possible pairs of regions
Thresholding makes sparse networks, removing some of the weak connections
Optionally, these networks can also be binarised
Describe absoloute vs proportional thresholding
Absolute - retaining edges above a certain weight
Proportional - retaining a fixed percentage of the strongest edges, e.g retain Y% of the strongest edges. This helps remove bias based on different density of edges for different ppts
