Pas de résumé en français / The cortex can be viewed as a network of functional areas. A cortical area, composed ofneurons forming local connections, interacts with other areas via long distance connections.Each neuron receives multiple inputs and has to integrate the incoming signals. This integrativecapacity is the basis of the computational power of the brain. Our work concentrates onunderstanding the principles that govern the structure of the cortical network i.e. the allocationof neural resources as well as the anatomical segregation between processing steams. Usingretrograde tracer injections we extract two quantitative parameters: (i) the proportion ofSupragranular Labelled Neurons (SLN) identifies the feedforward (FF) or feedback (FB)operation between the source and target area; (ii) the Fraction of Labelled Neurons (FLN)identifies the magnitude of a connection pathway.We have made repeat injections in V1, V2, V4 to investigate the consistency of corticalpathways. This showed that (i) connection weights are consistent between animals; (ii) the listof areas projecting to each injection site is highly reproducible. We find that there are fixedFLN values for each pair of interconnected areas. The FLN values of all the afferent pathwaysto a given target span over a factor of 6 levels of log and although there is some overdispersiontheir variability is not larger than one single level of log meaning that there is a specificconnectivity profile for each area. Futermore the FLN follow a lognormal distribution. Inlognormals the mode is lower than the median and the mean i.e. the majority of pathways haveFLN weaker than the average FLN, meaning that strong projections are rare. If instead thedistribution of FLN was to follow a power law, then high FLN values would have been evenrarer. We found, a regularity in that the strongest input is invariably from within the injectedarea, second strongest are the inputs from areas sharing common borders with the target area.Sub-cortical inputs have a weak FLN, even when they are associated with an importantfunctional role such as the LGN → V1 pathway. We found that projection distance is inverselyrelated to the FLN value and an exponential distance rule operates that constrains short distanceprojections to high FLN and long distance projections to low FLN.We injected a total of 26 cortical areas homogenously distributed across the cortex. Thisrevealed 1232 projection pathways. Roughly 30% of pathways that we reveal have notpreviously been reported in the literature. Our ability to find new connections is due to theimproved tracing and brain segmentation techniques. We scan the whole brain at up to 80μmintervals to detect projection neurons, and this, as discussed in the text, is a major advantage toexisting studies. The weak long distance connections were shown to contract the characteristicpath-length of the graph (number of hops needed to go between any two areas).Our analysis of the graph showed that contrary to current belief the cortical inter-areal networkis dense (i.e. 58% of the connection that could exist do exist). At such a density, models basedon binary features such as small world cannot capture the specificity of the graph. Hence thecortex does not correspond small–world network, with sparse clustered graph possessingempowered by few critical projecitons that ensure short characteristic path-lengths. Furtheranalysis of pathway efficiency showed that the short distance connections of high magnitudeprovide large bandwidth for local connectivity and form a backbone of clustered functionallyrelated areas. This backbone is embedded in a sea of weak connections providing direct linksbetween cortical areas. We refer to this architecture as a tribal–network. We speculate that thesmall scale and high density that characterize the cortico-cortical network is facilitating theemergence of synchrony between cortical areas.
| Identifer | oai:union.ndltd.org:theses.fr/2010LYO10079 |
| Date | 03 June 2010 |
| Creators | Markov, Nikola |
| Contributors | Lyon 1, Kennedy, Henry |
| Source Sets | Dépôt national des thèses électroniques françaises |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation, Text |
Page generated in 0.0024 seconds