Here are some examples illustrating the use of CoCoMac data for investigating the organisation of the primate brain.

In the past, most connectivity data were provided using a fixed brain map leaving very little flexibility. Here is a well known example of the visual cortical connections from Felleman & Van Essen 1991, which are the only review data contained in CoCoMac.

 connectivity matrix based on Felleman & Van Essen 1991 Cereb. Cx. 1: 1-47

The first column gives the areas of origin, the target areas are listed in the first row. Each entry in the matrix provides information on the state of the connection from a source area to a target area: 1 = connection found, 0 = no connection found, empty field = information missing.

We are working to create graphical representations of connectivity data. Some data can already be displayed using the CoCoMac-Catacomb Interface. Here is an example displaying the above data on a flat map (adapted from Kötter 2004):

flatmap of cerebral cortex with visual system connectivity

This example uses data that were already available for the parcellation scheme proposed by Felleman and Van Essen (1991). 

A more elaborate graphical interface was built on the basis of the "Rhesus Monkey Brain in Stereotaxic Coordinates", an atlas created by Paxinos and colleagues. After manual drawing of the 151 sections we assembled them to a 3D stack whose component structures can be individually viewed and manipulated in 3D space. The CoCoMac-Paxinos-3D viewer was built extending tools from the Virtual Rat Brain project and is freely available for download under the GNU General Public License.

Connections of mediodorsal thalamic nucleus

Different authors, however, partition the brain (particularly the cerebral cortex) in different ways so that the relationships between those parcellation schemes need to be noted if one wants to combine and compare data from several sources. We developed a method for coordinate-independent mapping of connectivity data: Objective Relational Transformation (ORT; Stephan et al. 2000a). Essentially this methods creates a graph of mapping relationships and identifies the best mapping paths in that graph. These mapping paths are subsequently used to transform the connectivity data to an arbitrary target map. Here is an overview of the different processing steps (Stephan et al. 2001):

flow of data processing in CoCoMac

Using the ORT procedure we have able to transform a diverse set of tracing data into an almost complete connectivty matrix of the prefrontal cortex using the map of Walker (Stephan et al. 2001).

binary version of prefrontal connectivity matrix obtained with ORT from CoCoMac data

In addition to the binary information on the presence or absence of a connections we can obtain an estimate on the strength of a connection usually on a rank order scale. The example below shows the same data with strengths of projections graded as 3=strong, 2=moderate, 1=weak, 0=absent connections. As before, empty fields indicate missing information (Stephan et al. 2001).

Alternative processing option preferring strong connections

Processing the data requires that some decisions are made. For cells in the above connectivity matrix we had different information on connection strength. In that case we chose the higher strength under the assumption that a single report of a dense projection supersedes other reports that may have missed the full strength of a projection.

The mapping graph itself contains some interesting information on the similarity of brain regions in terms of their mapping relationships. For example, areas in different brain maps that have the same mapping relationships as expressed in the mapping graph are very likely the same entities. The Multi-Dimensional Scaling procedures expresses similarities as proximities in space. The following graph indicates the similarities of the mapping pathways for a large set of auditory areas, which originate from different parcellation schemes. It shows the similarities among primary auditory areas on the left, variously called AI, A1, KAM, Core, etc.

Similarity of auditory areas as derived from the mapping relations

Many principles developed and implemented for CoCoMac apply also to other data modalities. Seeking for comparable functional measurements we processed data on the spread of epileptiform activity in the macaque cortex as ascertained by strychnine neuronography. Applying the ORT procedure we obtained the following directed functional connectivity matrix (Stephan et al. 2000b):

Functional connectivity matrix for macaque cortex from neuronographic experiments

To compare the correspondence of different data modalities ORT is useful to convert all data into the same brain map. Here is a crude example showing only presence or absence of connections from Stephan et al. (2000a) in the juxtaposition of data from anatomical tracing and from strychnine neuronography:

 Juxtaposition of structural and functional connectivity matrix

Connectivity data from CoCoMac have also been applied to interpret results from functional imaging studies (e.g. Northoff et al. 2000 Cereb Cortex 10: 93-107; Northoff et al. 2002 J Cogn Neurosci. 14: 348-370; Toni et al. 2002 Cereb Cortex 12: 1040-1047).

These are just small and preliminary examples of output from CoCoMac. Further data are continuously collated and lead to additional and improved connectivity tables. Note that an absence or bias of information may occur for any of the following reasons: 
1) no available data, 
2) available data not collated in CoCoMac, 
3) collated data incompatible with target map. 

Please contact us if you are interested in specific data or seek collaborations.


Computational | Systems | Neuroscience Group, 1997-2008