Can somebody advise me, where I can find the connection matrix for human brain regions (anatomical)? I've already found these matrices for animals and now I am interested in the human.
Thank you very much. Have a nice day.
Full Version: connection matrix
I have placed a link to the "neural systems" page at mind-brain.com in the left navigator (of the main site, not forum). At the page http://brainmeta.com/neuralsystems/connections.html , you will find a primate cortico-cortical connectivity matrix based on the literature up until (and including) 2001. Primate connectivity is a great interest of mine, and I intend to set up a connectivity database (with advanced search and sort features) as soon as time permits.
(btw, primate anatomical connectivity is the closest information we can get regarding human anatomical connectivity since tracer injection experiments in humans are not permitted, and since diffusion-tensor imaging is not good enough yet, nor likely ever, to replace tracer-injection experiments)
(btw, primate anatomical connectivity is the closest information we can get regarding human anatomical connectivity since tracer injection experiments in humans are not permitted, and since diffusion-tensor imaging is not good enough yet, nor likely ever, to replace tracer-injection experiments)
Thank you very much for the answer and for the advice.
I wonder about a few things in your matrix.
Why do you include selfconnection?
Why is your matrix symmetrical?
Do you know also the strength (weight) of the certain connections?
Why is there the difference between the matric of Felleman&Van Essen and your matrix (for ex. -connection between the regions V1 and V3)?
I'm looking forward to hear from you
I wonder about a few things in your matrix.
Why do you include selfconnection?
Why is your matrix symmetrical?
Do you know also the strength (weight) of the certain connections?
Why is there the difference between the matric of Felleman&Van Essen and your matrix (for ex. -connection between the regions V1 and V3)?
I'm looking forward to hear from you
>>Thank you very much for the answer and for the advice.
anytime.
>>Why do you include selfconnection?
because all cortical areas have local intra-areal connections, usually with related "modules" or cortical columns. For example, in primary visual cortex, orientation columns will tend to have connections with nearby orientation columns with similar orientation preference. The self-connections in the matrix are really just a reminder of these local intra-areal (or intrinsic) connections.
>>Why is your matrix symmetrical?
Good question! Cortical connectivity is really not symmetrical, though oftentimes, for the sake of simplicity (and/or to mask uncertainties about the existence of bidirectional connections), the assumption of symmetrical connectivity is made. In my particular matrix, it was mainly a convenience factor, because to drop the symmetry assumption would mean there would be twice as many entries in the connectivity matrix I'd have to ascertain. I know it's a lazy man's thing to do. I mean, you can present "reasonable" arguments from the literature showing that cortical connectivity is symmetric, but anyone who sees the actual data from tracer injection experiments knows for a fact that cortical connectivity is not symmetric.
As a general rule, if cortical area A sends feedforward (or lateral) projections to cortical area B, then area B will likely send feedback (or lateral) projections to area A (within the context of a cortical hierarchy). However, examples of seemingly unidirectional (or asymetrical) cortical connections between areas abound in the literature. For example, there is evidence for asymetrical connections involving cingulate cortex and other areas (though the details escape my mind right now), which is interesting considering how high up cingulate cortex is in the cortical hierarchy.
>>Do you know also the strength (weight) of the certain connections?
It's difficult to assign numerical estimates of connectivity weights, though a scheme of 0 (no connection) to 4 (very strong connection) certainly seems do-able. I haven't done it, yet. Part of the difficulty, too, involves assigning scalar magnitudes to qualitatively different types of connections. For example, how do you assign numerical estimates to feedforward vs. feedback connections between V1 and V4? Most feedforford connections are "driving" and terminate in layer 4 and lower layer 3, but feedback connections are more "modulatory".... how would you assign scalar magnitudes to these different types of connections to characterize their strength? It seems that maybe classifying the connections into different types, and then assigning magnitudes, might be a solution. In this way, connections wouldn't be characterized by a single scalar weight, but rather by an n-tuple (i.e., a vector weight).
>>Why is there the difference between the matric of Felleman&Van Essen and your matrix (for ex. -connection between the regions V1 and V3)?
I would have to look at the Felleman and Van Essen matrix again from their '91 Cerebral Cortex paper. Their matrix is very old. I did use it, though, as well as Malcolm Young's (also Hilgetag and Scannell?) more recent connectivity matrix, when I was putting together mine. I had a lot of problems with the Felleman and Van Essen matrix. They grossly over-emphasize visual areas at the expense of somato-motor and auditory areas. They do not even have an area 46vr, which is a prominent somatosensory area in dorsolateral prefrontal cortex that's connected with multiple sensori-motor areas, including SII and premotor cortex (even though evidence for this connection was made very explicit by Preuss and Goldman-Rakic in '89 in Journal of Comparative Neurology, a full two years before Felleman and Van Essen paper came out; how they could not have known of this connection and of this prefrontal somatosensory area amazes me).
In general, I tried to be more detailed in my connectivity matrix than Felleman and Van Essen, in part, because the work done in the decade after their paper was published contained a lot of new connectivity data that was not part of their matrix.
>>I'm looking forward to hear from you
It's been a pleasure. I do intend to have a real primate cortical connectivity database available soon, one that's searchable and sortable. I have 439 references now, ranging from 1945 up to the current date, but I first have to get into them into a database for publishing to the web. I will have this available as soon as possible. That connectivity matrix I have up now is in need of updating too. At the very least, I need to incorporate asymmetrical connections into it (and maybe even weights, as you suggested).
anytime.
>>Why do you include selfconnection?
because all cortical areas have local intra-areal connections, usually with related "modules" or cortical columns. For example, in primary visual cortex, orientation columns will tend to have connections with nearby orientation columns with similar orientation preference. The self-connections in the matrix are really just a reminder of these local intra-areal (or intrinsic) connections.
>>Why is your matrix symmetrical?
Good question! Cortical connectivity is really not symmetrical, though oftentimes, for the sake of simplicity (and/or to mask uncertainties about the existence of bidirectional connections), the assumption of symmetrical connectivity is made. In my particular matrix, it was mainly a convenience factor, because to drop the symmetry assumption would mean there would be twice as many entries in the connectivity matrix I'd have to ascertain. I know it's a lazy man's thing to do. I mean, you can present "reasonable" arguments from the literature showing that cortical connectivity is symmetric, but anyone who sees the actual data from tracer injection experiments knows for a fact that cortical connectivity is not symmetric.
As a general rule, if cortical area A sends feedforward (or lateral) projections to cortical area B, then area B will likely send feedback (or lateral) projections to area A (within the context of a cortical hierarchy). However, examples of seemingly unidirectional (or asymetrical) cortical connections between areas abound in the literature. For example, there is evidence for asymetrical connections involving cingulate cortex and other areas (though the details escape my mind right now), which is interesting considering how high up cingulate cortex is in the cortical hierarchy.
>>Do you know also the strength (weight) of the certain connections?
It's difficult to assign numerical estimates of connectivity weights, though a scheme of 0 (no connection) to 4 (very strong connection) certainly seems do-able. I haven't done it, yet. Part of the difficulty, too, involves assigning scalar magnitudes to qualitatively different types of connections. For example, how do you assign numerical estimates to feedforward vs. feedback connections between V1 and V4? Most feedforford connections are "driving" and terminate in layer 4 and lower layer 3, but feedback connections are more "modulatory".... how would you assign scalar magnitudes to these different types of connections to characterize their strength? It seems that maybe classifying the connections into different types, and then assigning magnitudes, might be a solution. In this way, connections wouldn't be characterized by a single scalar weight, but rather by an n-tuple (i.e., a vector weight).
>>Why is there the difference between the matric of Felleman&Van Essen and your matrix (for ex. -connection between the regions V1 and V3)?
I would have to look at the Felleman and Van Essen matrix again from their '91 Cerebral Cortex paper. Their matrix is very old. I did use it, though, as well as Malcolm Young's (also Hilgetag and Scannell?) more recent connectivity matrix, when I was putting together mine. I had a lot of problems with the Felleman and Van Essen matrix. They grossly over-emphasize visual areas at the expense of somato-motor and auditory areas. They do not even have an area 46vr, which is a prominent somatosensory area in dorsolateral prefrontal cortex that's connected with multiple sensori-motor areas, including SII and premotor cortex (even though evidence for this connection was made very explicit by Preuss and Goldman-Rakic in '89 in Journal of Comparative Neurology, a full two years before Felleman and Van Essen paper came out; how they could not have known of this connection and of this prefrontal somatosensory area amazes me).
In general, I tried to be more detailed in my connectivity matrix than Felleman and Van Essen, in part, because the work done in the decade after their paper was published contained a lot of new connectivity data that was not part of their matrix.
>>I'm looking forward to hear from you
It's been a pleasure. I do intend to have a real primate cortical connectivity database available soon, one that's searchable and sortable. I have 439 references now, ranging from 1945 up to the current date, but I first have to get into them into a database for publishing to the web. I will have this available as soon as possible. That connectivity matrix I have up now is in need of updating too. At the very least, I need to incorporate asymmetrical connections into it (and maybe even weights, as you suggested).
Thank you very much for the answer, Shawn (I'm sorry that so late, you know, summer....) I'm really looking forward to see your connectivity matrix in the future.
I have another question about microanatomy of these connections. As far as I know these fibers (tracti) are formed by large number of axons bundled together. I'm interesting for the exact structure. Do the axons communicate between themselves in the one fiber? Can I take the one (mean) value of action potential generated by them (like activity of whole fiber) or are their AP completely independent? And what about the percentage of neurons from one brain region incorporated by their axons in the certain fiber?
Does anybody have some idea about these things or some hints where to find such information?
Thank you.
I have another question about microanatomy of these connections. As far as I know these fibers (tracti) are formed by large number of axons bundled together. I'm interesting for the exact structure. Do the axons communicate between themselves in the one fiber? Can I take the one (mean) value of action potential generated by them (like activity of whole fiber) or are their AP completely independent? And what about the percentage of neurons from one brain region incorporated by their axons in the certain fiber?
Does anybody have some idea about these things or some hints where to find such information?
Thank you.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.