Right. And I think that if you put the *right input* into V1 (which means taking the visual scene and putting it through the right set of filters - something we could in principle do in silicon now), then you may be able to support a reasonable sense of visual perception.
To me the most important thing that is being missed is any subcortical role in visual attention. However, it could be that this might work out anyway, because of feedback connections from V1 back to LGN. In essence, these would take over the role of the minor (joking - but it is only 10%) projection from the retina to the LGN.
Of course, if you have the option of using the built-in hardware - retina, optic tract etc - you do it.
And by the way, you gave three examples which illustrate why there wouldn't be much point trying to reproduce the subcortical visual pathway in a visual prosthesis anyway.
* eye movements: we're talking about an artificial retina, eye movements don't help (or will have to be done by an entirely artificial eye-movement system, such as that being developed by Chris Diorio). * circadian rhythms... if you can't do without 'em, it's probably easier to just hook up a clock circuit, make it actually time-based rather than light-based. * pupil size, again no use if the eye isn't being used as the input device.
Ah, but we were talking about projections from the LGN to cortex.
The subcortical visual pathway was/is being discussed in response to another poster in the thread. Yes, of course it exists, its just not very important for visual perception.
Does this apply specially to the brain rather than the body in general? In the latter case, I'd mention as a counter-example the appendix...
Where is this mysterious place that the LGN projects to apart from V1, anyway? While I accept that it receives from a lot of places, I understood that its output is pretty much focused on layer IV of primary visual cortex.
That is arguable. See recent work by Paul Azzopardi and Alan Cowey: http://www.ncbi.nlm.nih.gov:80/entrez/quer y.fcgi?c md=Retrieve&db=PubMed&list_uids=11133785&dopt=Abst ract
In any case, lesioning V1 in that case does abolish the conscious perception of motion - all that may be left is some residual ability to say if something moved - while all the while denying you saw something.
There is no way in hell MT receptive fields will be normal after a V1 lesion, I am sure you will agree.
The question here is whether you need to bother with the subcortical visual pathway for a visual prosthesis - I say not.
Not very much. For practical purposes the LGN really only projects to layer IV of V1. It is from there that information spreads out through the cortex (including going back to other parts of the thalamus such as the pulvinar, and then on to other visual areas). The important thing that is being missed is not so much what the LGN is sending elsewhere, but what it is receiving (all the neuromodulatory input from the brainstem and back from cortex). Of course, all this stuff could be simulated too, but at some point it becomes not worth it.
PS It is true that the "subcortical visual pathway" via the superior colliculus is ignored. But I really doubt that matters.
I agree with most of what you say - however, repairing the optic nerve is not the *only* way to get real vision. Another way would be to reproduce retinal and LGN processing in silicon, and use that to drive stimulation of the primary visual cortex. Not that they do that in the study mentioned, but they should start, since a bit of preprocessing could increase the quality of the vision substantially, I suspect.
I don't think that this approach is as good as either (a) stimulating the optic nerve, (b) stimulating the retina, or (c) repairing the optic nerve/retina, if those can be done, but presumably they can't be done in all circumstances. Also, unless someone comes up with decent trascutaneous wireless transmission and a way to power an autonomous device, it is going to have to have a chronic (permanent) interface through the skull and skin. This is going to be a long-term nightmare in terms of infection (and the results of infection to the brain are *real* nasty), and would never fly commercially for lawsuit reasons.
Where does this stuff come from? Quartz anti-aliasing is not even subpixel sampled, it is just plain old-fashioned pixel anti-aliasing. Look closely at a Mac LCD display - if it was using pixel subsampling, you'd see little coloured bits around the edge. You don't.
The ClearType technology used in Windows XP is superior, being based on subpixel sampling. Hopefully Apple will include it in a future release of OS X - apparently they have access to the patent, but I suspect they have not been able to make it run fast enough to include so far. This may change with hardware acceleration in 10.2.
I'm going to start putting together a new linux box sometime in the next year, and am still open to the choice Intel vs Alpha. I want the box to be really spanking fast, of course, and it sounds like alpha is *still* going to be outperforming Intel even with the Merced (which is disappointing in a way, because I thought the whole point was that Intel were finally going to have a chip that could compete with alpha on grunt speed). But what worries me is: you can only buy Quake 3 in binary form, and I think I saw somewhere on the quake website that they're *only* releasing it for Intel. Plea: *please* release Quake3 for Alpha...!
Neither voice nor handwriting input would change much the `human-computer communication bottleneck'. Which is a very big bottleneck: while we receive very large amounts of information from the computer through our visual system, we talk back to it at (using a keyboard) only ~ 5 bits per second (much less than you got with even the very first modems!). [English text has around 1.5 bits per character, say there are on average 5 characters per word, and we type at 40 words per minute (for me only when typing gibberish)].
The answer I think is to sacrifice some small part of our primary visual cortex (the bit of the brain your little finger is over when you're sitting back with your hands behind your head) to interfacing with the computer. By surgically implanting an electrode chip with a radio transmitted that had just enough power to transmit through bone, and then letting it heal over so there was no risk of infection (which would happen if you had wires hanging out of your skull).
These electrodes would then be used to decode the nerve cell activity in a small patch of brain tissue. By visualising characters in that part of your visual field, you could communicate with the computer [the jury's still out on whether 'visual imagery' reaches all the way down to the primary visual cortex, but I hope it does].
The end result might be that we could have a unix command-line interface in a small strip in the lower part of our visual field, where we just visualise command-words. Of course, English is a pretty inefficient language for communicating in this way, so performance would be improved by operating with a specifically optimised alphabet (even Chinese would be much more efficient for this, you'd think).
Slashdot Mirror
That's cool.
My girlfriend bought me a lego kit, and I'd planned to build essentially the same thing - but I haven't had time to because of my research...
The $100 special edition gold coins just go for *miles*. Damned expensive hobby, though...
OK, so instead lets just ignore the subcortical pathway and do it the easy way...
Right. And I think that if you put the *right input* into V1 (which means taking the visual scene and putting it through the right set of filters - something we could in principle do in silicon now), then you may be able to support a reasonable sense of visual perception.
To me the most important thing that is being missed is any subcortical role in visual attention. However, it could be that this might work out anyway, because of feedback connections from V1 back to LGN. In essence, these would take over the role of the minor (joking - but it is only 10%) projection from the retina to the LGN.
Of course, if you have the option of using the built-in hardware - retina, optic tract etc - you do it.
Maybe. How about we agree to hold off building such a system until we actually know what the subcortical visual projection does? :-).
And by the way, you gave three examples which illustrate why there wouldn't be much point trying to reproduce the subcortical visual pathway in a visual prosthesis anyway.
... if you can't do without 'em, it's probably easier to just hook up a clock circuit, make it actually time-based rather than light-based.
* eye movements: we're talking about an artificial retina, eye movements don't help (or will have to be done by an entirely artificial eye-movement system, such as that being developed by Chris Diorio).
* circadian rhythms
* pupil size, again no use if the eye isn't being used as the input device.
Ah, but we were talking about projections from the LGN to cortex.
The subcortical visual pathway was/is being discussed in response to another poster in the thread. Yes, of course it exists, its just not very important for visual perception.
Does this apply specially to the brain rather than the body in general? In the latter case, I'd mention as a counter-example the appendix...
Where is this mysterious place that the LGN projects to apart from V1, anyway? While I accept that it receives from a lot of places, I understood that its output is pretty much focused on layer IV of primary visual cortex.
That is arguable. See recent work by Paul Azzopardi and Alan Cowey:r y.fcgi?c md=Retrieve&db=PubMed&list_uids=11133785&dopt=Abst ract
http://www.ncbi.nlm.nih.gov:80/entrez/que
In any case, lesioning V1 in that case does abolish the conscious perception of motion - all that may be left is some residual ability to say if something moved - while all the while denying you saw something.
There is no way in hell MT receptive fields will be normal after a V1 lesion, I am sure you will agree.
The question here is whether you need to bother with the subcortical visual pathway for a visual prosthesis - I say not.
Not very much. For practical purposes the LGN really only projects to layer IV of V1. It is from there that information spreads out through the cortex (including going back to other parts of the thalamus such as the pulvinar, and then on to other visual areas). The important thing that is being missed is not so much what the LGN is sending elsewhere, but what it is receiving (all the neuromodulatory input from the brainstem and back from cortex). Of course, all this stuff could be simulated too, but at some point it becomes not worth it.
PS It is true that the "subcortical visual pathway" via the superior colliculus is ignored. But I really doubt that matters.
I agree with most of what you say - however, repairing the optic nerve is not the *only* way to get real vision. Another way would be to reproduce retinal and LGN processing in silicon, and use that to drive stimulation of the primary visual cortex. Not that they do that in the study mentioned, but they should start, since a bit of preprocessing could increase the quality of the vision substantially, I suspect.
I don't think that this approach is as good as either (a) stimulating the optic nerve, (b) stimulating the retina, or (c) repairing the optic nerve/retina, if those can be done, but presumably they can't be done in all circumstances. Also, unless someone comes up with decent trascutaneous wireless transmission and a way to power an autonomous device, it is going to have to have a chronic (permanent) interface through the skull and skin. This is going to be a long-term nightmare in terms of infection (and the results of infection to the brain are *real* nasty), and would never fly commercially for lawsuit reasons.
To clarify: I meant you should see little coloured bits around the edges of font characters if you look very closely.
Where does this stuff come from? Quartz anti-aliasing is not even subpixel sampled, it is just plain old-fashioned pixel anti-aliasing. Look closely at a Mac LCD display - if it was using pixel subsampling, you'd see little coloured bits around the edge. You don't.
The ClearType technology used in Windows XP is superior, being based on subpixel sampling. Hopefully Apple will include it in a future release of OS X - apparently they have access to the patent, but I suspect they have not been able to make it run fast enough to include so far. This may change with hardware acceleration in 10.2.
"Greater civilian casualties than any war in US history"
So, how many people were killed in Hiroshima and Nagasaki again?
You have the right to bear arms ...
We have the right to bare arms.
-------------
ans =
NaN
You mean New Psientist?
I'm going to start putting together a new linux box sometime in the next year, and am still open to the choice Intel vs Alpha. I want the box to be really spanking fast, of course, and it sounds like alpha is *still* going to be outperforming Intel even with the Merced (which is disappointing in a way, because I thought the whole point was that Intel were finally going to have a chip that could compete with alpha on grunt speed). But what worries me is: you can only buy Quake 3 in binary form, and I think I saw somewhere on the quake website that they're *only* releasing it for Intel. Plea: *please* release Quake3 for Alpha ...!
Neither voice nor handwriting input would change much the `human-computer communication bottleneck'. Which is a very big bottleneck: while we receive very large amounts of information from the computer through our visual system, we talk back to it at (using a keyboard) only ~ 5 bits per second (much less than you got with even the very first modems!). [English text has around 1.5 bits per character, say there are on average 5 characters per word, and we type at 40 words per minute (for me only when typing gibberish)].
The answer I think is to sacrifice some small part of our primary visual cortex (the bit of the brain your little finger is over when you're sitting back with your hands behind your head) to interfacing with the computer. By surgically implanting an electrode chip with a radio transmitted that had just enough power to transmit through bone, and then letting it heal over so there was no risk of infection (which would happen if you had wires hanging out of your skull).
These electrodes would then be used to decode the nerve cell activity in a small patch of brain tissue. By visualising characters in that part of your visual field, you could communicate with the computer [the jury's still out on whether 'visual imagery' reaches all the way down to the primary visual cortex, but I hope it does].
The end result might be that we could have a unix command-line interface in a small strip in the lower part of our visual field, where we just visualise command-words. Of course, English is a pretty inefficient language for communicating in this way, so performance would be improved by operating with a specifically optimised alphabet (even Chinese would be much more efficient for this, you'd think).