May 15 2020

Stimulating the Visual Cortex

For adults who had vision but then lose it due to eye disease or damage to the optic nerve, their visual cortex is still intact. It is deprived if input, but is theoretically capable of functioning normally to create images. The ultimate technological expression of this potential would be something like the visor of Geordi La Forge – a device that can see (even in frequencies and particles humans cannot normally see) and transfer that information to the visual cortex. Obviously we are a long way away from any such technology, but we have taken the first baby steps in that direction, including a recent study which makes one tiny advance (but more on that shortly).

Early research into this approach involved animals and simply tried to determine if the monkeys could “see” the stimulation. Often simple behaviors, like moving their eyes, were used to see if the stimulation was having any effect. Some of the research also comes from trying to map the visual cortex, not necessary allow the blind to see. This research has been encouraging, because it shows that the primary visual cortex is arranged in a way to reflect the images it sees (so-called retinal mapping). The neurons, in short, are like a bitmap of an image. So if you stimulated a circle of neurons in the primary visual cortex, subjects would see a circle.

Of course there is more to vision than the primary visual cortex. A lot of processing occurs in the nervesĀ  and pathways carrying information to the cortex. After the basic image is formed it is then sent to higher visual cortical areas for further processing, so a two-dimensional image is given shape, shading, movement, distance, three-dimensionality, and ultimately meaning. But hopefully we wouldn’t need to worry about all that higher levels of processing because once the image is presented to the primary visual cortex, the rest should take care of itself.

So far researchers have tried to stimulate the visual cortex with an array of electrodes, either on the scalp or implanted on the brain surface. Often the brain surface subjects are those who are getting electrodes placed anyway as part of their epilepsy management. Actually the first time this was done was in 1968, by Dr. Giles Brindley. Then Dr. William Dobelle took inspiration from this preliminary research and in the 1970s conducted extensive research on stimulating the visual cortex to see if limited sight could be restored to the blind. So actually this research has already been going on for over half a century. Enthusiasts hearing about this research in the 1970s would probably be disappointed to learn that we have made precious little progress in 50 years – we are still struggling to get subjects to see basic shapes by stimulating the visual cortex.

What went wrong? Why has progress been so slow? I think the primary reason is technological. In all of these studies arrays of dozens of electrodes are used. Meanwhile there are an estimated 140 million neurons in the primary visual cortex (V1). What ends up happening is that subjects see indistinct blobs, not the letters or shapes that are being stimulated. There has been some limited success, but nothing yet achieving a useful technology. The ultimate “Geordi La Forge” goal would be to stimulate individually each of those 140 million neurons. But the question is – where is the threshold for useful vision? Do we need hundreds of electrodes, thousands, tens of thousands? And what constitutes useful vision?

This brings us to the current study, which does not really address the underlying problem, but does provide perhaps a useful “hack” that may lead to some practical application. The researchers report that instead of stimulating a static shape, they used dynamic stimulation. (As an aside, although the study is now published, I found a reports of this same research from two years ago.) Instead of just stimulating electrodes in the shape of a “Z”, they stimulate the electrodes in sequence, dynamically forming the shape of a “Z”. They report that their six subjects were better able to identify the letters imaged in this way than through static stimulation. Dynamic stimulation, therefore, may get us to the point more quickly that subjects can be artificially stimulated to see shapes, letters, and numbers. Whether this will prove practical remains to be seen.

This result does make me wonder if this hack can be exploited in more subtle ways. If you have a picture of a more complex object, for example, like a horse, would it help to dynamically stimulate the pixels making up the horse in some way? Perhaps the blob problem comes from tonically stimulated neurons just bleeding over. If the pixels were instead flashing on and off or moving in some way the image might be more clear in the cortex. If this works out, this discovery could prove to be one small piece of the full technology necessary to artificially produce useful images in the cortex of blind subjects.

Here is another piece of speculation – we know that our eyes are constantly moving. The primary known reason for this is so that our retina does not burn out, meaning that if the same retinal neurons are stimulated over time they uses up all their neurotransmitter can stop working. So our eyes move, so that retinal neurons have a chance to reset. You can test this by staring at one spot and fixing your gaze – the image tends to fade away over time. But perhaps the visual cortex also needs this constant movement. At the very least it evolved in adaptation to this constant movement. Dynamic stimulation may be more the norm for the visual cortex.

Again – one tiny baby step, but hopefully we are moving in the right direction.

I have to say, however, that I think we are only going to make real progress when the technology significantly improves, by orders of magnitude. We are making progress in this areas as well. I wrote not too long ago about using microwires for brain electrodes. This is the kind of direction we need to go in, some technology to effectively produce thousands of electrode connections to brain cells – or more. We also need the electrodes to be stable in their connection and last a long time.

This research is also a good example of how an idea can stall simply because the technology is not ready. Fifty years with little practical progress in sobering. It reinforces the notion that predicting future technology is really difficult, especially when it comes to overcoming roadblocks. Artificial vision can join the jet pack and flying car as technologies dreamed of for decades, but not practical because of technological limitations.


No responses yet