الخميس، 21 مايو 2020

New Bionic Eye Might See Better Than We Do

The ability to restore sight to the blind is one of the most profound acts of healing medicine can achieve, in terms of the impact on the affected patient’s life — and one of the most difficult for modern medicine to achieve. We can restore vision in a limited number of scenarios and there are some early bionic eyes on the market that can restore limited vision in very specific scenarios. Researchers may have taken a dramatic step towards changing that in the future, with the results of a new experiment to design a bionic retina.

The research team in question has published a paper in Nature detailing the construction of a hemispherical retina built out of high-density nanowires. The spherical shape of the retina has historically been a major challenge for biomimetic devices.

EyeComparison

Light enters the eye through the lens, which is curved — which means the light that hits the retina has already been curved. When you use a flat sensor to capture it, there’s an intrinsic limit to how much the image can be focused. This seems like the sort of thing cutting-edge AI might be able to help with, but the amount of processing power available at the back of a human eyeball is limited and the latency requirement for vision is pretty much nil. Alternatively, we could solve the hemisphere problem. That’s what Zhiyong Fan, an electronic and computer engineer at the Hong Kong University of Science and Technology, and the rest of the research team did.

They started with a hemisphere of aluminum foil (as one does). Electrochemical treatment transformed the foil into an insulator known as aluminum oxide, and left it studded with nanoscale pores across its service. These densely-clustered holes became the channels for the perovskite nanowires that mimic the function of the retina itself. Perovskite is used in the manufacture of solar cells. Once the nanowires grew, the researchers capped the eye with an artificial lens and filled it with an ionic liquid to mimic the vitreous humor in our own eyeball.

This ionic liquid is important to the process, allowing the nanowires to detect light and transmit its signals to external, image-processing electronics.

The performance of the artificial eye is impressive. Because it isn’t limited by the biological parameters of our own lens, it can respond to wavelengths of light up to 800nm. The human visual range tops out around 740mm; colors above this wavelength appear black to us. If we could see at 800nm, we’d be seeing into the near-infrared band (considered to be 750 – 1400nm). Processing time for light patterns is ~19ms, or half the time of the human eye. Cutting the eye’s reaction speed to 19ms might reduce total human reaction time — and the artificial eye’s image sharpening and overall clarity were better than those produced by the Mark I Eyeball.

Note: Do not read that as a comment on the nature of frame rates and whether humans can see above a particular framerate threshold. Measured response and recovery times on the human eye range from 40ms to 150ms. Average total human reaction time is between 200ms and 250ms. Exceptional individuals sometimes exceed these speeds; 150ms reaction times are not unknown.

In short, this artificial retina sees better than we do in multiple respects, and as far as I’m aware, this is the first time anything like it has been built. The new retina even lacks a blind spot.

The Long Road Ahead

As Scientific American details, there’s a lot of work to do before a system like this could be integrated into a functional device. Systems like Second Sight (a company we’ve covered before) integrate directly with the brain. This artificial retina doesn’t, which is why I haven’t referred to it as a bionic eye. It is a proof-of-concept artificial retina that might one day be deployed in a bionic eye, provided current problems can be overcome.

Overcoming those problems is going to be difficult. The human visual system is not a camera, even if it can be conceptually described in similar terms. The idea that we’d benefit from the features the sensor offers implicitly assumes we can connect it to the brain seamlessly enough to allow these benefits to manifest. Because there are different forms of blindness, solutions that work for one type may not work for another. Blindness caused by brain damage would be unlikely to be helped by this kind of solution — even a flawless artificial eye won’t let us restore sight to every single person.

Still, the long-term potential here is tremendous. It’s been less than a decade since the first grayscale, low-resolution artificial sensors came to market. Now we’re trying to figure out how to build a plausibly superior system and connect it to the server backend, if you’ll pardon the metaphor. Hopefully we’ll see further advances in the field over the next decade.

Now Read:



sourse ExtremeTechExtremeTech https://ift.tt/36vEmM9

ليست هناك تعليقات:

إرسال تعليق