What if, in the not-too-distant future, people with partial or total blindness due to degenerative retinal diseases could regain their sight with unexpected quality? That is the mission undertaken by doctoral student Marwan Besrour and his team at the Groupe de recherche en appareillage médical de Sherbrooke (GRAMS).
Photo: Provided
Driven by a desire to contribute to society's well-being, Marwan, originally from Tunisia, chose to pursue his PhD in electrical engineering at Sherbrooke, motivated by the retinal implant project using artificial intelligence led by Professor Réjean Fontaine, from the Department of Electrical Engineering and Computer Engineering. These implants are intended to restore sight to people suffering from macular degeneration or retinitis pigmentosa.
A giant leap
Between the first retinal implants, commercialized in the 2010s and allowing patients to perceive light "flashes" corresponding to objects or lights in their environment, and the implant developed by Marwan and his team, there is a giant leap.
The problem with these early implants lies in visual acuity, Marwan explains. Of course, people who had lost their sight were happy to see something, but the resolution remained poor and the vision was essentially made up of irregular flashes with random contours.
The retinal implant developed at GRAMS, on the other hand, attempts to mimic the complexity of the human eye's functioning as closely as possible using second-generation artificial intelligence. "The retina works like a symphony orchestra where each cell would be a musician, each color or detail of an image would be a musical note, and each instrument an image function. To see clearly, every note must be played at the right time by the right instrument."
Indeed, retinal neurons interpret colors, brightness, and contrasts according to temporal and spatial settings. An artificial neuron, however, is based on an abstract model and certainly does not possess the full complexity and subtlety of a real neuron. Therefore, several artificial neurons are needed to replace a single biological neuron. "So we are working to make our implant mimic these temporal and spatial aspects of the retina as closely as possible in order to obtain images that are as close to reality as possible."
To achieve this, the team uses a complex network of 48 artificial neurons placed on a very small chip, itself attached to the implant that will be inserted at the back of the eye, to send electrical signals to the retina. "These artificial neurons capture images like a video, then coordinate this visual information to obtain what I call a retinal symphony, without false notes. To do this, we trained them to recognize each component of thousands upon thousands of images, so that the network can activate the right neuron at the right moment and thus the brain can understand the information."
Photo: Provided
When will it be commercialized?
"Sometimes, after my presentations, people contact me to ask if they can get our implants," Marwan says. "However, commercialization is not imminent, because we want to achieve a very high-quality product that will allow people to regain sight comparable to what you could see on the screens of early flip phones."
"First-generation implants used 16 electrodes," he continues. "Our implants use 288, which will allow people to distinguish shapes much more clearly, such as door frames, cars, and street signs. But our goal is to reach 10,000 electrodes. That will take some time."
It is no surprise that the doctoral student aspires to become a professor at UdeS and recruit students to whom he can pass on his passion in order to push this project further. "Professor Réjean Fontaine told me: you are laying the first stone of a great pyramid," he concludes. For Marwan Besrour, this project could well become the project of a lifetime.