Researchers at the École Polytechnique Fédérale de Lausanne (EPFL) have made significant strides in developing next-generation visual prosthetics. By employing specially designed artificial intelligence (AI) models, they have achieved a high level of precision in predicting which areas of the brain’s cortex require electrical stimulation to generate complex visual perceptions.
The challenge lies in the intricate architecture of the higher visual cortex, which is responsible for recognizing complex objects such as cars and human faces. Scientists previously struggled to determine the optimal points and intensity for electrical stimulation.
Under the leadership of Johannes Merero, the research team addressed this issue using AI:
- Topographic Neural Networks: The researchers developed a virtual model that simulates millions of stimulation parameter combinations, significantly reducing the time and financial resources typically required for laboratory testing.
- Response Prediction: The AI analyzes the desired image and generates an optimal pattern of electrical impulses to achieve specific visual perception.
Practical applications of these computer simulations were validated by a team from Amsterdam, who tested the AI-generated stimulation patterns on two primates with pre-existing brain implants.
The results were promising: The AI accurately predicted how stimulation would influence the animals’ behavior during object recognition tasks.
At this stage, researchers have learned to adjust and transform the images that the subjects already perceive. The next major milestone will involve generating a complex visual image from scratch—essentially creating a clear picture in the brain without any input from the eyes.
Future Directions: Addressing Hearing Impairments
Martin Schrimpf, head of the NeuroAI laboratory, noted that the success of this model paves the way for advancements in other types of prosthetics, particularly auditory devices. Despite the effectiveness of current cochlear implants, they still fall short of perfect sound reproduction.
With a new financial grant from the Horton Health Foundation, the research team is now beginning to adapt their topographic AI models for stimulating the auditory cortex, aiming to assist individuals with severe hearing impairments.
Researchers at EPFL have developed AI models that enhance the precision of brain stimulation for visual prosthetics. The success of these models opens avenues for future advancements in auditory prosthetics, aiming to assist those with hearing impairments.