EPFL Develops Innovative Machine Learning Technique for Neural Prostheses
Researchers at EPFL (École Polytechnique Fédérale de Lausanne) in Switzerland have made a significant breakthrough in the field of neural prostheses with a groundbreaking machine learning technique. Their innovation aims to enhance the functionality of retinal implants and other medical electronic devices, revolutionizing the process of sensory encoding.
Neural prostheses are designed to convert environmental information into neural signals that can be interpreted by the human nervous system. One of the challenges in this field has been downsampling, which is the process of reducing environmental input while preserving data quality. Traditional downsampling methods lack a learning component, but EPFL’s new AI-based approach fills this gap.
The research team, led by Demetri Psaltis, Christophe Moser, and Diego Ghezzi, developed a novel method using two neural networks within an actor-model framework to optimize sensory encoding. The model network acts as a digital twin of the retina, learning to convert high-resolution images into a binary neural code similar to the biological retina. The actor network then learns to downsample high-resolution images, closely mimicking the neural code produced by the biological retina.
Preliminary results have shown great promise, demonstrating that the unconstrained neural network can successfully mimic aspects of retinal processing. This breakthrough not only offers a more precise image compression technique for visual prostheses but also has the potential to be applied to other types of neural prostheses, such as auditory or limb prostheses. The implications of this research go beyond vision restoration, redefining our approach to sensory prostheses as a whole.
The EPFL researchers aim to continue refining their machine learning technique and explore its applications further. The advanced AI-based method has the potential to revolutionize the field of neural prostheses, improving the quality and functionality of medical electronic devices.
