Brain-Computer Interface (BCI) technology, a cutting-edge innovation explored by countries worldwide, holds groundbreaking potential for helping patients with severe motor disabilities restore communication and physical control. It may even extend to fields like speech synthesis and handwriting assistance. However, current BCI devices are bulky, power-hungry, and have limited practical applications.
Researchers from the Integrated Neurotechnologies Laboratory at EPFL’s (École Polytechnique Fédérale de Lausanne) IEM and Neuro X institutes have developed a groundbreaking miniaturized brain-computer interface (MiBMI). Their findings, published on August 23 in the IEEE Journal of Solid-State Circuits (DOI: 10.1109/JSSC.2024.3443254) and presented at the International Solid-State Circuits Conference, represent a significant leap forward.
The MiBMI is the world's first high-performance, miniature brain-computer interface system designed to provide patients with amyotrophic lateral sclerosis (ALS) a small, low-power, highly precise, and multifunctional solution. It enables direct brain-to-text communication on a chip, enhancing the efficiency and scalability of BCI technology. This innovation paves the way for fully implantable BCIs, potentially transforming the quality of life for patients suffering from ALS and spinal cord injuries.
Comprised of two ultra-thin chips with a combined area of just 8 mm², MiBMI is significantly smaller than devices like Neuralink’s 23 x 8 mm interface. Its compact size and low power consumption make it ideal for implantable applications, minimizing invasiveness while ensuring safety and practicality in clinical and everyday settings. MiBMI is a fully integrated system capable of both data recording and processing.
"MiBMI allows us to convert complex neural activity into readable text with high precision and low power," said Mahsa Shoaran, one of the project's lead researchers. "This advancement brings us closer to a practical, implantable solution that can significantly enhance communication for patients with severe motor disabilities."
By detecting neural signals when a person imagines writing, MiBMI’s electrodes, implanted in the brain, record neural activity related to handwriting movements. The chip then translates these neural commands into text in real-time.
This technology allows individuals, particularly those with locked-in syndrome or other severe motor impairments, to communicate by simply thinking about writing. Their thoughts are converted into text displayed on a screen for others to read.
Researchers discovered that each letter has a unique neural signature, which they named Distinct Neural Codes (DNC). Instead of processing the full data for each letter, the BCI system processes just the DNC (about 100 bytes), allowing for faster and more accurate performance while maintaining low power consumption. This breakthrough also reduces training time, making it easier for patients to learn how to use the BCI device.
Compared to traditional low-complexity linear discriminant analysis (LDA) systems without DNC, MiBMI’s decoder achieved significant breakthroughs in memory efficiency (~100x) and computational complexity (~320x).
Lead researcher Mohammed Ali Shaeri noted that MiBMI was able to convert these handwriting neural signals into text with 91.3% accuracy. In an in-vivo experiment, researchers also decoded the neural response of rats to six types of acoustic stimuli with 87% accuracy.
MiBMI can decode 31 different characters, a feat unmatched by any other integrated system. "We believe we can decode up to 100 characters, but we currently lack enough handwriting data samples for that," Shaeri added.
Existing BCI devices typically rely on external computers to decode signals, with electrodes implanted in the brain solely responsible for signal collection. In contrast, MiBMI processes data while recording it. It integrates a 192-channel neural recording system with a 512-channel neural decoder, marking a breakthrough in extreme miniaturization, combining expertise in integrated circuits, neuroengineering, and AI.
"Our goal is to develop a universal BCI that can be customized for various neurological disorders, offering a broader range of solutions for patients," Shoaran concluded.
