Research on a novel Real-Time Neural Network platform "BRAND" published in the Journal of Neural Engineering

Researchers, including CNTR faculty members, recently published a paper in the Journal of Neural Engineering that uncovers a state-of-the-art tool for processing Artificial Neural Networks (ANNs) in closed-loop experiments with minimal latency. ANNs are used to model and decode neural activity, but their integration in closed-loop experiments with tight time constraints has previously been a challenge. A platform is needed to support high-level languages such as Python and Julia on which ANNs run, while also supporting languages with low latency data acquisition and processing in languages such as C and C++. The groundbreaking platform known as the Backend for Realtime Asynchronous Neural Decoding, or BRAND, was developed to address some of these challenges.

BRAND has been shown to be successful in processing and sending large amounts of data with minimal latency. In a real-word application of the platform, a participant in the brain-computer interface BrainGate2 clinical trial performed a cursor control task where data processing, neural decoding, task control, and graphics were all carried out on BRAND. From the input of neural activity to the decoder prediction, the BRAND platform performed with a latency of less than 8 ms. 

Through the fast and reliable framework created by BRAND, it is evident that significant efforts are being made to integrate neuroscience and machine learning techniques in closed-loop experiments.