Content area
Brain-computer interfaces (BCIs) offer a way for people with quadriplegia to regain lost abilities and live more independently. BCIs have demonstrated various applications in laboratory research, such as controlling a computer cursor or a prosthetic limb. Additionally, they enable users to interact with a menu interface or operate assistive devices like wheelchairs. However, despite their potential in restoring motor function, performance is often affected by factors such as neural signal variability and electrode instability. In these studies, neural data were recorded using the Blackrock NeuroPort® Utah Electrode Arrays (UEAs) implanted in the brain of one human participant and two nonhuman primates (NHPs). Rather than focusing on a specific decoding algorithm, this research presents framework-based solutions to address challenges in BCI reliability, particularly neural signal variability and electrode performance.
First, this research integrates assistive vehicle technologies with BCIs, using recalibration to mitigate neural variability. An example is presented of recalibrating using labels assigned by the navigation system as an additional feedback source. Unlike conventional recalibration methods that rely solely on neural signals, this approach leverages external system outputs to enhance classifier adaptation, resulting in improved classification time and performance. These findings demonstrate the potential of recalibration in BCIs as an effective strategy for improving system reliability and addressing neural variability.
Second, decoder analysis revealed that although decoders are very effective in the movement state, they can still predict trajectories based on brain activity in the rest state, also because the brain does not stop its activity due to the termination of movement, causing the decoder to output unwanted trajectories (e.g., cursor drift), thus affecting the user experience. Based on this, a new signal processing framework was proposed to better exclude the impact of brain activity at rest state.
Finally, signals from the UEAs implanted in NHPs were statistically analyzed. The spike firing performance recorded by electrodes in the premotor and motor cortices was compared across different movement stages, with a focus on the spike patterns generated post firing. Hypotheses were proposed to investigate potential factors influencing electrode performance and implantation outcomes, such as implantation depth, cortical curvature, and boundary effects.