Retrospectively supervised click decoder calibration for self-calibrating point-and-click brain-computer interfaces

Brain-computer interfaces (BCIs) can help people with motor disabilities control devices using their brain activity. However, changes in the brain signals can make it hard to maintain stable control over time. To address this, the researchers developed a method to automatically recalibrate the BCI click decoder during normal use. This method, called retrospectively supervised (RS) calibration, uses data from regular BCI use to relabel neural activity patterns as either "click" or "non-click". This relabeled data is then used to retrain the click decoder, improving its accuracy without requiring separate calibration tasks. A participant with ALS used this self-calibrating BCI to type freely for multiple sessions over a month. The results showed that the BCI maintained high performance without needing any interruptions for traditional calibration, suggesting that this approach could improve the practicality of BCIs.

• 1
  The study demonstrated a “retrospectively supervised” (RS) method for calibrating the click decoder of a brain-computer interface (BCI) using data acquired during practical BCI use.
• 2
  The RS calibration, combined with real-time bias correction and baseline firing rate tracking, enabled a participant with ALS to type freely across 11 research sessions spanning 29 days.
• 3
  The participant maintained high-performance neural control over cursor movement and click without needing to interrupt virtual keyboard use for explicit calibration tasks.