Improving motor imagery decoding methods for an EEG-based mobile brain-computer interface in the context of the 2024 Cybathlon.

Only two pilots were included, so the results cannot show how a wider group of clients would perform. The reported outcomes are classifier accuracy, game success, and latency, not daily function. Offline accuracy was high, but real competition performance was limited: the pilot solved one task in 8 minutes. The total delay was 1617 ms, while the paper notes that delays below 250 ms are recommended for closed-loop systems. The classifier needed about 30 min of retraining before each online session. The feedback screen was not used during the competition because the pilot had to focus on the official game display. The tasks were software games, and the paper notes hardware applications mostly do not allow smaller errors without impacting safety.

Funding was provided by Technische Universität München. No other conflicts of interest are reported.

Do not read the 84% or 93% offline accuracy as evidence that this BCI restores daily function for tetraplegic clients. The study had two pilots, and in competition the pilot solved only one task in 8 minutes.