A BCI review makes generalization the clinical test

A review published on October 23 argues that the central test for a clinical BCI is whether performance survives new sessions, users, tasks and hardware. Strong accuracy during development is insufficient if a system requires frequent recalibration afterward.

The review organizes evidence by what was held out during evaluation, what target-user data were available and how adaptation was performed. Sustained performance in one person is different from transfer to a new participant.

That makes evaluation details part of the result. A claim that a decoder generalizes is only useful when the reader can see which conditions actually changed and how much new calibration the system received.