Stroke Rehab BCI Replay

Session

Press play

Trial 1

C3 (over the left motor area, runs the right hand) C4 (over the right motor area, runs the left hand) shaded = imagery window the decoder reads

decoder: RIGHT handwaitingLEFT hand

LEFT hand

RIGHT hand

Feedback rule

This session

--0 of 0 trials read correctly

Why it matters

More than 795,000 people in the US have a stroke each year (CDC). When the hand will not move, ordinary therapy has nothing to practice and the brain gets no sign that its attempt did anything.

The BCI reads the attempt from the brain and answers it right away, on screen or with a small pulse that moves the real muscle. That restores the "I tried, it happened" loop that rehab relies on.

All six sessions (unseen test runs)

Patient	First	Last	Change

Chance is 50%. Changes of -6, +18 and +2 points across 3 people do not show that therapy worked.

What this is, and is not

Real: recordings from 3 stroke patients imagining left or right hand movement. The decoder (standard CSP + LDA) was built from each session's training run and scored on the test run it never saw.
Simulated: the hand and the stimulation. No patient, no stimulator.
Not shown: that the therapy improves recovery. Not new either: g.tec's recoveriX already does this clinically.
Cost: each session needs about 17 minutes of calibration recording.
Protocol note: in the recorded therapy, feedback was given only when the detection matched the instruction. A real home system cannot know the instruction, so switch to "Real use" to see wrong guesses move the wrong hand.