A Stentrode study decodes a small set of attempted characters

A study published on October 22 tests attempted-handwriting decoding in one person implanted with a Stentrode endovascular BCI. The decoder uses how high-gamma activity changes over time, rather than relying only on spatial differences across electrodes.

Across four online sessions, the best three-class model reached 65% character recall and 59% precision. Aligning trial durations improved performance compared with using raw timing.

This is a small-vocabulary selection result, not unrestricted handwriting transcription. It suggests that temporal structure can carry useful information through a limited recording interface, while leaving larger vocabularies, more participants and sustained communication to future studies.