Reading Imagined Letter Shapes from the Mind's Eye Using Real-time 7 Tesla fMRI

We present a 7 Tesla fMRI proof-of-concept study of the first letter speller BCI that decodes imagined letter shapes from activity patterns in early visual cortical areas. New tools are developed to enable real-time population receptive field retinotopic mapping for encoding and decoding. Using two different letter shapes (H and T), classification performance of generated activity patterns during imagery reaches 80% accuracy in each individual. Using a denoising autoencoder, recognizable letter shapes could be reconstructed and displayed as feedback to participants in the scanner.

Automated summary

This study presents a proof-of-concept research using a 7 Tesla fMRI to decode imagined letter shapes in the first letter speller BCI. New tools were developed to enable real-time retinotopic mapping for encoding and decoding. The results showed that the classification performance of generated activity patterns during imagery reached 80% accuracy in each individual using two different letter shapes.