🔗 ohbm-environment.org/ohbm-satelli...

This approach opens doors to adaptive, individualized BCI training protocols informed by model-tracked neural changes.

Huge thanks to the FACE Foundation, our teams and institutions for their support: IIT Madras, @nerv-lab.bsky.social @institutducerveau.bsky.social

These parameter shifts are robust across EEG and MEG, and localize to sensorimotor regions—crucial hubs for motor imagery in BCI.

We introduce mi-NMM: a linear neural mass model that nails power spectral density shapes—both at rest and during motor imagery tasks.
It tracks intra-regional connectivity strength and E/I population dynamics, revealing how training reshapes neural activity.

These parameter shifts are robust across EEG and MEG, and localize to sensorimotor regions—crucial hubs for motor imagery in BCI.

We introduce a linear neural mass modeling approach, mi-NMM, that accurately captures power spectral density shapes in resting state & while performing a motor imagery task. It tracks intra-regional connectivity strength and E/I population dynamics, revealing how training reshapes neural activity.

Key findings:
✅ Neuronal avalanches are robust biomarkers of individual BCI learning.
✅ Features like avalanche duration and spatiotemporal size correlate with BCI performance across sessions.
✅ Longitudinal models using these features achieve up to 91% accuracy in predicting BCI success.

Why it matters?
Up to 30% of users struggle with motor imagery-based BCI, a challenge known as "BCI inefficiency." Current protocols use fixed-length sessions, ignoring individual variability. Our study introduces a novel approach that rely on neuronal avalanches to tackle this issue.

Main features of the workshop are available through the dedicated github page (that will be updated with some surprises soon!): github.com/mccorsi/BCI-...

José del R. Millán from the University of Texas presented his latest findings on transfer learning based on Riemannian geometry and transcutaneous electrical spinal stimulation to foster BCI learning.