From Minutes to Days: Scaling Intracranial Speech Decoding with Supervised Pretraining

Decoding speech from brain activity has typically relied on limited neural recordings collected during short and highly controlled experiments. Here, we introduce a framework to leverage week-long intracranial and audio recordings from patients undergoing clinical monitoring, effectively increasing the training dataset size by over two orders of magnitude. With this pretraining, our contrastive learning model substantially outperforms models trained solely on classic experimental data, with gains that scale log-linearly with dataset size. Analysis of the learned representations reveals that, while brain activity represents speech features, its global structure largely drifts across days, highlighting the need for models that explicitly account for cross-day variability. Overall, our approach opens a scalable path toward decoding and modeling brain representations in both real-life and controlled task settings.

New modalities of cortical electrophysiology, perspectives in medical research and human physiology

Recent advances in material technology and in micro- and nano-electronics have profoundly changed the design of intracranial electrophysiology electrodes. It is now possible to manufacture electrodes that record cortical activity at a spatial resolution that was previously unthinkable. This high spatial resolution enables recording of the functional structures of the brain, and differentiation of the activity of the different types of neurons composing them. In this paper, we present a review of the different types of electrodes now available, and then suggest one of the first applications for such high resolution electrodes, namely a means to better characterise the mechanisms that generate focal seizures in epileptics. Finally, we reflect more broadly on prospects for their future use.