Brain activity before stimulation predicts treatment responses

Brain stimulation can produce very different effects at different moments, even when the stimulus is delivered the exact same way. A new study published in Brain Stimulation, based on detailed EBRAINS-hosted brain data, shows that the brain's ongoing activity before stimulation can predict much of this variability, offering a path toward more reliable neuromodulation therapies.

Brain stimulation is a therapeutic and research technique that uses electrical or magnetic fields to modulate neural activity, helping treat disorders such as depression, Parkinson's and epilepsy. Yet even when the stimulation site and parameters are identical, brain responses can vary several-fold from one trial to the next – a problem that undermines both research reproducibility and therapeutic reliability. This study asked whether the brain's ongoing state at the moment of stimulation could explain that variability, and if so, which features provide the most reliable predictors.

To answer this question, the researchers analysed an EBRAINS-hosted dataset, from the EBRAINS Knowledge Graph, comprising simultaneous intracranial (SEEG) and scalp EEG (hd-EEG) recordings from 36 epilepsy patients who underwent intracranial electrical stimulation as part of clinically indicated seizure-mapping procedures. Across roughly 320 sessions and more than 10,000 individual stimulations, the team tested 125 different measures of pre-stimulus brain activity – including markers of network synchronisation, functional connectivity, and signal complexity – to determine which ones best predicted the brain's response to stimulation.

The results showed that whole-brain activity patterns predicted stimulation outcomes far better than local, single-site recordings, and that predictability was strongest when the sensorimotor and visual networks were targeted. Restricting stimulation to trials with specific pre-stimulus signatures reduced response variability by up to 24%, suggesting that timing stimulation according to the brain's ongoing state could make interventions more consistent and reliable. Notably, these predictive patterns were consistent across both intracranial and scalp EEG, supporting the robustness of the approach and its potential for non-invasive clinical applications.

This study highlights the value of EBRAINS' open datasets for enabling large-scale, reproducible analyses of brain stimulation data, and points toward a future in which real-time monitoring of brain state could make neuromodulation therapies more precise and predictable.