Not all networks behave the same. Sensorimotor regions showed stronger state-dependence than higher-order association areas, revealing a hierarchy—especially in SEEG analyses. 6/n

Conditioning stimulation on favorable pre-stimulus states reduced response variability by over 20%. This shows that brain-state monitoring can make interventions more reproducible, offering a concrete step toward refined closed-loop and precision neuromodulation. 5/n

Whole-brain context matters. When predictions are based only on activity close to the stimulation site, accuracy is limited. Expanding to include wider pre-stimulus dynamics across the whole brain consistently improves predictability of outcomes. 4/n

Across 125 candidate metrics, measures of synchronization, connectivity, and spatiotemporal signal diversity consistently predicted responses—sometimes explaining up to 80% of the variability within a session. 3/n

We analyzed a rare dataset: 36 patients, ~320 sessions, and more than >10,000 stimulations, combining intracranial and high-density EEG recordings. We asked: which pre-stimulus features are most reliable in predicting stimulation outcomes? 2/n