Stabilizing fractional dynamical networks suppresses epileptic seizures

Why calming storms in the brain matters

For millions of people living with epilepsy, seizures can strike without warning, disrupting work, school, and everyday life. Many patients do not respond well to drugs, and surgery or implanted devices do not always bring relief. This study explores a new way to read and gently steer brain activity using mathematical tools that capture how signals unfold over time, with the goal of making seizures weaker and less likely to spread.

Looking at seizures as shifting brain states

The researchers focused on four stages that surround a seizure: quiet periods between events, the minutes just before a seizure, the seizure itself, and the recovery afterward. Using recordings from electrodes placed directly on the brain surface in 10 people with hard-to-treat epilepsy, they sliced the data into short time windows and treated each window as a snapshot of a changing brain network. Instead of assuming the brain responds only to its most recent activity, they used a mathematical description that allows the present to depend on a long history of past signals, better reflecting the brains memory like behavior.

Finding hidden patterns in brain rhythms

With this approach, the team extracted two key features from the recordings. One feature described how strongly current brain activity depends on its past, capturing multi scale, or long range, memory in the signals. The other feature described how stable or unstable the network is at a given moment. Across patients, quiet periods between seizures showed one characteristic pattern, early warning periods before seizures showed another, and the seizure and recovery stages each had their own signatures. In particular, as the brain moved from quiet to seizure, its activity became more history dependent, suggesting that once a seizure pattern emerges, it can feed on its own past and become self sustaining.

How a gentle steering signal can tame seizures

Armed with these patterns, the researchers designed a control strategy that computes small, targeted adjustments to the brain network, similar to a thermostat nudging a room back to a comfortable temperature. Using the recorded data, they simulated what would happen if such stabilizing signals were applied at seizure onset. In 27 of 35 recorded seizures, the adjusted networks became mathematically stable, and across all seizures the simulated brain signals dropped in strength by about half on average. Only a handful of seizures could not be stabilized, which the authors traced to numerical issues that may reflect especially complex brain changes in those cases.

Personal differences and what they reveal

When the team compared results across all patients together, the four brain states looked different but overlapped. When they examined each person separately, the differences between states became much clearer. This suggests that seizure dynamics are highly personal, shaped by each individuals brain structure and history of disease. In many patients, the transition from the quiet state to the pre seizure state was easier to detect than the moment the seizure visibly began, hinting that early warning may come from subtle shifts long before outward signs appear.

What this could mean for future care

In simple terms, the study shows that seizures can be viewed as storms in a networked brain that carry a memory of what came before, and that carefully designed control signals might help calm those storms. While these results come from computer simulations based on real patient data, they point toward future implanted or noninvasive devices that could sense a persons unique seizure patterns and deliver personalized, low strength stimulation to keep brain activity within a healthier range.

Citation: Wang, Y., Ashourvan, A., Ramos, G. et al. Stabilizing fractional dynamical networks suppresses epileptic seizures. Sci Rep 16, 16037 (2026). https://doi.org/10.1038/s41598-026-43151-1

Keywords: epilepsy, seizures, brain networks, neurostimulation, EEG