Related Experiment Video
Updated: Sep 12, 2026
Time-dependent Increase in the Network Response to the Stimulation of Neuronal Cell Cultures on Micro-electrode Arrays
Published on: May 29, 2017
Dynamical network models predict spectral response to single-pulse electrical stimulation
Helen Brinyark1, Caila Coyne2, Nicholas Reyes3
1Biomedical Engineering, The University of Alabama at Birmingham School of Engineering, 0511 GH, Birmingham, Birmingham, Alabama, 35294, United States.
Objective:
Drug-resistant epilepsy can be treated by surgically removing the seizure onset zone (SOZ), but localizing this zone is challenging because no clinically accepted biomarker of the SOZ exists. A novel computational biomarker, neural resonance, utilizes control theory to model the brain's response to stimulation. However, model validation is limited, and it is necessary to further contextualize the relationship between the model's prediction and the brain's true response to stimulation.
Approach:
Single-pulse electrical stimulation (SPES) provides an opportunity to bridge this gap as Bode plots derived from dynamical network models built from SPES data can be directly compared to the spectral response of the SPES-evoked waveform, the cortico-cortical spectral response (CCSR). In this study, we tested whether neural resonance, predicted by dynamical network models, reflects the SPES-evoked CCSR. Using SPES data from 33 patients, we constructed dynamical network models and compared the model-predicted magnitude and frequency behavior (Bode plot) to the CCSR curve measured for each patient.
Main Results:
Across patients, Bode plots correlated best with the CCSR curves taken at the time of the N2 component of the evoked response and taken as the average post-stimulus CCSR response (median Pearson correlation coefficient, r= 0.45 and 0.60, respectively), and significantly better than CCSR curves from randomly selected brain regions (Wilcoxon signed-rank test, p < 〖1x10〗^(-5)). We also found that resonant frequencies in the Bode plot were significantly more co-expressed in CCSR curves coinciding with the time of the N1 and N2 components of the evoked response, features associated with early direct connectivity and delayed indirect connectivity, respectively (Wilcoxon rank-sum test, both p < 〖1x10〗^(-3)).
Significance:
These findings validate the use of dynamical network models to predict neural connectivity and further support the use of neural resonance as a biomarker helping to localize the SOZ through guiding stimulation to induce seizures.
