Exploratory Study on the Dynamic Characteristics of Brain Electrical Impedance During Epileptic Seizures

Electrical impedance tomography (EIT), as an emerging imaging method, has the potential to become a portable, non-invasive and real-time monitoring technique for seizure detection. However, the existing studies lacked impedance characteristics that could prove the availability of EIT for detecting seizures. This study aimed to explore the characteristic parameters of impedance changes and evaluate their effectiveness in seizure recognition. Seizures kindled by kainic acid were monitored by EEG and EIT synchronously. EEG was used to determine ictal periods, and impedance changes during the same period were analyzed. Five impedance characteristic parameters, including linear and nonlinear indicators, were analyzed. The efficacy of these parameters in identifying seizures was verified through support vector machine. The impedance in the ROI decreased significantly during seizures. In most seizures, all characteristic parameters showed significant differences between seizures and normal states. The feature Mean performed best in single feature-based seizure recognition, with accuracy, sensitivity and specificity of 88.7%, 97.4%, and 77.5%, respectively. The combination of Mean, FI and SampEn improved the performance, achieving accuracy, sensitivity and specificity of 98.5%, 98.3% and 98.8%, respectively. The characteristic parameters proposed in this study may be used as typical features of seizures to identify the epileptic impedance, which is helpful to realize seizure detection based on EIT.