Seizure Control by a Learning Type Active Disturbance Rejection Approach

Epilepsy is one of the most common neurological disorders. Neuro-modulation becomes a promising way to address it. For an effective modulation, closed-loop mode is necessary but difficult. A control algorithm, which can adjust itself to get desired suppression of epileptic activity, is in great need. In this paper, active disturbance rejection control (ADRC) is utilized for its satisfied disturbance rejection and regulation performance. However, fixed observer parameters are difficult to fit the time-varying electrophysiological signals. Therefore, based on the estimation errors, an iterative learning approach is designed to get the parameters of an extended state observer (ESO). By combining the advantages of ADRC and the iterative learning, a learning type ADRC (LTADRC) is proposed to suppress the high amplitude epileptiform waves generated by the Jansen's neural mass model (NMM). For those variable parameters of an ESO, scalable bandwidths can be obtained to adapt to time-varying disturbance signals. It is of great significance for both ADRC and the neuro-modulation of epilepsy. Simulation results show that, compared with ADRC, much better performance can be obtained. It may provide a promising closed-loop regulation way for epilepsy in clinics.