978-1-6654-0810-3/21/$31.00 ©2021 IEEE System-Level Modeling of a Safe Autonomous Closed-loop Epileptic Seizure Control Implant Keyvan Farhang Razi, Mohammad Javad Karimi, Catherine Dehollain, Alexandre Schmid Department of Electrical and Micro Engineering Swiss Federal Institute of Technology Lausanne (EPFL) Lausanne, Switzerland Keyvan.farhangrazi@epfl.ch Abstract—A system-level model of a closed-loop epilepsy control implant is designed and simulated using the Simulink software for the first time in this work. The proposed model consists of multi-channel signal acquisition, ADC, seizure detector, multi-channel electrical stimulator, temperature sensor, wireless data and power transmission blocks. The power dissipation and temperature elevation of implants are crucial parameters which must be precisely controlled to ensure the safety of patients. The system is optimized to offer a low-power design. In addition, the controller of the system modifies the stimulation parameters to maintain the temperature elevation at the tissue within 1℃. Intracranial EEG (iEEG) datasets of five patients from the Inselspital Bern are used to evaluate the performance of the presented system. A two-stage seizure detection method used in the bio-signal processor enables achieving accurate detection with an energy efficient-approach. Keywords—Implantable medical device, Epileptic seizure control, Feature extraction, Electrical stimulation, Wireless power and data transmission I. INTRODUCTION Epilepsy is a neurological disorder that affects more than 65 million people worldwide. Approximately 30% of people living with epilepsy can be treated neither by anti- seizure medicine nor by the brain resection surgery [1]. Nowadays, implantable medical devices (IMD) are gaining significance as an emerging treatment of intractable epilepsy. An IMD with automatic seizure detection and responsive neural stimulation can significantly improve the life quality of epileptic patients [1]. Commercially available epilepsy control IMDs are categorized into open-loop and closed-loop devices. Closed-loop devices, apply electrical stimulation under the control of a digital signal processor. A significant amount of research has been done to improve the signal acquisition techniques compatible with implantable applications. [1] and [2] developed closed-loop epilepsy control implants focusing on enhancing the analog-front-end (AFE). [1] extracts multiple time-domain features while [5] uses both time and frequency domain features to detect seizures. Inductive links are frequently utilized to wirelessly supply power for all units of the implant as well as enabling data communication in various range of sizes from one- stage to free-floating implants [7]. Seizure suppression is realized by a proper electrical stimulation. The current-mode is the most common electrical stimulation technique due to its accurate charge transfer and miniaturized size [4]. A critical consideration of a safe long-term stimulation is residual charges that accumulate on the electrodes. [4] employs the anodic current pulse modulation active charge balancing to ensure maintaining the accumulated charge within an acceptable window range. In this work, a system-level design of a wirelessly powered epileptic seizure control implant is presented using Simulink software. The principal characteristics of the AFE, analog-to-digital converter (ADC), bio-signal processor, neural stimulator, wireless power and data transmission blocks are modeled. An energy-efficient seizure detection system is employed to activate the responsive electrical stimulator. Furthermore, the controller block is designed to continuously monitor the temperature elevation caused by the implant and apply necessary adjustments to the system’s parameters to guarantee safety of patients. The rest of the paper is organized as follows. Section II introduces the system overview and explains how to model all units of an autonomous implant. Section III describes the controller which is in charge of controlling the temperature elevation at tissue caused by power dissipation of the implant. The seizure detection results of the implant are given in Section IV. Finally, the main conclusions of the paper are summarized in Section V. II. SYSTEM OVERVIEW OF THE IMPLANT The epileptic seizure control implant consists of several components shown in Fig. 1 and Fig. 2. Fig. 1 includes the neural signal recording and stimulation channels, signal processing and controller blocks. Fig. 2 illustrates wireless power and data transmission as well as the temperature sensor block. The details pertaining to the system-level model of the implant units are given in the following. A. Signal acquisition recording channels The AFE comprises a Low-noise amplifier (LNA), low- pass filter, additional gain stage and sample-and-hold buffer. The LNA is the first and fundamental block of all neural recording systems. The role of this block is to amplify weak neural signals in the frequency band of interest and reject the large DC offset existing at the electrode-tissue interface. A transfer function and a gain block are utilized in order to model the performance of the LNA. Considering a typical LNA reported in [1], the passband frequency and the mid-band gain of the LNA are 30 Hz ~ 144 kHz and 30 dB, respectively. A 14 th -order Butterworth band-pass transfer function is employed to realize the transfer function of the LNA with the aforementioned properties. The power spectral density (PSD) of the input and output of the LNA is depicted in Fig. 3. It can be observed that the LNA successfully amplifies the weak mid-band neural signals while the large DC components are rejected thanks to its band-pass transfer function.