Machine Learning-Based Rapid Analysis of the Failure Progress of Thin-film Electrodes from Electric Impedance Spectroscopy (EIS) Data

Abstract

Abstract Parylene C is a common substrate and encapsulation material used in implantable microelectrodes. Its reliability and failure are of great significance in the research and application of microelectrodes. In this study, three different failure stages of Parylene C thin-film electrodes were modeled using equivalent circuits, and the electric impedance spectroscopy of the electrodes were rapidly analyzed 9 different machine learning algorithms to identify the failure stages. The results showed that the three equivalent circuit models can represent the dynamics of the three failure stages of the Parylene C thin-film electrodes. The support vector machine (SVM) algorithm achieves more than 93% accuracy in identifying the equivalent circuit models from electric impedance spectroscopy data with an average time of 0.0273s. The SVM algorithm has great potential in fast analysis of electric impedance spectroscopy for the endurability study and application of implantable microelectrodes.