Rapid Prediction of Ablation Zones of Irreversible Electroporation With Electrochemical Impedance Spectroscopy and Artificial Neural Network in a Heterogeneous Model

In this study, we combined electrochemical impedance spectroscopy (EIS) and artificial neural network (ANN) to predict the ablation zone of irreversible electroporation (IRE) in a heterogeneous plant model. The heterogeneous plant model was built by implanting an exotic material with a different electrical conductivity (copper or wood) into potato cubes. For each heterogeneous model, 55 IRE trials were performed with the pulse strength of 300–1300 V and the pulse number of 30, 60, or 90 ( $100 \mu \text{s}$ in the pulsewidth and the frequency of 1 Hz) for different positions of exotic implants. The ANN for each model was trained, tested, and validated by a total of 165 experimental data with five inputs (pulse strength, pulse number, implant $x$ -/ $y$ -axis values, and impedance variation parameter) and four outputs (ablation area, major axis length (MaAL), minor axis length (MiAL), and ablation boundary in the first quadrant). Both the experiment and simulation results showed that the two implants with different electrical conductivities could distort the electric field distribution in the plant model. This study concludes that the method combining ANN and EIS can be used to predict the ablation zone of heterogeneous IRE with acceptable accuracy ( $>$ 90%), which might bring a hint to the rapid monitoring of IRE in the treatment of tumors.