Application of artifical neural network methods in HTS RF coil design for MRI

Abstract

This article presents a new method of simulating high-temperature superconducting (HTS) RF coils using an electromagnetically trained artificial neural network (EM-ANN). This design is based on a spiral planar coil with distributed capacitance fabricated with Y1Ba2Cu3O7 (YBCO) films. Simulation time with this new method can be reduced to only one millionth of the time required by the commercial electromagnetic software programme HP Momentum. The new method can also exploit the properties of an artificial neural network by providing an inverse algorithm based on a resonant frequency input to derive other properties of an RF coil. This inverse algorithm using EM-ANN is easier, faster, and more interactive than the traditional “moment method.” The simulation results also show excellent agreement with experimental measurements, with a margin of error of less than 3%. © 2003 Wiley Periodicals, Inc. Concepts Magn Reson Part B (Magn Reson Engineering) 18B: 9–14, 2003