Active Control of Acoustic Noise by Magnetic Resonance Imaging: A Simulation Study

In medical imaging for clinical diagnosis and biomedical research magnetic resonance imaging (MRI) is one of the prevailing techniques. A lot of subjects of MRI are annoyed by the loud noise that MRI generates. In the present study computer simulation of active control of the noise induced by MRI was performed. The sound generated by the measurement sequences for the brain was recorded with an IC recorder in an MRI scanner room. Prediction of the time series of the sound was attempted using linear prediction and artificial neural networks (ANN). By subtraction of the predicted value from the sound value the sound intensity is reduced. With several optimization techniques of ANN such as steepest decent method, LevenbergMarquadt method, etc, active control of the noise was examined. In the present paper, the comparison of the simulation results of the prediction with those methods is described. Key-Words: MRI, acoustic noise, active noise control, linear prediction, artificial neural networks, steepest decent method, Levenberg-Marquadt method