Volterra series modelling and compensation of non-linear distortions caused by susceptibility difference artefacts related to the presence of ferromagnetic implants in magnetic resonance imaging

Magnetic resonance imaging is popular in medical settings due to its unique technical characteristics. However, its full potential has been limited by imaging artefacts caused by various phenomena. Previously, a methodology w as proposed to characterize and reduce artefacts caused specifically by mag netic susceptibility differences. In the present work, the Volterra series approach is suggested as an alternative method for describing non-linear di stortions induced by susceptibility artefacts. A second-order Volterra seri es is utilized for characterizing the image non-linearities using a block-b y-block processing approach. Subsequently, a corresponding second-order inv erse Volterra series is applied to compensate for the quantified distortion s. In addition, a technique for automatic demarcation of recoverable and no n-recoverable regions in magnetic resonance images is proposed to ameliorat e the developed methodology. Experimental results indicate that this approa ch offers improved accuracy and flexibility in reducing nonlinear distortio ns caused by magnetic susceptibility artefacts. (C) 2001 IPEM. Published by Elsevier Science Ltd. All rights reserved.