An adaptive finite element method to cope with a large scale lung deformation in magnetic resonance images

The purpose of this study is to present an adaptive deformable image registration method to improve the performance of a multi-resolution “demons” registration algorithm in handling large scale lung deformation observed in 4D-MR images. Specifically, a finite element method (FEM) was integrated with MR tagging information to correct registration errors in the lung region. The displacements of 349 tagged grids were calculated with an average of 3.5 cm. The mean error of the demons registration over the tags was 2.5 cm which was reduced to 0.7 cm by the FEM registration. The FEM-generated transformation was merged to the demons deformation map without introducing any discontinuity. This method can help correct deformable registration errors identified in the clinical setting.