Affine transformation registers small scale lung deformation

Abstract

To evaluate the nature of small scale lung deformation between multiple pulmonary magnetic resonance images, two different kinematic intensity based image registration techniques: affine and bicubic Hermite interpolation were tested. The affine method estimates uniformly distributed deformation metrics throughout the lung. The bicubic Hermite method allows the expression of heterogeneously distributed deformation metrics such as Lagrangian strain. A cardiac triggered inversion recovery technique was used to obtain 10 sequential images of pulmonary vessel structure in a sagittal plane in the right lung at FRC in 4 healthy subjects (Age: 28.5(6.2)). One image was used as the reference image, and the remaining images (target images) were warped onto the reference image using both image registration techniques. The normalized correlation between the reference and the transformed target images within the lung domain was used as a cost function for optimization, and the root mean square (RMS) of image intensity difference was used to evaluate the quality of the registration. Both image registration techniques significantly improved the RMS compared with non-registered target images (p= 0.04). The spatial mean (µE) and standard deviation (σ(E)) of Lagrangian strain were computed based on the spatial distribution of lung deformation approximated by the bicubic Hermite method, and were measured on the order of 10(-3) or less, which is virtually negligible. As a result, small scale lung deformation between FRC lung volumes is spatially uniform, and can be simply characterized by affine deformation even though the bicubic Hermite method is capable of expressing complicated spatial patterns of lung deformation.