On a Spectral Image Metric for Non-Rigid Group-wise Registration of Dynamic MR Image Series

Through the recent developments of new MRI sequences, MRI became an alternative modality for functional pulmonary imaging. Ventilation and perfusion maps from such dynamic MR image series are of great diagnostic interest. The calculation of those maps requires a precise registration of the image series. However, the registration of dynamic MR image series of the lung is challenging, since the images are often blurred and affected by intensity changes and sliding organ motion. In this paper, we adapt a group-wise spectral image metric to apply it on dynamic 2d+t MR image series of the lung. The proposed image metric is then embedded in a parametric group-wise registration framework. For the evaluation we compared our method against three pairwise and two group-wise image metrics. All metrics are tested on nine dynamic MR image series of the lung. The results show that our metric extension performs statistically better than the state-of-the-art methods.