IEEE Transactions on Pattern Analysis and Machine Intelligence
Discriminative sliding preserving regularization in medical image registration
ISBI'09 Proceedings of the Sixth IEEE international conference on Symposium on Biomedical Imaging: From Nano to Macro
From box filtering to fast explicit diffusion
Proceedings of the 32nd DAGM conference on Pattern recognition
Efficient and reliable schemes for nonlinear diffusion filtering
IEEE Transactions on Image Processing
Hi-index | 0.00 |
The accurate estimation of respiratory lung motion by non-linear registration is currently an important topic of research and required for many applications in pulmonary image analysis, e.g. for radiotherapy treatment planning. A special challenge for lung registration is the sliding motion between visceral an parietal pleurae during breathing, which causes discontinuities in the motion field. It has been shown that accounting for this physiological aspect by modeling the sliding motion using a direction-dependent regularization approach can significantly improve registration results. While the potential of such physiology-based regularization methods has been demonstrated in several publications, so far only simple explicit solution schemes were applied due to the computational complexity. In this paper, a numerical solution of the direction-dependent regularization based on Fast Explicit Diffusion (FED) is presented. The approach is tested for motion estimation on 23 thoracic CT images and a significant improvement over the classic explicit solution is shown.