Lung Image Volume Registration Using the Affine Transformation and Demons Algorithm

Abstract

A method to register the expiration and inspiration breath-hold HRCT lung image volumes was presented. We considered that the deformation between the expiration and inspiration of lung can be decomposed into a global affine transformation and a local deformation. Before registration, we segmented the lung parenchyma from thoracic HRCT slices. Then, we used a prior anatomy knowledge based method to find the global affine transformation parameters, and re-sampled a side of lung to align the corresponding one. The Demons algorithm was employed to realize local deformation. In the experiment results, the images after registration exhibit a good alignment about the lung shape and the inner structures. The quantitative criterions indicate that the average volume overlap before registration is 0.7982, but after the global transformation, it becomes 0.8936 and 0.9544 after Demons. The great descending of RMSE shows that the average intensities of corresponding voxels become more and more similar. Our method about lung image volume registration can help the analysis of lung respiration function.