Three-dimensional lung medical image registration based on improved demons algorithm

To put forward an accurate and effective registration method based on improved Demons algorithm, and to register three-dimensional pulmonary medical images of the same individual whose lung has the deformation under different respiration state. The experimental data are the maximum and minimum respiratory phase of the three-dimensional pulmonary images within a cycle of respiratory movement. First, images for registration are registered globally and non-rigidly. Feature points are extracted and matched by scale invariant feature transform algorithm. Afterwards, the global registration is finished according to the transformation parameter computed based on matching results. Finally, images after global registration are registered non-rigidly utilizing improved Demons algorithm. Image registration of human lung is realized employing improved method. The mean-square error between images before registration is 25,835.3 and it is reduced to 11,790.9 after registration. After further deal with improved Demons algorithm, the mean-square error between images is reduced to 3726.31 and the descent rate of mean-square is up to 85.58%. The proposed method effectively registers three-dimensional pulmonary images, which provides a solid foundation for doctors to estimate pulmonary respiratory movement and analyze respiratory function.