Regularized Nonrigid Registration of Lung CT Images by Preserving Tissue Volume and Vesselness Measure

Image registration is an important research area within pulmonary image analysis. Accurate registration is critical to post-analysis of lung mechanical properties and useful for clinical applications. To improve registration accuracy, we design a nonrigid registration algorithm to preserve both parenchymal tissue volume and vesselness measure. In addition, the transformation is regularized using a Laplacian constraint. Comparison experiments are performed and evaluation statistics demonstrate this algorithm has achieved better registration accuracy in the alignment of lung boundaries, fissures, and landmarks. Visual inspection shows obvious improvement on matching accuracy in the lung regions near the thoracic cage. This algorithm also results in physiologically more plausible Jacobian patterns.