Compressible image registration for thoracic computed tomography images

We developed a method for the calculation of dynamic ventilation images from four dimensional computed tomography (4D CT) images. A voxel mapping produced by applying deformable image registration to the components of the 4D CT image data set is central to the calculation. Current algorithms, such as optical flow, assume incompressibility in their formulation which is inaccurate for lung tissue and a potential source of error. To improve the quality of the deformable image registration for lung tissue, we introduce the Compressible Combined Local Global (CCLG) method for the accurate registration of lung tissue CT images. The CCLG formulation accounts for: (1) the compressible nature of the lungs, (2) noise in the images, (3) the high computational workload required. In order to account for lung compressibility, voxel displacement is modeled using the conservation of mass equation rather than an assumption of constant voxel intensity as in optical flow. Secondly, the effects of noise are alleviated by applying the local-global approach to the conservation of mass setting. Finally, the resulting large scale linear systems are solved using a parallelizable, preconditioned conjugate gradient algorithm. The new method has been implemented and tested on two dimensional synthetic images and thoracic CT images with promising results. *To whom correspondance and reprint requests should be addressed: Professor Yin Zhang, Department of Computational and Applied Mathematics, Rice University, 6100 Main St. – MS 134, Houston, TX 77005-1892 E-mail: [email protected] Compressible image registration for thoracic computed tomography images 2