Peter Rösch

Respiratory motion in coronary magnetic resonance angiography: a comparison of different motion models.

To assess respiratory motion models for coronary magnetic resonance angiography (CMRA). In this study various motion models that describe the respiration‐induced 3D displacements and deformations of the main coronary arteries were compared.

Model evaluation and calibration for prospective respiratory motion correction in coronary MR angiography based on 3-D image registration

Image processing was used as a fundamental tool to derive motion information from magnetic resonance (MR) images, which was fed back into prospective respiratory motion correction during subsequent data acquisition to improve image quality in coronary MR angiography (CMRA) scans. This reduces motion artifacts in the images and, in addition, enables the usage of a broader gating window than commonly used today to increase the scan efficiency. The aim of the study reported in this paper was to find a suitable motion model to be used for respiratory motion correction in cardiac imaging and to develop a calibration procedure to adapt the motion model to the individual patient. At first, the performance of three motion models [one-dimensional translation in feet-head (FH) direction, three-dimensional (3-D) translation, and 3-D affine transformation] was tested in a small volunteer study. An elastic image registration algorithm was applied to 3-D MR images of the coronary vessels obtained at different respiratory levels. A strong intersubject variability was observed. The 3-D translation and affine transformation model were found to be superior over the conventional FH translation model used today. Furthermore, a new approach is presented, which utilizes a fast model-based image registration to extract motion information from time series of low-resolution 3-D MR images, which reflects the respiratory motion of the heart. The registration is based on a selectable global 3-D motion model (translation, rigid, or affine transformation). All 3-D MR images were registered with respect to end expiration. The resulting time series of model parameters were analyzed in combination with additionally acquired motion information from a diaphragmatic MR pencil-beam navigator to calibrate the respiratory motion model. To demonstrate the potential of a calibrated motion model for prospective motion correction in coronary imaging, the approach was tested in CMRA examinations in five volunteers.

Gray-Value Based Registration of CT and MR Images by Maximization of Local Correlation

For gray-value based multi-modality registration the similarity measure is essential. Excellent results have been obtained with mutual information for various modality combinations. In this contribution we consider local correlation as similarity measure for multi-modality registration. Using a software phantom it is analyzed why local correlation is suitable for this registration task whereas direct gray-value correlation itself is usually not. It is shown that registration with local correlation can be done using only a fraction of the image volume offering an opportunity to accelerate the algorithm. Within validation, registration of the phantom images, two simultaneously acquired dual contrast MR images, and a clinical CT-MR data set has been studied. For comparison, the data sets have also been registered with mutual information. The results show that not only mutual information, but also local correlation is suitable for gray-value based multi-modality registration.

Computer-aided segmentation of pulmonary nodules: automated vasculature cutoff in thin- and thick-slice CT

Abstract Consistent volume measurement of pulmonary nodules is essential for follow-up monitoring for diagnosis and therapy purposes. Automated segmentation algorithms have to cope in a robust manner with nodules which are attached to surrounding vessels, the lung walls, or the diaphragm. In this paper we introduce such a nodule segmentation algorithm, and study the effect of slice spacing (thickness) and Hounsfield threshold on the estimated volume. With thin-slice data from multi-array CT scanners, the delicate vasculature structure surrounding most pulmonary nodules becomes visible. Even though the connectivity of the nodules to the surrounding vasculature varies with slice thickness and Hounsfield threshold, we find that the vasculature cutoff decisions of the proposed segmentation algorithm yield consistent measurement characteristics which are robust enough to compare nodule volumes between follow-up CT studies even of different slice thickness.

Robust 3D Deformation Field Estimation by Template Propagation

A new robust method to automatically determine a 3D motion vector field for medical images in the presence of large deformations is proposed. The central idea of this approach is template propagation. Starting from an image position where valid starting estimates are known, small sub-volumes (templates) are registered rigidly. Parameters of successfully registered templates serve as starting estimates for its neighbors. The registration proceeds layer by layer until the relevant image volume is covered. Based on this principle, a template-based registration algorithm has been implemented. Using the resulting set of corresponding points, the parameters of a non-rigid transformation scheme are determined. The complete procedure has been validated using four MR image pairs containing considerable deformations. In order to obtain an estimate for the accuracy, homologous points determined by template propagation are compared to corresponding landmarks defined by an expert. For landmarks with sufficient structure, the average deviation is well below the voxel size of the images. Because of the larger number of homologous points available, transformations incorporating the output of template propagation yielded a larger similarity between the reference image and the transformed image than an elastic transformation based on landmark pairs.