· · · Repository hosted by TU Delft Library

Breathing motion complicates many image-guided interventions working on the thorax or upper abdomen. However, prior knowledge provided by a statistical breathing model, can reduce the uncertainties of organ location. In this paper, a prediction framework for statistical motion modeling is presented and different representations of the dynamic data for motion model building of the lungs are investigated. Evaluation carried out on 4D-CT data sets of 10 patients showed that a displacement vector-based representation can reduce most of the respiratory motion with a prediction error of about 2 mm, when assuming the diaphragm motion to be known.

A fully automatic iterative training approach to generate discriminative shape models for usage in the Generalized Hough Transform (GHT) is presented. The method aims at capturing the shape variability of the target object contained in the training data as well as identifying confusable structures (anti-shapes) and to integrate this information into one model. To distinguish shape and anti-shape points and to determine their importance, an individual positive or negative weight is estimated for each model point with means of a discriminative training technique. The model is built from edge points surrounding the target point and the most confusable structure as identified by the GHT. Through an iterative approach, the performance of the model is gradually improved by extending the training dataset with images, where the current model failed to localize the target point. The proposed method is successfully tested on a set of 670 long-leg radiographs, where it achieves a localization rate of 74-97%.

The Discriminative Generalized Hough Transform (DGHT) is a method for object localization, which combines the standard GHT with a discriminative training technique. Thereby the aim of the discriminativetraining is to equip the models used in the GHT with individual model point weights such that the localization error in the GHT becomesminimal. In this paper we want to introduce an extension of the DGHTusing a multi-level approach to reduce processing time and to improve localization accuracy. The approach searches for the target object on multiple resolution levels and combines this information for faster and better results. The advantage of the approach is demonstrated on low-resolution, whole-body MR images, which are intended for PET attenuation correction.

Temporal subtraction techniques using 2D image registration improve the detectability of interval changes from chest radiographs. Although such methods are well known for some time they are not widely used in radiologic practice. The reason are strong pose differences between these follow-up acquisitions with a time interval of months to years in between. Such strong perspective differences occur in a reasonable number of cases. They cannot be compensated by available image registration methods and thus mask interval changes to be undetectable. A method is proposed to estimate a 3D pose difference by the adaptation of a 3D rib cage model to both projections. The difference between both is then compensated for, thus producing a subtraction image with virtually no change in pose. No 3D image data is used. The accuracy of pose estimation is validated with chest phantom images under controlled geometric conditions.

The delineation of anatomical structures in medical images can be achieved in an efficient and robust manner using statistical anatomical organ models, which has been demonstrated for an already considerable set of organs, including the heart. While it is possible to provide models with sufficient shape variability to cope, to a large extent, with inter-patient variability, as long as object topology is conserved, it is a fundamental problem to cope with topological organvariability. We address this by creating a set of model variants and selecting the most appropriate model variant for the patient at hand. We propose a hybrid method combining model-based image analysiswith a guided region growing approach for automated anatomical variant selection and apply it to the left atrium in cardiac CT images. Concerning the human heart, the left atrium is the most variable sub-structure with a variable number of pulmonary veins drainng into it.It is of large clinical interest in the context of atrial fibrillation and related interventions.

Electrophysiological simulations of the atria could improve diagnosis and treatment of cardiac arrhythmia, like atrial fibrillation or flutter. For this purpose, a precise segmentation of both atria is needed. However, the atrial epicardium and the electrophysiological structures needed for electrophysiological simulations are barely or not at all detectable in CT-images. Therefore, a model based segmentation of only the atrial endocardium was developed as a landmark generator to facilitate the registration of a finite wall thickness model of the right and left atrial myocardium. It further incorporates atlas information about tissue structures relevant for simulation purposes like Bachmann’s bundle, terminal crest, sinus node and the pectinate muscles. The correct model based segmentation of the atrial endocardium was achieved with a mean vertex to surface error of 0.53 mm for the left and 0.18 mm for the right atrium respectively. The atlas based myocardium segmentation yields physiologically correct results well suited for electrophysiological simulations.