Dominik Fritz

Automatic segmentation of the left ventricle and computation of diagnostic parameters using regiongrowing and a statistical model

The manual segmentation and analysis of high-resolution multi-slice cardiac CT datasets is both labor intensive and time consuming. Therefore it is necessary to supply the cardiologist with powerful software tools to segment the myocardium and compute the relevant diagnostic parameters. In this work we present a semi-automatic cardiac segmentation approach with minimal user interaction. It is based on a combination of an adaptive slice-based regiongrowing and a modified Active Shape Model (ASM). Starting with a single manual click point in the ascending aorta, the aorta, the left atrium and the left ventricle get segmented with the slice-based adaptive regiongrowing. The approximate position of the aortic and mitral valve as well as the principal axes of the left ventricle (LV) are determined. To prevent the regiongrowing from draining into neighboring anatomical structures via CT artifacts, we implemented a draining control by examining a cubic region around the currently processed voxel. Additionally, we use moment-based parameters to integrate simple anatomical knowledge into the regiongrowing process. Using the results of the preceding regiongrowing process, a ventricle-centric and normalized coordinate system is established which is used to adapt a previously trained ASM to the image, using an iterative multi-resolution approach. After fitting the ASM to the image, we can use the generated model-points to create an exact surface model of the left ventricular myocardium for visualization and for computing the diagnostically relevant parameters, like the ventricular blood volume and the myocardial wall thickness.

Multi-Scale Feature Extraction for Learning-Based Classification of Coronary Artery Stenosis

Assessment of computed tomography coronary angiograms for diagnostic purposes is a mostly manual, timeconsuming task demanding a high degree of clinical experience. In order to support diagnosis, a method for reliable automatic detection of stenotic lesions in computed tomography angiograms is presented. Thereby, lesions are detected by boosting-based classification. Hence, a strong classifier is trained using the AdaBoost algorithm on annotated data. Subsequently, the resulting strong classification function is used in order to detect different types of coronary lesions in previously unseen data. As pattern recognition algorithms require a description of the objects to be classified, a novel approach for feature extraction in computed tomography angiograms is introduced. By generation of cylinder segments that approximate the vessel shape at multiple scales, feature values can be extracted that adequately describe the properties of stenotic lesions. As a result of the multi-scale approach, the algorithm is capable of dealing with the variability of stenotic lesion configuration. Evaluation of the algorithm was performed on a large database containing unseen segmented centerlines from cardiac computed tomography images. Results showed that the method was able to detect stenotic cardiovascular diseases with high sensitivity and specificity. Moreover, lesion based evaluation revealed that the majority of stenosis can be reliable identified in terms of position, type and extent.

Automatic Liver Segmentation Using the Random Walker Algorithm

In this paper we present a new method for fully automatic liver segmentation in computed tomography images. First, an initial set of seed points for the random walker algorithm is created. In this context, voxels belonging to air, fat tissue and ribcage are labeled as background. Furthermore, depending on the shape of the ribcage and voxel intensities, several seed points inside the liver are automatically selected as foreground. This seed mask is then used to initialize the segmentation algorithm. Our method was successfully tested on data of 22 patients.

MEDIASSIST: medical assistance for intraoperative skill transfer in minimally invasive surgery using augmented reality

Minimally invasive surgery is a highly complex medical discipline with various risks for surgeon and patient, but has also numerous advantages on patient-side. The surgeon has to adapt special operation-techniques and deal with difficulties like the complex hand-eye coordination, limited field of view and restricted mobility. To alleviate with these new problems, we propose to support the surgeons spatial cognition by using augmented reality (AR) techniques to directly visualize virtual objects in the surgical site. In order to generate an intelligent support, it is necessary to have an intraoperative assistance system that recognizes the surgical skills during the intervention and provides context-aware assistance surgeon using AR techniques. With MEDIASSIST we bundle our research activities in the field of intraoperative intelligent support and visualization. Our experimental setup consists of a stereo endoscope, an optical tracking system and a head-mounted-display for 3D visualization. The framework will be used as platform for the development and evaluation of our research in the field of skill recognition and context-aware assistance generation. This includes methods for surgical skill analysis, skill classification, context interpretation as well as assistive visualization and interaction techniques. In this paper we present the objectives of MEDIASSIST and first results in the fields of skill analysis, visualization and multi-modal interaction. In detail we present a markerless instrument tracking for surgical skill analysis as well as visualization techniques and recognition of interaction gestures in an AR environment.

Segmentation of the left and right cardiac ventricle using a combined bi-temporal statistical model

The manual segmentation and analysis of high-resolution multislice cardiac CT datasets is both labor intensive and time consuming. Therefore it is necessary to supply the cardiologist with powerful software tools to segment the myocardium as well as the cardiac cavities and to compute the relevant diagnostic parameters. In this paper we present an automatic cardiac segmentation procedure with minimal user interaction. It is based on a combined bi-temporal statistical model of the left and right ventricle using the principal component analysis (PCA) as well as the independent component analysis (ICA) to model global and local shape variation. To train the model we used manually drawn end-diastolic as well as end-systolic contours of the right epi- and of the left and right endocardium to create triangular surfaces of training datasets. These surfaces were used to build a mean triangular surface model of the left and right ventricle for the end-diastolic and end-systolic heart phase and to compute the PCA and ICA decorrelation matrices which are used in a point distribution model (PDM) to model the global and local shape variations. In contrast to many previous attempts of model based cardiac segmentation we do not create separate models for the left and the right ventricle and for different heart phases, but instead create one single parameter vector containing the information of both ventricles and both heart phases. This enables us to use the correlation between the phases and between left and right side to create a model which is more robust and less sensitive e.g. to poor contrast at the right ventricle.

Robust model-based centerline extraction of vessels in CTA data

Extracting the centerline of blood vessels is a frequently used technique to assist the physician in the diagnosis of common artery disease in CTA images. Thereby, a robust and precise computation of the centerline is an essential prerequisite. In this paper we present a novel approach to robustly model the vessel tree and to compute its centerline. The algorithm is initialized with two clicks from the physician, which mark the start and end point of the vessel to be examined. Our approach is divided into two consecutive steps. In the first step, a section of the vessel tree is mapped to the model so that the desired centerline is entirely included. After the generation of the model, the centerline can easily be extracted in the second step. The robust and efficient extraction of required model parameters is performed by a ray-casting approach. The proposed method determines a set of points on the vascular wall. The analysis of these points using the principal component analysis provides all parameters needed for modeling the vessel. The proposed technique reduces computation time and does not require a segmentation of the vessel lumen to determine the centerline of the vessel. Furthermore, a priori knowledge of vessel structures is incorporated to improve robustness in the presence of pathological deformations.

Automatic 4D-Segmentation of the Left Ventricle in Cardiac-CT-Data

The manual segmentation and analysis of 4D high resolution multi slice cardiac CT datasets is both labor intensive and time consuming. Therefore, it is necessary to supply the cardiologist with powerful software tools, to segment the myocardium and the cardiac cavities in all cardiac phases and to compute the relevant diagnostic parameters. In recent years there have been several publications concerning the segmentation and analysis of the left ventricle (LV) and myocardium for a single phase or for the diagnostically most relevant phases, the enddiastole (ED) and the endsystole (ES). However, for a complete diagnosis and especially of wall motion abnormalities, it is necessary to analyze not only the motion endpoints ED and ES, but also all phases in-between. In this paper a novel approach for the 4D segmentation of the left ventricle in cardiac-CT-data is presented. The segmentation of the 4D data is divided into a first part, which segments the motion endpoints of the cardiac cycle ED and ES and a second part, which segments all phases in-between. The first part is based on a bi-temporal statistical shape model of the left ventricle. The second part uses a novel approach based on the individual volume curve for the interpolation between ED and ES and afterwards an active contour algorithm for the final segmentation. The volume curve based interpolation step allows the constraint of the subsequent segmentation of the phases between ED and ES to very small search-intervals, hence makes the segmentation process faster and more robust.

Robuste Verzweigungserkennung von Gefäßen in CTA-Datensätzen zur modellbasierten Extraktion der Centerline

Bei der Befundung und Visualisierung von Blutgefasen ist deren Centerline von zentraler Bedeutung. Die Unterscheidung zwischen unverzweigten Abschnitten des Gefases und Verzweigungsbereichen ermoglicht den Einsatz spezialisierter und sehr effizienter Algorithmen zur modellbasierten Extraktion der Centerline. In diesem Artikel wird ein robustes Verfahren zur Verzweigungserkennung vorgestellt. Das Verfahren beruht auf einem Front-Propagation-Ansatz mit dynamisch angepassten Schwellwerten und einer anschliesenden Clusteranalyse. Die vorgestellte Methode zur Verzweigungserkennung wurde als Komponente einer Architektur zur Extraktion der Centerline auf handannotierten Datensatzen getestet. Erste Ergebnisse sind sehr vielversprechend und ermoglichen auch bei pathologischen Gefasen eine robuste Detektion von Gefasverzweigungen.

Fully automatic detection and visualization of patient specific coronary supply regions

Coronary territory maps, which associate myocardial regions with the corresponding coronary artery that supply them, are a common visualization technique to assist the physician in the diagnosis of coronary artery disease. However, the commonly used visualization is based on the AHA-17-segment model, which is an empirical population based model. Therefore, it does not necessarily cope with the often highly individual coronary anatomy of a specific patient. In this paper we introduce a novel fully automatic approach to compute the patient individual coronary supply regions in CTA datasets. This approach is divided in three consecutive steps. First, the aorta is fully automatically located in the dataset with a combination of a Hough transform and a cylindrical model matching approach. Having the location of the aorta, a segmentation and skeletonization of the coronary tree is triggered. In the next step, the three main branches (LAD, LCX and RCX) are automatically labeled, based on the knowledge of the pose of the aorta and the left ventricle. In the last step the labeled coronary tree is projected on the left ventricular surface, which can afterward be subdivided into the coronary supply regions, based on a Voronoi transform. The resulting supply regions can be either shown in 3D on the epicardiac surface of the left ventricle, or as a subdivision of a polarmap.

Technical experience from clinical studies with INPRES and a concept for a miniature augmented reality system

This paper is going to present a summary of our technical experience with the INPRES System -- an augmented reality system based upon a tracked see-through head-mounted display. With INPRES a complete augmented reality solution has been developed that has crucial advantages when compared with previous navigation systems. Using these techniques the surgeon does not need to turn his head from the patient to the computer monitor and vice versa. The systems purpose is to display virtual objects, e.g. cutting trajectories, tumours and risk-areas from computer-based surgical planning systems directly in the surgical site. The INPRES system was evaluated in several patient experiments in craniofacial surgery at the Department of Oral and Maxillofacial Surgery/University of Heidelberg. We will discuss the technical advantages as well as the limitations of INPRES and present two strategies as a result. On the one hand we will improve the existing and successful INPRES system with new hardware and a new calibration method to compensate for the stated disadvantage. On the other hand we will focus on miniaturized augmented reality systems and present a new concept based on fibre optics. This new system should be easily adaptable at surgical instruments and capable of projecting small structures. It consists of a source of light, a miniature TFT display, a fibre optic cable and a tool grip. Compared to established projection systems it has the capability of projecting into areas that are only accessible by a narrow path. No wide surgical exposure of the region is necessary for the use of augmented reality.