Tomographic reconstruction is the process of reconstructing a 3-D object or its cross section from several of its 2-D projection images. The object is illuminated by a cone-beam of Xrays, where the signal is attenuated by the object. Due to its speed filtered back projection (FBP) still is state-of-the-art in 3-D reconstruction for clinical use where time matters. But considering the accuracy and number of projections required for FBP, as shown in [1], an algebraic reconstruction technique (ART) is superior. Our current focus lies on 3-D angiography using C-arm systems. But this new approach should also be applicable on many real world reconstruction problems. Within ART, the object is represented as a linear combination of basis functions, typically voxels, with some unknown coefficients. The observations can also be expressed as a linear combination of these coefficients. This results in a linear system of equations with a sparse system matrix, because each X-ray intensity observation is influenced only by the pixels on the corresponding beam path. If enough measures are available, one has an over-determined system, which is solved in the leastsquares sense. On the other hand, if there are not enough measures in a region to determine the coefficient values, one is faced with an under-determined problem. In this case, one solves the regularized version of the problem which supplies the additional constraints. Due to the large number of unknowns in real applications, an iterative instead of a direct linear solver has to be used. Techniques such as Kaczmarz’s algorithm or CAV (component averaging) are currently used as iterative solvers, but for large problems, their computational costs are high. In addition, these solvers tend to improve the solution very much only in the first few iterations. An efficient ART is therefore essential to compete with FBP successfully. In this paper we think of these iterative methods as smoothers within a multigrid solver. It should be noted that because of the structure of the system matrix, the standard multigrid ∗University of Erlangen-Nuremberg, Germany, pruemmer@informatik.uni-erlangen.de †University of Erlangen-Nuremberg, Germany, Harald.Koestler@informatik.uni-erlangen.de ‡University of Erlangen-Nuremberg, Germany, Ulrich.Ruede@informatik.uni-erlangen.de §University of Erlangen-Nuremberg, Germany, joachim.hornegger@informatik.uni-erlangen.de

The detection of organs from full-body PET images is a challenging task due to the high noise and the limited amount of anatomical information of PET imaging. The knowledge of organ locations can support many clinical applications like image registration or tumor detection. This paper is the first to propose an organ localization framework tailored on the challenges of PET. The algorithm involves intensity normalization, feature extraction and regression forests. Linear and nonlinear intensity normalization methods are compared theoretically and experimentally. From the normalized images, long-range spatial context visual features are extracted. A regression forest predicts the organ bounding boxes. Experiments show that percentile normalization is the best preprocessing method. The algorithm is evaluated on 25 clinical images with a spatial resolution of 5mm. With 13.8mm mean absolute bounding box error, it achieves state-of-the-art results.

Time-resolved 3-D imaging of the heart is a major research topic in the medical imaging community. Recent advances in the interventional cardiac 3-D imaging from rotational angiography (C-arm CT) are now also making 4-D imaging feasible during procedures in the catheter laboratory. State-of-the-art reconstruction algorithms try to estimate the cardiac motion and utilize the motion field to enhance the reconstruction of a stable cardiac phase (diastole). The available data offers a handful of opportunities during interventional procedures, e. g. the ECG-synchronized dynamic roadmapping or the computation and analysis of functional parameters. In this paper we will demonstrate that the motion vector field (MVF) that is output by motion compensated image reconstruction algorithms is in general not directly usable for animation and motion analysis. Dependent on the algorithm different defects are investigated. A primary issue is that the MVF needs to be inverted, i.e. the wrong direction of motion is provided. A second major issue is the non-periodicity of cardiac motion. In algorithms which compute a non-periodic motion field from a single rotation the in depth motion information along viewing direction is missing, since this cannot be measured in the projections. As a result, while the MVF improves reconstruction quality, it is insufficient for motion animation and analysis. We propose an algorithm to solve both problems, i.e. inversion and missing in-depth information in a unified framework. A periodic version of the MVF is approximated. The task is formulated as a linear optimization problem where a parametric smooth motion model based on B-splines is estimated from the MVF. It is shown that the problem can be solved using a sparse QR factorization within a clinical feasible time of less than one minute. In a phantom experiment using the publicly available CAVAREV platform, the average quality of a non-periodic animation could be increased by 39% by applying the proposed periodization and inversion method.

Zur Bewältigung komplexer Berechnungen wird in der medizinischen Bildverarbeitung immer häufiger Spezialhardware eingesetzt. Die Open Computing Language offeriert die Möglichkeit eines gleichzeitig hardware-unabhängigen und performanten Programms. Dies wurde von uns am Beispiel der Bildrekonstruktion untersucht und gezeigt, dass sich mit Hilfe von OpenCL auf CPU-Systemen Leistungssteigerungen einfach erzielen lassen. Des weiteren wird ein hohe Unabhängigkeit der Implementierung von der Hardware erreicht und somit die Nutzung moderner Technologien, wie z.B. Grafikprozessoren, erleichtert. Die Laufzeit unseres Problems konnten wir auf einer Vierkern-CPU von 40min auf 6, 5min reduzieren. Durch die Verwendung einer Grafikkarte und einfache Optimierungen wurde schließlich eine Laufzeit von 17 s erreicht.

High-density objects, like catheters, pacemakers or even contrast agent-filled vessels, cause characteristic streak artifacts in computed tomography (CT). Similar to metal artifacts, these streaks can be reduced by removing the dense object using segmentation and interpolation. First, we compare state-of-the-art interpolation methods like linear, spline and higher-order methods to the Healing Brush technique. Second, a new method is presented, that extracts a low-frequency model of the dense object and restores the decomposed X-ray intensity of the remaining tissue. This method is henceforth called Subtract-and-Shift. Compared to standard interpolation methods, it retains the measured structure that is superimposed and dominated by the dense object. The extracted structure is then used to replace the segmented pixel intensities of the object. The introduced method is compared to state-of-the-art interpolation methods using in-vivo data. First preliminary results show that Subtract-and-Shift can be superior to these interpolation methods.

Heart motion is a crucial problem in cardiac tomographic cone-beam image reconstruction. It requires special treatment to avoid motion related image artifacts. Analytic and iterative algorithms for approximative and exact motion compensated 3-D reconstruction are known. The estimation of the motion field from the projection data is still an open problem. The inherent assumption of recent publications is a periodic heart motion. The electrocardiogram (ECG) is used as an estimate for the periodically repeating heart phase. In those approaches the heart motion is averaged over several heart cycles. As a consequence heart beat variabilities cannot be captured. However, frequently arrhytmic heart cycles can be observed in a clinical environment. In addition breathing motion can still occur. We present a reconstruction method based on a 4-D timecontinuous B-spline motion field which is parameterized by the acquisition time and not the quasi-periodic heart phase. A timecorrelated objective function is introduced which measures the error between the measured projection data and the dynamic forward projection of the motion compensated reconstruction. For reconstruction an analytic motion compensation algorithm is used. Our objective function formulation exploits the fact that the desired motion compensated reconstruction is totally determined by a given motion field. The motion model parameters are estimated using an iterative optimization scheme. Simulation results are provided for a synthetic cardiac vasculature phantom undergoing deformable motion which could be well recovered using the presented framework without using the ECG and assuming periodicity of the motion.

Anatomical and functional information of cardiac vasculature is a key component of future developments in the field of interventional cardiology. With the technology of C-arm CT it is possible to reconstruct intraprocedural 3-D images from angiographic projection data. Current approaches attempt to add the temporal dimension (4-D) by ECG-gating in order to distinct physical states of the heart. This model assumes that the heart motion is periodic. However, frequently arrhytmic heart signals are observed in a clinical environment. In addition breathing motion can still occur. We present a reconstruction method based on a 4-D time-continuous motion field which is parameterized by the acquisition time and not the quasi-periodic heart phase. The output of our method is twofold. It provides a motion compensated 3-D reconstruction (anatomic information) and a motion field (functional information). In a physical phantom experiment a vessel of size 3.08 mm undergoing a non-periodic motion was reconstructed. The resulting diameters were 3.42 mm and 1.85 mm assuming non-periodic and periodic motion, respectively. Further, for two clinical cases (coronary arteries and coronary sinus) it is demonstrated that the presented algorithm outperforms periodic approaches and is able to handle realistic irregular heart motion.

Generating 3-D images of the heart during interventional procedures is a significant challenge. In addition to real time fluoroscopy, angiographic C-arm systems can also now be used to generate 3-D/4-D CT images on the same system. One protocol for cardiac CT uses ECG triggered multisweep scans. A 3-D volume of the heart at a particular cardiac phase is then reconstructed by applying Feldkamp (FDK) reconstruction to the projection images with retrospective ECG gating. In this work we introduce a unified framework for heart motion estimation and dynamic cone-beam reconstruction using motion corrections. The benefits of motion correction are 1) increased temporal and spatial resolution by removing cardiac motion which may still exist in the ECG gated data sets and 2) increased signal-to-noise ratio (SNR) by using more projection data than is used in standard ECG gated methods. Three signal-enhanced reconstruction methods are introduced that make use of all of the acquired projection data to generate a 3-D reconstruction of the desired cardiac phase. The first averages all motion corrected back-projections; the second and third perform a weighted averaging according to 1) intensity variations and 2) temporal distance relative to a time resolved and motion corrected reference FDK reconstruction. In a comparison study seven methods are compared: nongated FDK, ECG-gated FDK, ECG-gated, and motion corrected FDK, the three signal-enhanced approaches, and temporally aligned and averaged ECG-gated FDK reconstructions. The quality measures used for comparison are spatial resolution and SNR. Evaluation is performed using phantom data and animal models. We show that data driven and subject-specific motion estimation combined with motion correction can decrease motion-related blurring substantially. Furthermore, SNR can be increased by up to 70% while maintaining spatial resolution at the same level as is provided by the ECG-gated FDK. The presented framework provides excellent image quality for cardiac C-arm CT.

For many interventional procedures the 3-D reconstruction of dynamic high contrast objects from C-arm data is desirable. We present a method for compensating artifacts from periodic motions by providing a modified filtered backprojection algorithm. The proposed algorithm comprises three steps: First, the reconstruction of an initial reference volume from a phase-consistent subset of the projection data. Secondly, the selection of proper data for a motion corrected reconstruction using as many projections as possible in the third step. The first step is addressed by gating in combination with a modified backprojection operator which reduces streak artifacts, the second by analysis of the cardiac motion characteristics and the impact on gated reconstruction quality and the third by accumulating gated sub-reconstructions registered with the reference volume. We present first clinical results from real patient data for the reconstruction of the coronary sinus.

Cardiac C-arm CT is a promising technique that enables 3D cardiac image acquisition and real-time fluoroscopy on the same system. Retrospective ECG gating techniques have already been adapted from clinical cardiac CT that allow 3D reconstruction using retrospectively gated projection images of a multi-sweep C-arm CT scan according to the desired cardiac phase. However, it is known that retrospective gating of projection data does not provide an optimal signal-to-noise-ratio (SNR) since the measured projection data is only partially considered during the reconstruction. In this work we introduce a new reconstruction technique for cardiac C-arm CT that provides increased SNR by including additional corrected and resampled filtered back-projections (FBP) from temporal windows outside of the targeted reconstruction phase. We take advantage of several motion corrected FDK-like reconstructions of the subject to increase SNR. In the presented results, using in vivo data from an animal model, the SNR could be increased by approximately 30 percent.

Cardiac C-arm CT is a promising technique that enables 3D cardiac image acquisition and real-time fluoroscopy on the same system. Retrospective ECG gating techniques have already been adapted from clinical cardiac CT that allow 3D reconstruction using retrospectively gated projection images of a multi-sweep C-arm CT scan according to the desired cardiac phase. However, it is known that retrospective gating of projection data does not provide an optimal signal-to-noise-ratio (SNR) since the measured projection data is only partially considered during the reconstruction. In this work we introduce a new reconstruction technique for cardiac C-arm CT that provides increased SNR by including additional corrected and resampled filtered back-projections (FBP) from temporal windows outside of the targeted reconstruction phase. We take advantage of several motion corrected FDK-like reconstructions of the subject to increase SNR. In the presented results, using in vivo data from an animal model, the SNR could be increased by approximately 30 percent.

In this paper we introduce a multigrid method for sparse, possibly rank-deficient and inconsistent least squares problems arising in the context of tomographic image reconstruction. The key idea is to construct a suitable AMG method using the Kaczmarz algorithm as smoother. We first present some theoretical results about the correction step and then show by our numerical experiments that we are able to reduce the computational time to achieve the same accuracy by using the multigrid method instead of the standard Kaczmarz algorithm.

We present a new level-set based method to segment and quantify stenosed internal carotid arteries (ICAs) in 3D contrast-enhanced computed tomography angiography (CTA). Within these data sets it is a difficult task to evaluate the degree of stenoses deterministically even for the experienced physician because the actual vessel lumen is hardly distinguishable from calcified plaque and there is no sharp border between lumen and arterial wall. According to our knowledge no commercially available software package allows the detection of the boundary between lumen and plaque components. Therefore in the clinical environment physicians have to perform the evaluation manually. This approach suffers from both intra- and inter-observer variability. The limitation of the manual approach requires the development of a semi-automatic method that is able to achieve deterministic segmentation results of the internal carotid artery via level-set techniques. With the new method different kinds of plaques were almost completely excluded from the segmented regions. For an objective evaluation we also studied the method’s performance with four different phantom data sets for which the ground truth of the degree of stenosis was known a priori. Finally, we applied the method to 10 ICAs and compared the obtained segmentations with manual measurements of three physicians.

Ziele: Entwicklung und klinische Evaluierung eines semiautomatischen Segmentierungsverfahrens zur Quantifizierung von Karotisstenosen in der CT-Angiographie. Methode: Der neu entwickelte Algorithmus wurde an CT-Angiographie Datensätzen von 10 Patienten mit hochgradigen Karotisstenosen getestet. Die Datenakquisition erfolgte an einem 16-MSCT Gerät. Es wurden Untersuchungen mit nicht-verkalkten Plaques, gemischten Plaques und stark verkalkten Plaques sowie zwei Untersuchungen mit deutlicher venöser Kontrastierung ausgewählt um die Praxistauglichkeit des Algorithmus zu überprüfen. Die errechneten Ergebnisse wurden mit den manuell gemessenen Werten verglichen. 3 Untersucher werteten die CTA-Daten unabhängig voneinander manuell aus. Die manuelle Auswertung wurde 2 mal pro Untersucher zu unterschiedlichen Zeitpunkten durchgeführt. Ergebnis: Es gelang mithilfe des Algorithmus eine zuverlässige Segmentierung der Karotis durchzuführen und die Plaqueanteile vom Gefäßlumen zu separieren. Wiederholte Messungen ergaben eine bessere Korrelation zwischen den semiautomatisch ermittelten Werten (r>0.9) als den manuell ermittelten Werten. Schlussfolgerung: Durch den Segmentierungsalgorithmus verbessert sich die Reproduzierbarkeit von Messergebnissen in der Quantifizierung von Karotisstenosen. Die Ergebnisse können durch überlagernde Venen beeinträchtigt werden, so dass eine visuelle Kontrolle der Segmentierung notwendig ist.

In diesem Beitrag beschreiben wir die Möglichkeiten der Steuerung von Geräten mittels natürlicher Sprache am Beispiel eines sprachgesteuerten 3D-Gefäßanalysesystems. Das System versteht ganze Sätze und erkennt selbständig, ob eine Äußerung an das System gerichtet ist oder an eine andere Person. Die Sprachsteuerung wurde am Lehrstuhl für Mustererkennung der Universität Erlangen-Nürnberg in Zusammenarbeit mit der Firma Sympalog Voice Solutions GmbH für ein Gerät zur Stenosenvermessung der Firma Siemens Medical Solutions (Leonardo Workstation) entwickelt und erfolgreich einer klinischen Erprobung unterzogen.