Jonathan Lessick

Segmentation of the heart and great vessels in CT images using a model-based adaptation framework

Recently, model-based methods for the automatic segmentation of the heart chambers have been proposed. An important application of these methods is the characterization of the heart function. Heart models are, however, increasingly used for interventional guidance making it necessary to also extract the attached great vessels. It is, for instance, important to extract the left atrium and the proximal part of the pulmonary veins to support guidance of ablation procedures for atrial fibrillation treatment. For cardiac resynchronization therapy, a heart model including the coronary sinus is needed. We present a heart model comprising the four heart chambers and the attached great vessels. By assigning individual linear transformations to the heart chambers and to short tubular segments building the great vessels, variable sizes of the heart chambers and bending of the vessels can be described in a consistent way. A configurable algorithmic framework that we call adaptation engine matches the heart model automatically to cardiac CT angiography images in a multi-stage process. First, the heart is detected using a Generalized Hough Transformation. Subsequently, the heart chambers are adapted. This stage uses parametric as well as deformable mesh adaptation techniques. In the final stage, segments of the large vascular structures are successively activated and adapted. To optimize the computational performance, the adaptation engine can vary the mesh resolution and freeze already adapted mesh parts. The data used for validation were independent from the data used for model-building. Ground truth segmentations were generated for 37 CT data sets reconstructed at several cardiac phases from 17 patients. Segmentation errors were assessed for anatomical sub-structures resulting in a mean surface-to-surface error ranging 0.50-0.82mm for the heart chambers and 0.60-1.32mm for the parts of the great vessels visible in the images.

Single-seeded coronary artery tracking in CT angiography

A new algorithm for rapid and accurate segmentation of the coronary arterial tree in CT angiography data sets is presented. Each artery is fully tracked from the aortic origin to its distal end, following a single touch by the user anywhere along the vessel. A two-level front propagation technique is applied for the tracking of each vessel. An initial bottom-hat filtering, an adaptive threshold and a heuristic credit system are incorporated to overcome the pitfalls characterizing the coronary environment like the adjacent veins, the proximate chambers and the changing gray level along the artery. The aortic origin and the distal end of each vessel are determined automatically based on specific stopping criteria. An additional seed placed at the aortic root is used to complete the extraction of the entire coronary tree for an overall examination. The algorithm used for the segmentation of the aortic root region is a planar front propagation based on watershed segmentation and an adaptive erosion technique that are needed to avoid leakage into the atria in cases of artifacts or poor image quality. The results of the proposed methodology were evaluated in 34 patients. Artery tracking was proved to be highly robust for images of medium and high quality.

Accurate segmentation of the left ventricle in computed tomography images for local wall thickness assessment

In recent years, the fully automatic segmentation of the whole heart from three-dimensional (3D) CT or MR images has become feasible with mean surface accuracies in the order of 1mm. The assessment of local myocardial motion and wall thickness for different heart phases requires highly consistent delineation of the involved surfaces. Papillary muscles and misleading pericardial structures lead to challenges that are not easily resolved. This paper presents a framework to train boundary detection functions to explicitly avoid unwanted structures. A two-pass deformable adaptation process allows to reduce false boundary detections in the first pass while detecting most wanted boundaries in a second pass refinement. Cross-validation tests were performed for 67 cardiac datasets from 33 patients. Mean surface accuracies for the left ventricular endo- and epicardium are 0.76mm and 0.68mm, respectively. The percentage of local outliers with segmentation errors > 2mm is reduced by a factor of 3 as compared to a previously published approach. Wall thickness measurements in full 3D demonstrate that artifacts due to irregular endo- and epicardial contours are drastically reduced.

Fast automatic delineation of cardiac volume of interest in MSCT images

Computed Tomography Angiography (CTA) is an emerging modality for assessing cardiac anatomy. The delineation of the cardiac volume of interest (VOI) is a pre-processing step for subsequent visualization or image processing. It serves the suppression of anatomic structures being not in the primary focus of the cardiac application, such as sternum, ribs, spinal column, descending aorta and pulmonary vasculature. These structures obliterate standard visualizations such as direct volume renderings or maximum intensity projections. In addition, outcome and performance of post-processing steps such as ventricle suppression, coronary artery segmentation or the detection of short and long axes of the heart can be improved. The structures being part of the cardiac VOI (coronary arteries and veins, myocardium, ventricles and atria) differ tremendously in appearance. In addition, there is no clear image feature associated with the contour (or better cut-surface) distinguishing between cardiac VOI and surrounding tissue making the automatic delineation of the cardiac VOI a difficult task. The presented approach locates in a first step chest wall and descending aorta in all image slices giving a rough estimate of the location of the heart. In a second step, a Fourier based active contour approach delineates slice-wise the border of the cardiac VOI. The algorithm has been evaluated on 41 multi-slice CT data-sets including cases with coronary stents and venous and arterial bypasses. The typical processing time amounts to 5-10s on a 1GHz P3 PC.

An automatic method for the identification and quantification of myocardial perfusion defects or infarction from cardiac CT images

The current study presents an automatic algorithm for detection of myocardial infarction and ischemia using cardiac CT image data. The classification is based on probabilistic tissue modeling, where a pixel is classified according to its maximum a-posteriori probability (MAP) as belonging to a normal or abnormal tissue segment. The pixels are represented in a two-dimensional space, where the first dimension is based on pixel intensity and the second relates to pixel position in the radial (transmural) direction. By means of this method, optimal thresholds for separating abnormal from normal pixels are calculated and clusters of abnormal pixels are identified. The methods performance was evaluated in comparison to an expert analysis of the cardiac CT images and showed good agreement.

Spatially Resolved Automatic Cardiac Rest Phase Determination in Coronary Computed Tomography Angiography (CTA)

We present a method to determine separately and automatically the individual cardiac rest phases of the right coronary artery (RCA), left anterior descending artery (LAD), and circumflex artery (LCX) for optimal image quality in retrospectively ECG-gated coronary CTA. We reconstruct a series of low resolution 3D images over the cardiac cycle to represent the different motion states of the heart. An average surface model of the heart is then automatically adapted to one of these images. The model contains labeled mesh regions corresponding to the typical locations of the three main coronary arteries. From the starting point, the model adaptation propagates over the image series for the entire cardiac cycle and follows the location of the arteries. Optimal phase points for diagnostic image reconstruction are determined as the points of minimal motion calculated from the 4D heart shape model on a per-artery basis. In an initial study of clinical cases, we determined cardiac rest phases individually for the RCA, LAD and LCX, and performed ECG-gated image reconstruction in the artery specific rest phase instead of a default rest phase of the whole heart. In cases where the rest phases differ between the three arteries, image quality is significantly improved using the presented approach. Motion artifacts are reduced and vessels appear less disrupted.

Streak artifact reduction in cardiac cone beam CT

Cone beam reconstructed cardiac CT images suffer from characteristic streak artifacts that affect the quality of coronary artery imaging. These artifacts arise from inhomogeneous distribution of noise. While in non-tagged reconstruction inhomogeneity of noise distribution is mainly due to anisotropy of the attenuation of the scanned object (e.g. shoulders), in cardiac imaging it is largely influenced by the non-uniform distribution of the acquired data used for reconstructing the heart at a given phase. We use a cardiac adaptive filter to reduce these streaks. In difference to previous methods of adaptive filtering that locally smooth data points on the basis of their attenuation values, our filter is applied as a function of the noise distribution of the data as it is used in the phase selective reconstruction. We have reconstructed trans-axial images without adaptive filtering, with a regular adaptive filter and with the cardiac adaptive filter. With the cardiac adaptive filter significant reduction of streaks is achieved, and thus image quality is improved. The coronary vessel is much more pronounced in the cardiac adaptive filtered images, in slab MIP the main coronary artery branches are more visible, and non-calcified plaque is better differentiated from vessel wall. This improvement is accomplished without altering significantly the border definition of calcified plaques.