VCBM: Eurographics Workshop on Visual Computing for Biomedicine

Permanent URI for this community

https://diglib.eg.org/handle/10.2312/463

Browse

Now showing 1 - 20 of 287

Fully Automatic Skull-Stripping in 3D Time-of-Flight MRA Image Sequences
(The Eurographics Association, 2008) Forkert, Nils Daniel; Säring, Dennis; Fiehler, Jens; Illies, Till; Färber, Matthias; Möller, Dietmar; Handels, Heinz; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
In this paper we present a robust skull-stripping method for the isolation of cerebral tissue in 3D Time-of-Flight (TOF) magnetic resonance angiographic images of the brain. 3D TOF images are often acquired in case of cerebral vascular diseases, because of their good blood-to-background-contrast. Skull-stripping is an essential preprocessing step towards a better segmentation as well as direct visualization of the vascular system. Our approach consists of three main steps. After preprocessing in order to reduce signal inhomogeneities and noise the first main step is the segmentation of the surrounding skull using a region growing approach. The second step is the automatic extraction of distinctive points at the border of the brain, based on the segmentation of the skull, which are then used as supporting points for a graph based contour extraction. The third step is a slicewise correction based on a non-linear registration in order to improve sub-optimal segmentation results. The method proposed was validated using 18 manually stripped datasets. The calculated similarity measures show that the proposed method leads to good segmentation results with only a few segmentation errors. At the same time the mean rate of vessel voxels included by the brain segmentation is 99.18%. In summary the procedure suggested allows a fast and fully automatic segmentation of the brain and is especially helpful as a preprocessing step towards an automatic segmentation of the vessel system or direct volume rendering.
Illustrated Ultrasound for Multimodal Data Interpretation of Liver Examinations
(The Eurographics Association, 2008) Viola, Ivan; Nylund, Kim; Øye, Ola Kristoffer; Ulvang, Dag Magne; Gilja, Odd Helge; Hauser, Helwig; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Traditional visualization of real-time 2D ultrasound data is difficult to interpret, even for experienced medical personnel. To make the interpretation during the education phase easier, we enhance the visualization during liver examinations with an abstracted depiction of relevant anatomical structures, here denoted as illustrated ultrasound. The specifics of enhancing structures are available through an interactively co-registered computed tomography, which has been enhanced by semantic information. To assist the orientation in the liver, we partition the liver into Couinaud segments. They are defined in a rapid segmentation process based on linked 2D slice views and 3D exploded views. The semantics are interactively related from the co-registered modality to the real-time ultrasound via co-registration. During the illustrated ultrasound examination training we provide visual enhancements that depict which liver segments are intersected by the ultrasound slice.
Surface Curvature Line Clustering for Polyp Detection in CT Colonography
(The Eurographics Association, 2008) Zhao, Lingxiao; Ravesteijn, Vincent F. van; Botha, Charl P.; Truyen, Roel; Vos, Frans M.; Post, Frits H.; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Automatic polyp detection is a helpful addition to laborious visual inspection in CT colonography. Traditional detection methods are based on calculating image features at discrete positions on the colon wall. However large-scale surface shapes are not captured. This paper presents a novel approach to aggregate surface shape information for automatic polyp detection. The iso-surface of the colon wall can be partitioned into geometrically homogeneous regions based on clustering of curvature lines, using a spectral clustering algorithm and a symmetric line similarity measure. Each partition corresponds with the surface area that is covered by a single cluster. For each of the clusters, a number of features are calculated, based on the volumetric shape index and the surface curvedness, to select the surface partition corresponding to the cap of a polyp. We have applied our clustering approach to nine annotated patient datasets. Results show that the surface partition-based features are highly correlated with true polyp detections and can thus be used to reduce the number of false-positive detections.
GPU Accelerated Normalized Mutual Information and B-Spline Transformation
(The Eurographics Association, 2008) Teßmann, Matthias; Eisenacher, Christian; Enders, Frank; Stamminger, Marc; Hastreiter, Peter; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Visualization of multimodal images in medicine and other application areas requires correct and efficient registration. Optimally, the alignment operation is made an integral part of the rendering process. Voxel based approaches using mutual information ensure high quality similarity measurement. As a limiting factor, high computational load is caused since for every iteration of the optimization procedure one volume is transformed into the coordinate system of the other, a 2D histogram is generated and mutual information is computed. The expensive trilinear interpolation operations are well supported by 3D texture mapping hardware. However, existing strategies compute the histogram and mutual information on the CPU and thus require a cost intensive data transfer. Overcoming this considerable bottleneck, we introduce a new approach that efficiently supports all computations on modern graphics cards. This makes expensive data transfers from GPU to main memory dispensable. Due to its modularity, the presented approach can be integrated into general frameworks. As a major result, the speed improvement over existing GPU-CPU strategies amounts to a factor of 4 and over pure conventional CPU techniques to more than a factor of 10. Overall, the suggested strategy contributes considerably to integration of multimodal registration directly into interactive volume visualization.
An Integrated Platform for Dynamic Cardiac Simulation and Image Processing: Application to Personalised Tetralogy of Fallot Simulation
(The Eurographics Association, 2008) Toussaint, Nicolas; Mansi, T.; Delingette, H.; Ayache, N.; Sermesant, M.; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Processing and visualisation of dynamic data is still a common challenge in medical imaging, especially as for many applications there is an increasing amount of clinical data as well as generated data, such as in cardiac modelling. In this context, there is a strong need for software that can deal with dynamic data of different kinds (i.e. images, meshes, signals, etc.). In this paper we propose a platform that aims at helping researchers and clinicians to visualise and process such dynamic data, as well as evaluate simulation results. To illustrate this platform we chose to follow a concrete clinical application, the personalised simulation of the Tetralogy of Fallot. We show that the software provides the user with a significant help in the assessment and processing of the 3D+t raw data, as well as an adapted framework for visualisation and evaluation of various dynamic simulation results.
Generation of a Mean Motion Model of the Lung Using 4D-CT Image Data
(The Eurographics Association, 2008) Ehrhardt, Jan; Werner, René; Richberg, Alexander Schmidt -; Schulz, Benny; Handels, Heinz; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Modeling of respiratory motion gains in importance within the field of radiation therapy of lung cancer patients. Current modeling approaches are usually confined to intra-patient registration of 3D image data representing the individual patient's anatomy at different breathing phases. We propose an approach to generate a mean motion model of the lung based on thoracic 4D CT data of different patients to extend motion modeling capabilities. Our modeling process consists of two main parts: an intra - subject registration to generate subject - specific motion models and an inter - subject registration to combine these subject - specific motion models into a mean motion model. Further, we present methods to adapt the mean motion model to a patient-specific lung geometry. A first evaluation of the model was done by using the generated mean motion model to predict lung and tumor motion of individual patients and comparing the prediction quality to non - linear registration. Our results show that the average difference in prediction quality (measured by overlap coefficients) between non - linear registration and model - based prediction is approx. 10%. However, the patient - specific registration relies on individual 4D image data, whereas the model - based prediction was obtained without knowledge of the individual breathing dynamics. Results show that the model predicts motion patterns of individual patients generally well and we conclude from our results that such a model has the capability to provide valuable a-priori knowledge in many fields of applications.
Reconstruction of Blood Vessels from Neck CT Datasets using Stable 3D Mass-Spring Models
(The Eurographics Association, 2008) Dornheim, Jana; Lehmann, Dirk J.; Dornheim, Lars; Preim, Bernhard; Strauß, Gero; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Preoperative neck dissection planning benefits from a smooth, organic visualization of the main blood vessels of the neck, in particular the carotid artery and jugular vein. While most reconstruction techniques for vasculature are designed for segmenting the complete vessel tree, our goal is to isolate these specific blood vessels of the neck from the CT dataset, and to exclude irrelevant vasculature from the visualization. Pure threshold- and iso value-based reconstruction techniques do not allow such a selective segmentation and often lead to undersegmentation at the lower parts of the blood vessels, due to inhomogeneous contrast agent diffusion. In order to avoid staircase artifacts in the visualizations of the reconstructed vascular structures, a subvoxel accuracy of the reconstruction technique is also required. We present a model-based reconstruction technique to isolate blood vessels from neck CT datasets using Stable 3D Mass-Spring Models. The results can be visualized directly without staircase artifacts. The interaction needed for the reconstruction is reduced substantially to only a few clicks along the blood vessels. The presented method was evaluated with 30 blood vessels from 14 CT datasets of the neck and could be shown to be accurate, while leading to smooth visualizations of the neck blood vessels.
Automatic Hepatocyte Quantification from Histological Images: Comparing Pre-smoothing filters
(The Eurographics Association, 2008) Ivanovska, Tetyana; Schenk, Andrea; Dahmen, Uta; Hahn, Horst K.; Linsen, Lars; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Quantity of hepatocytes in the liver can reveal a lot of information for medical researchers. In our project, it is needed for evaluation of the liver regeneration rate. In this paper, we present a processing pipeline for automatic counting of hepatocytes from images of histological sections. In particular, we propose to introduce a preprocessing step in form of image smoothing. We apply five different smoothing techniques, namely Gaussian smoothing, nonlinear Gaussian smoothing, median filtering, anisotropic diffusion, and minimum description length segmentation, and compare them to each other. The processing pipeline is completed by subsequent automatic thresholding using Otsu s method and hepatocyte detection using Hough transform. We compare the quantification results in terms of quality (sensitivity and specificity rates) against the manually specified ground truth. We discuss the results and limitations of the individual processing steps as well as of the overall automatic quantification approach.
CT Late Enhancement Segmentation for the Combined Analysis of Coronary Arteries and Myocardial Viability
(The Eurographics Association, 2008) Hennemuth, Anja; Mahnken, Andreas; Kühnel, Caroline; Oeltze, Steffen; Peitgen, Heinz-Otto; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Non-invasive imaging techniques become more and more important as diagnostic tools for the assessment of coronary heart disease (CHD). While CT is widely applied for the inspection of the coronary arteries, the state of myocardial tissue is normally analyzed with MRI or nuclear imaging methods such as PET or SPECT. The effect of late enhancement, the accumulation of contrast agent in defective tissue, is used to assess tissue viability with MR imaging. Studies have shown, that this effect can be observed for iodine based contrast agents, which are used for CT coronary artery imaging, as well. Thus the goal of this work is the development of segmentation and visualization methods, which allow a combined inspection of the coronary arteries and the viability of the corresponding myocardial tissue. We therefore present a new segmentation method for the analysis of CT late enhancement images and the integration with methods for the inspection of the coronary arteries. In preliminary tests by a radiologist, the methods are applied to 4 datasets to compare the segmentation with the reference method, test the combined inspection for data with a known relation between infarction and supplying artery and test the general applicability to patient data. The preliminary results are promising, and further studies will focus on the evaluation of the segmentation method as well as on the clinical benefit through the combined analysis.
Focus + Context Rendering of Structured Biomedical Data
(The Eurographics Association, 2008) Abellán, Pascual; Puig, Anna; Tost, Dani; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Biomedical data can be classified according to different taxonomies. Understanding the relationships between different data categories is essential for an in-depth knowledge of the data. We present a volume rendering system aimed at outlining structural relationships between different classification criteria of a biomedical voxel model. The system clusterizes the model into subsets of voxels sharing the same classification criteria. It constructs a labelled voxel model storing for each voxel an identifier of its associated cluster. We represent the classification space as a graph and we render it in the application interface. This way, clinicians can specify their visualization queries by selecting nodes of the graph and boolean operations between them. Given a rendering query, the system computes a transfer function on the labelled voxel model domain. This transfer function, together with the original voxel model and the labelled voxel model, are used during rendering to visualize the selected data more or less colored according to level of the graph at which they have been selected, and contextualized with the other parts of the model to which they are related. We demonstrate the utility of our approach on several biomedical datasets.
Ontology-Based Visualization of Hierarchical Neuroanatomical Structures
(The Eurographics Association, 2008) Kuß, Anja; Prohaska, Steffen; Meyer, Björn; Rybak, Jürgen; Hege, Hans-Christian; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
This work presents a method for generating intuitive visualizations for high-level user queries to an hierarchical surface-based neuroanatomical atlas. We combine a spreading activation approach for computing focus and context in an ontology with a specific level-of-detail strategy for hierarchical structures. We demonstrate our method on an atlas of the bee brain.
FluoroSim: A Visual Problem-Solving Environment for Fluorescence Microscopy
(The Eurographics Association, 2008) Quammen, Cory W.; Richardson, Alvin C.; Haase, Julian; Harrison, Benjamin D.; II, Russell M. Taylor; Bloom, Kerry S.; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Fluorescence microscopy provides a powerful method for localization of structures in biological specimens. However, aspects of the image formation process such as noise and blur from the microscope s point-spread function combine to produce an unintuitive image transformation on the true structure of the fluorescing molecules in the specimen, hindering qualitative and quantitative analysis of even simple structures in unprocessed images. We introduce FluoroSim, an interactive fluorescence microscope simulator that can be used to train scientists who use fluorescence microscopy to understand the artifacts that arise from the image formation process, to determine the appropriateness of fluorescence microscopy as an imaging modality in an experiment, and to test and refine hypotheses of model specimens by comparing the output of the simulator to experimental data. FluoroSim renders synthetic fluorescence images from arbitrary geometric models represented as triangle meshes. We describe three rendering algorithms on graphics processing units for computing the convolution of the specimen model with a microscope s point-spread function and report on their performance. We also discuss several cases where the microscope simulator has been used to solve real problems in biology.
Information-based Transfer Functions for Multimodal Visualization
(The Eurographics Association, 2008) Haidacher, Martin; Bruckner, Stefan; Kanitsar, Armin; Gröller, M. Eduard; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Transfer functions are an essential part of volume visualization. In multimodal visualization at least two values exist at every sample point. Additionally, other parameters, such as gradient magnitude, are often retrieved for each sample point. To find a good transfer function for this high number of parameters is challenging because of the complexity of this task. In this paper we present a general information-based approach for transfer function design in multimodal visualization which is independent of the used modality types. Based on information theory, the complex multi-dimensional transfer function space is fused to allow utilization of a well-known 2D transfer function with a single value and gradient magnitude as parameters. Additionally, a quantity is introduced which enables better separation of regions with complementary information. The benefit of the new method in contrast to other techniques is a transfer function space which is easy to understand and which provides a better separation of different tissues. The usability of the new approach is shown on examples of different modalities.
Computation and Visualization of Asynchronous Behavior of the Heart
(The Eurographics Association, 2008) Wesarg, Stefan; Lacalli, Christina; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Nowadays, computer-aided diagnosis is widely used in the analysis of cardiac image data. Especially, for the investigation of the dynamic behavior of the heart, automated analysis tools for 4D data sets have been developed. A small set of descriptors of the heart s dynamics are established and used in the clinical routine. However, there exists a whole lot more of such parameters that can be extracted by analyzing 4D data sets. But, many of them are not used due to several reasons: time-consuming computation, no intuitive meaning, little clinical relevance, etc.. In this work we propose a novel descriptor for the dynamic behavior of the heart that can easily be computed from 4D data sets. It describes to which extent the heart exhibits an asynchronous movement. This novel descriptor ASYNCHRONISM is based on the already established measures WALL MOTION and WALL THICKENING, but reveals new, valuable information that is not available when relying only upon the two aforementioned parameters. The ASYNCHRONISM has an intuitive meaning, since it corresponds to the clinical classification scheme of wall motion abnormalities. Beyond its computation we present in this work also methods for its visualization as well as first preliminary results for 4D cardiac magnetic resonance image data.
A Haptic Rendering Algorithm for Molecular Interaction
(The Eurographics Association, 2008) Stocks, Matthew B.; Laycock, Stephen D.; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Haptic rendering is the process of calculating and displaying physical forces to a user. Used concomitantly with a virtual environment it can further enhance a user's immersive experience whilst they interact with computer graphics. Haptic Feedback has been applied to the study of molecular systems for several years, however, computation requirements have hampered progress. Most popular representations of molecules comprise of primitive shapes like spheres. Many molecules, especially proteins, potentially contain thousands of atoms each of which can be represented as a single sphere and will need to be processed for collision in the haptic rendering loop. Current systems often simulate stiff contacts with a proxy system based on tracking a point over planar surfaces. In this paper a novel method for the haptic rendering of a space filling molecule representation is presented. The technique reduces the time taken to detect and respond to the collisions and improves the overall spherical feel of the molecule by using the implicit description of spheres to track the surface as opposed to a polygonal approximation.
Robust Classification and Analysis of Anatomical Surfaces Using 3D Skeletons
(The Eurographics Association, 2008) Reniers, Dennie; Jalba, Andrei; Telea, Alexandru; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
We present a method for computing a surface classifier that can be used to detect convex ridges on voxel sur- faces extracted from 3D scans. In contrast to classical approaches based on (discrete) curvature computations, which can be sensitive to various types of noise, we propose here a new method that detects convex ridges on such surfaces, based on the computation of the surface s 3D skeleton. We use a suitable robust, noise-resistant skeletonization algorithm to extract the full 3D skeleton of the given surface, and subsequently compute a surface classifier that separates convex ridges from quasi-flat regions, using the feature points of the simplified skeleton. We demonstrate our method on voxel surfaces extracted from actual anatomical scans, with a focus on cortical surfaces, and compare our results with curvature-based classifiers. As a second application of the 3D skeleton, we show how a partitioning of the brain skeleton can be used in a preprocessing step for the brain surface analysis.
Computation of More Channels in Protein Molecules
(The Eurographics Association, 2008) Bene , Petr; Medek, Petr; Sochor, Jiri; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
In the process of designing drugs it is crucial to perform various analyses of cavities and channels in protein molecules. Chemists also require that more than one ideal channel be computed in a static protein molecule. Three basic approaches for computation of more than a single channel were introduced in recent publications. However, these approaches have several disadvantages. In this paper we propose a new adaptive method for computation of more channels. This new method is piloted on a real data and results are compared with channels identified by chemists as relevant. The comparison indicates that this method is a significant improvement over previous methods, as the method computes less number of similar and biochemically insignificant channels.
High-Quality Multimodal Volume Visualization of Intracerebral Pathological Tissue
(The Eurographics Association, 2008) Rieder, Christian; Schwier, Michael; Hahn, Horst K.; Peitgen, Heinz-Otto; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Parallel visualization of multiple MRI sequences in 2D is a standard method for exploration of pathological structures for neurosurgery planning. In this work our aim is to support visualization techniques that allow medical experts a fast and comprehensive combined exploration of anatomical structures with inhomogeneous pathological tissue in the three-dimensional volume rendering. The prototypical software solution presented in this paper addresses the issue that a high amount of interaction is commonly needed to merge different MRI sequences and that the resulting visualization does not allow to recognize anatomical details of the brain and pathological tissue at the same time without loss of information. We also present novel clipping methods for neurological volume exploration and emphasize important structures as well as suspicious high intensity signals from multiple sequences in the volume rendering. We demonstrate that our methods facilitate comprehensive volume visualization for neurosurgery.
Automatic Segmentation of the Pelvic Bones from CT Data Based on a Statistical Shape Model
(The Eurographics Association, 2008) Seim, Heiko; Kainmueller, Dagmar; Heller, Markus; Lamecker, Hans; Zachow, Stefan; Hege, Hans-Christian; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
We present an algorithm for automatic segmentation of the human pelvic bones from CT datasets that is based on the application of a statistical shape model. The proposed method is divided into three steps: 1) The averaged shape of the pelvis model is initially placed within the CT data using the Generalized Hough Transform, 2) the statistical shape model is then adapted to the image data by a transformation and variation of its shape modes, and 3) a final free-form deformation step based on optimal graph searching is applied to overcome the restrictive character of the statistical shape representation. We thoroughly evaluated the method on 50 manually segmented CT datasets by performing a leave-one-out study. The Generalized Hough Transform proved to be a reliable method for an automatic initial placement of the shape model within the CT data. Compared to the manual gold standard segmentations, our automatic segmentation approach produced an average surface distance of 1.2 ± 0.3mm after the adaptation of the statistical shape model, which could be reduced to 0.7±0.3mm using a final free-form deformation step. Together with an average segmentation time of less than 5 minutes, the results of our study indicate that our method meets the requirements of clinical routine.
Glyph-Based Visualization of Myocardial Perfusion Data and Enhancement with Contractility and Viability Information
(The Eurographics Association, 2008) Oeltze, Steffen; Hennemuth, Anja; Glaßer, Sylvia; Kühnel, Caroline; Preim, Bernhard; Charl Botha and Gordon Kindlmann and Wiro Niessen and Bernhard Preim
Perfusion data characterize the regional blood flow in human tissue. In the diagnosis of the Coronary Heart Disease, they are acquired to detect hypoperfused regions of the myocardium (heart muscle) at an early stage or to evaluate the hemodynamical relevance of a known pathologic vessel narrowing. For each voxel in the data, a time-intensity curve describes the enhancement of a contrast agent. Parameters derived from these curves characterize the regional perfusion and have to be integrated for diagnosis. The diagnostic evaluation of this multi-field data is challenging and time-consuming due to its complexity. We tackle this problem by developing a glyph-based integrated visualization of perfusion parameters in 3D-space with the patient-individual ventricular anatomy as context information. Besides the assessment of myocardial perfusion, current cardiac imaging technology allows for the investigation of myocardial contractility as well as for the detection of non-viable tissue. The combined inspection of these data supports diagnosis finding and therapy planning by allowing for the discrimination of healthy, hypoperfused and non-viable tissue as well as between non-viable and temporarily inactive tissue. To facilitate such an inspection, we apply registration methods that cope with differences in orientation and coverage between these three datasets. We enhance the glyph-based visualization of perfusion parameters by integrating parameters describing the myocardial contractility and viability.

Browse

Browsing VCBM: Eurographics Workshop on Visual Computing for Biomedicine by Issue Date

Results Per Page

Sort Options