A recent trend in medical virtual reality is to include information from multiple sources, especially about physiology, into one model and one single visualization. Computer graphics must therefore deal with a huge amount of information in real time. The latest developments in computer graphics hardware allow not only implementing direct volume rendering on the graphics processing unit (GPU). The emerging compute languages enable us to address volume rendering problems of arbitrary complexity without being limited to formulating visualization techniques in an awkward fashion to match the GPU execution model. Utilizing the arising new possibilities we meet next generation’s demands in medical visualization.

We present an interactive rendering method for isosurfaces in a voxel grid. The underlying trivariate function is represented as a spline wavelet hierarchy, which allows for adaptive (view-dependent) selection of the desired level-of-detail by superimposing appropriately weighted basis functions. Different root finding techniques are compared with respect to their precision and efficiency. Both wavelet reconstruction and root finding are implemented in CUDA to utilize the high computational performance of Nvidia's hardware and to obtain high quality results. We tested our methods with datasets of up to 5123 voxels and demonstrate interactive frame rates for a viewport size of up to 1024 × 768 pixels.

We present a new GPU-based rendering system for ray casting of multiple volumes. Our approach supports a large number of volumes, complex translucent and concave polyhedral objects as well as CSG intersections of volumes and geometry in any combination. The system (including the rasterization stage) is implemented entirely in CUDA, which allows full control of the memory hierarchy, in particular access to high bandwidth and low latency shared memory. High depth complexity, which is problematic for conventional approaches based on depth peeling, can be handled successfully. As far as we know, our approach is the first framework for multivolume rendering which provides interactive frame rates when concurrently rendering more than 50 arbitrarily overlapping volumes on current graphics hardware.

We present a semiautomatic approach to generate high quality digital terrain models (DTM) from digital surface models (DSM). A DTM is a model of the earths surface, where all man made objects and the vegetation have been removed. In order to achieve this, we use a variational energy minimization approach. The proposed energy functional incorporates Huber regularization to yield piecewise smooth surfaces and an L1 norm in the data fidelity term. Additionally, a minimum constraint is used in order to prevent the ground level from pulling up, while buildings and vegetation are pulled down. Being convex, the proposed formulation allows us to compute the globally optimal solution. Clearly, a fully automatic approach does not yield the desired result in all situations. Therefore, we additionally allow the user to affect the algorithm using different user interaction tools. Furthermore, we provide a real-time 3D visualization of the output of the algorithm which additionally helps the user to assess the final DTM. We present results of the proposed approach using several real data sets.

To estimate the pose of a C-Arm during interventions therapy we have developed a small sized X-Ray Target including a special set of beads with known locations in 3D space. Since the patient needs to remain in the X-Ray path for all feasible poses of the C-Arm during the intervention, we cannot construct a single marker which is entirely visible in all images. Therefore finding 2D-3D point correspondences is a non-trivial task. The marker pattern has to be chosen in a way such that its projection onto the image plane is unique in a minimal-sized window for all relevant poses of the C-Arm. We use a two dimensional adaption of a linear feedback shift register (LFSR) to generate a two-dimensional pattern with unique sub-patterns in a certain window range. Thereby uniqueness is not achieved by placing unique 2D sub patterns side by side but by the code property itself. The code is designed in a way that any sub window of a minimal size guarantees uniqueness and that even occlusions from medical instruments can be handled. Experiments showed that we were able to estimate the C-Arm's pose from a single image within one second with a precision below one millimeter and one degree.

A common task in medical image analysis is the alignment of data from different sources, e.g., X-ray images and computed tomography (CT) data. Such a task is generally known as registration. We demonstrate the applicability of automatic differentiation (AD) techniques to a class of 2D/3D registration problems which are highly computationally intensive and can therefore greatly benefit from a parallel implementation on recent graphics processing units (GPUs). However, being designed for graphics applications, GPUs have some restrictions which conflict with requirements for reverse mode AD, in particular for taping and TBR analysis. We discuss design and implementation issues in the presence of such restrictions on the target platform and present a method which can register a CT volume data set (512 × 512 × 288 voxels) with three X-ray images (512 × 512 pixels each) in 20 seconds on a GeForce 8800GTX graphics card.

We report on a study carried out within the 3D MURALE project, in which the complete workflow necessary for automatic recording, reconstruction and interactive visualization of artefacts on archaeological excavation sites has been tested. With state-of-the-art technologies from digital photogrammetry, computer vision and computer graphics a prototypical pipeline for recording, 3D modeling and visualization of archaeological artefacts has been implemented and tried out on site on the excavation of Sagalassos (Turkey). The purpose of the presented work was to test the complete pipe-line in a real archaeologic environment and assess its usability.

The purpose of this study was to measure the ankle joint contact area under physiological load magnitudes using a stereophotography technique that allows accurate analysis of the entire joint surface without disrupting the joint during loading. Ten cadaveric foot and ankle specimens were loaded to 1000 N in neutral, and 20 degrees dorsiflexion, supination, pronation, and plantarflexion. Photo targets rigidly fixed to each of the bones were imaged in the loaded joint position using a high-resolution stereophotography system. After testing, each ankle was disarticulated and the joint surfaces imaged relative to the photo targets. The photo targets were then used to spatially register the joint surfaces into the loaded joint position; the overlap of the surfaces was used to determine the joint contact area. The mean talo-tibia contact area was greatest in dorsiflexion 7.34 +/- 1.69 cm(2) and was significantly larger than in plantar flexion (p < 0.05), which showed the smallest joint contact area 4.39 +/- 1.41 cm(2). Considering talo-fibula, the maximum contact area was measured in dorsiflexion, 2.02 +/- 0.78 cm(2), and the minimum contact area occurred in pronation, 0.77 +/- 0.49 cm(2), respectively (p < 0.05). The reported stereophotography technique allows measurement of the joint contact area without disrupting the joint during loading. The contact area is larger than previously reported, as the entire joint surface was analyzed. Joint contact extends over both the talar dome and the talar shoulders where osteochondritis dissecans lesions commonly occur.

We present a system to measure geometric properties (e.g., volume, surface area, and contact area) of the cartilage layer in human ankle joints. The cartilage and subchondral bone surfaces are sampled with a stereophotogrammetric device. A configurable pipeline of processing steps is applied to the sampled surface data to compute the desired quantities. The meshes are cleaned (i.e., disconnected parts removed, holes filled, and noise reduced) and aligned to each other such that they represent the cartilage layer as close as possibly. A hierarchical stitching approach creates from two separate (opposing) meshes a single closed surface, which is used for volume calculations. The method has been evaluated with data from 12 lower leg specimens. While our approach confirms the results of previous contact area studies for low curvature regions of the inspected cartilage and bone surfaces, it is superior to existing (mostly mechanical) methods for highly curved regions.

This paper combines two powerful approaches: variational methods and graphics hardware. Variational methods have demonstrated considerable success in com- puter vision for such diverse tasks as denoising, segmenta- tion, registration, stereo matching etc. Their main advan- tage is a mathematically clean and powerful formulation of the vision problem in terms of energy functionals that have to be minimized. However, due to their iterative nature thes e approaches tend to be slow and far from real-time capable. Recent progress in graphics hardware (the computational power grows much faster than for standard CPUs) makes this hardware interesting for computer vision application s. For example floating point arithmetic and high-level pro- gramming languages such as Cg are now available for mod- ern graphics cards. In this paper we demonstrate that by a careful analysis and formulation of variational methods and exploitation of the parallelism of modern GPUs we can achieve real-time performance without complex multi-grid optimization schemes (we obtain speed-ups of a factor of more than200 compared to Matlab implementations). This opens several new application areas for variational method s (e.g. realtime algorithms).

Aim To characterise the mechanical properties of the ankle, it is essential to have accurate joint morphology and measurements of the cartilage thickness and its variation across the joint. Thickness and volume measurements are also useful tools for detecting and monitoring degenerative change, however baseline measurements are required, to act as a ‘gold standard’. We present details of ankle cartilage thickness and distribution over the entire ankle joint, using a high precision stereophotogrammetry system. Method Twelve cadaveric ankles surfaces with photo targets, rigidly attached, were imaged using a stereo-photographic system, which generates a dense 3D point cloud of co-ordinates on the surface (typically 70,000 points per surface, accuracy ±2 μm). After imaging the surface, the cartilage was dissolved using 5% sodium hypochlorite to reveal the subchondral bone and the process was repeated. The two surfaces were combined and the normal distance from cartilage surface to bone was calculated at every point on the cartilage surface. Results The mean cumulative cartilage thickness of the ankle joint was 1.18±0.23 mm, the mean maximum cumulative cartilage thickness of the entire ankle joint was 2.17±0.46 mm. When considering the cartilage layers of the talus and the tibia-fibula complex separately, the mean and mean maximum thickness for the talus was 1.17±0.18 mm and 2.12±0.54 mm respectively. For the tibia-fibula complex, the mean and mean maximum thickness was 1.18±0.28 mm and 2.3±0.57 mm respectively. 3D cartilage thickness maps were also produced Conclusion The cartilage maps show that the thickest cartilage occurs at the shoulders of the talus, as opposed to the talar dome, as reported in earlier studies, which were unable to assess the highly curved regions of the ankle. This method also provides a gold standard for validating MRI cartilage measurements.

The scientific visualization and computer graphics communities have witnessed a tremendous rise in graphics processing unit (GPU) related literature and methodology recently. This is due in part to the rapidly increasing processing speed offered by graphics cards. Parallel to this, we have seen several advances made in the area of texture-based flow visualization. We present a texture-based flow visualization technique, Image Space Advection (ISA), that takes advantage of the computing power offered by recent, state-of-the-art GPUs. We have implemented a completely GPU-based version of the ISA algorithm. Here we describe our implementation in detail, including both the advantages and disadvantages of implementing ISA on the GPU. The result is state-of-the-art technique that demonstrates the latest in terms of both flow visualization methodology and GPU programming.

Future emission regulations and increased customer demands require the utilization of all available tools in developing engines of the next generation. 3D CFD simulation is one of these tools [1]. Due to well-advanced and welldeveloped simulation technology and improved computing power, the generated pack of data becomes complex and difficult to evaluate. In principle, technical problems can be treated with simulation, but the analysis of data very often leads to unsatisfactory results due to the limited evaluation functionality of the tools available at present. On the one hand, the reason is to be found in the visualization of 3D data, which is nowadays almost always performed in 2D only, and on the other hand in the physical relation of parameters, which is usually not reproduced in the analysis. Recently developed visualization techniques and modern graphics hardware make it possible to interact with vast amounts of multidimensional data in an intuitive way. A novel approach called SimVis [2] allows the engineer to specify certain subsets of the data – so-called features – in terms of data attributes under consideration (e.g., pressure, temperature, etc., and combinations thereof) and to immediately get a 3D view of the CFD data set with the selected features highlighted. This is useful to understand statistical properties of the data and the corresponding spatial and temporal distribution. Geometric properties can be presented and explored by means of virtual reality (VR) methods [3], which allow the user to become immersed in the dataset and to interact with the model in a natural way. Progress in PC graphics hardware technology makes it possible to perform sophisticated tasks in real-time in such an environment. By a simulation of a scavenge and combustion process in a 2-stroke engine, this paper investigates new solutions to the evaluation and visualization problem using SimVis and VR. Moreover, an outlook on further new developments is given.

Recently developed methods address the problem of denoising polygonal meshes derived from automatic scanning procedures. Nonlinear filtering is applied iteratively based on the estimated statistical properties of errors associated with the measured vertex locations. This procedure requires a rather uniform sampling interval across the surface and cannot remove noise below a certain frequency due to the local support of the filter kernel. We consider an approach to remedy these problems by interleaving mesh denoising and mesh simplification based on the quadric error metric. Instead of directly operating on vertex locations, the error quadrics (which are evaluated to determine vertex locations) are manipulated. Therefore denoising is applied at different scales according to the progress of simplification, being less sensitive to varying resolution and low-frequency noise. This is work in progress, which reveals some interesting properties in the context of denoising and simplification. However, the desired results have not yet been achieved.

We present a simple and efficient representation to store and transmit connectivity data of view dependent meshes for out-of-core rendering of large datasets. Resident mesh data available for rendering is organized as vertex tree to support real-time visualization of huge models. Our approach is not restricted to manifold meshes and can be used in the presence of non topology-preserving refinement operations as well. In contrast to progressive mesh compression our representation allows selective access to relevant fractions of the refinement hierarchy. After new mesh refinements are received, the update of the resident hierarchy available for display does not rely on known topology in the neighborhood of the patches to be refined. Therefore our approach does not require dependencies. We compare our connectivity coding with an existing framework for efficient transmission of view-dependent meshes and obtained substantially better compression results.