Search

Purpose Whole-body MRI is seeing increasing use in the study and diagnosis of disease progression. In this, a central task is the visual assessment of the progressive changes that occur between two whole-body MRI datasets, taken at baseline and follow-up. Current radiological workflow for this consists in manual search of each organ of interest on both scans, usually on multiple data channels, for further visual comparison. Large size of datasets, significant posture differences, and changes in patient anatomy turn manual matching in an extremely labor-intensive task that requires from radiologists high concentration for long period of time. This strongly limits the productivity and increases risk of underdiagnosis. Materials and methods We present a novel approach to the comparative visual analysis of whole-body MRI follow-up data. Our method is based on interactive derivation of locally rigid transforms from a pre-computed whole-body deformable registration. Using this approach, baseline and follow-up slices can be interactively matched with a single mouse click in the anatomical region of interest. In addition to the synchronized side-by-side baseline and matched follow-up slices, we have integrated four techniques to further facilitate the visual comparison of the two datasets: the “deformation sphere”, the color fusion view, the magic lens, and a set of uncertainty iso-contours around the current region of interest. Results We have applied our method to the study of cancerous bone lesions over time in patients with Kahler’s disease. During these studies, the radiologist carefully visually examines a large number of anatomical sites for changes. Our interactive locally rigid matching approach was found helpful in localization of cancerous lesions and visual assessment of changes between different scans. Furthermore, each of the features integrated in our software was separately evaluated by the experts. Conclusion We demonstrated how our method significantly facilitates examination of whole-body MR datasets in follow-up studies by enabling the rapid interactive matching of regions of interest and by the explicit visualization of change.

Purpose: Little is known about the way orbital fat slides and/or deforms during eye movements. We compared two deformation algorithms from a sequence of MRI volumes to visualize this complex behavior. Methods: Time-dependent deformation data were derived from motion-MRI volumes using Lucas and Kanade Optical Flow (LK3D) and nonrigid registration (B-splines) deformation algorithms. We compared how these two algorithms performed regarding sliding and deformation in three critical areas: the sclera-fat interface, how the optic nerve moves through the fat, and how the fat is squeezed out under the tendon of a relaxing rectus muscle. The efficacy was validated using identified tissue markers such as the lens and blood vessels in the fat. Results: Fat immediately behind the eye followed eye rotation by approximately one-half. This was best visualized using the B-splines technique as it showed less ripping of tissue and less distortion. Orbital fat flowed around the optic nerve during eye rotation. In this case, LK3D provided better visualization as it allowed orbital fat tissue to split. The resolution was insufficient to visualize fat being squeezed out between tendon and sclera. Conclusion: B-splines performs better in tracking structures such as the lens, while LK3D allows fat tissue to split as should happen as the optic nerve slides through the fat. Orbital fat follows eye rotation by one-half and flows around the optic nerve during eye rotation. Translational Relevance: Visualizing orbital fat deformation and sliding offers the opportunity to accurately locate a region of cicatrization and permit an individualized surgical plan.

Multi-fields are widely used in areas ranging from physical simulations to medical imaging. Illustrative visualization techniques can help to effectively communicate features of interest found in a given field. Current techniques for multi-field visualization are mostly focused on showing subsets of local attributes such as single values or vector directions, e.g., using colors, texture, streamlines or glyphs. Instead, we present an approach based on highlighting areas with similar characteristics, considering all attributes of the field. Our approach is example-based and interactive. A user simply selects a point within the field, upon which the system automatically derives the characteristic combination of attributes for that point. Our system then automatically creates a visualization highlighting areas within the field which are similar to the example point with respect to these characteristics. The visualizations are presented using sparse, illustrative techniques, using contours and colors to clearly delineate and identify separate areas. Users can interact with the visualizations in real-time, by moving the example point or, optionally, by changing the characteristics or adjusting other parameters used to determine similarity.

Statistical shape modeling is a widely used technique for the representation and analysis of the shapes and shape variations present in a population. A statistical shape model models the distribution in a high dimensional shape space, where each shape is represented by a single point. We present a design study on the intuitive exploration and visualization of shape spaces and shape models. Our approach focuses on the dual-space nature of these spaces. The high-dimensional shape space represents the population, whereas object space represents the shape of the 3D object associated with a point in shape space. A 3D object view provides local details for a single shape. The high dimensional points in shape space are visualized using a 2D scatter plot projection, the axes of which can be manipulated interactively. This results in a dynamic scatter plot, with the further extension that each point is visualized as a small version of the object shape that it represents. We further enhance the population-object duality with a new type of view aimed at shape comparison. This new "shape evolution view" visualizes shape variability along a single trajectory in shape space, and serves as a link between the two spaces described above. Our three-view exploration concept strongly emphasizes linked interaction between all spaces. Moving the cursor over the scatter plot or evolution views, shapes are dynamically interpolated and shown in the object view. Conversely, camera manipulation in the object view affects the object visualizations in the other views. We present a GPU-accelerated implementation, and show the effectiveness of the three-view approach using a number of realworld cases. In these, we demonstrate how this multi-view approach can be used to visually explore important aspects of a statistical shape model, including specificity, compactness and reconstruction error.

Deformation is a topic of interest in many disciplines. In particular in medical research, deformations of surfaces and even entire volumetric structures are of interest. Clear visualization of such deformations can lead to important insight into growth processes and progression of disease. We present new techniques for direct focus+context visualization of deformation fields representing transformations between pairs of volumetric datasets. Typically, such fields are computed by performing a non-rigid registration between two data volumes. Our visualization is based on direct volume rendering and uses the GPU to compute and interactively visualize features of these deformation fields in real-time. We integrate visualization of the deformation field with visualization of the scalar volume affected by the deformations. Furthermore, we present a novel use of texturing in volume rendered visualizations to show additional properties of the vector field on surfaces in the volume.

Objectives and background: Quantitative analysis of intracoronary optical coherence tomography (OCT) image data (QOCT) is currently performed by a time-consuming manual contour tracing process in individual OCT images acquired during a pullback procedure (frame-based method). To get an efficient quantitative analysis process, we developed a fully automatic three-dimensional (3D) lumen contour detection method and evaluated the results against those derived by expert human observers. Methods: The method was developed using Matlab (The Mathworks, Natick, MA). It incorporates a graphical user interface for contour display and, in the selected cases where this might be necessary, editing. OCT image data of 20 randomly selected patients, acquired with a commercially available system (Lightlab imaging, Westford, MA), were pulled from our OCT database for validation. Results: A total of 4,137 OCT images were analyzed. There was no statistically significant difference in mean lumen areas between the two methods (5.03 +/- 2.16 vs. 5.02 +/- 2.21 mm(2); p = 0.6, human vs. automated). Regression analysis showed a good correlation with an r value of 0.99. The method requires an average 2-5 sec calculation time per OCT image. In 3% of the detected contours an observer correction was necessary. Conclusion: Fully automatic lumen contour detection in OCT images is feasible with only a select few contours showing an artifact (3%) that can be easily corrected. This QOCT method may be a valuable tool for future coronary imaging studies incorporating OCT. (C) 2009 Wiley-Liss, Inc.

Computer-aided diagnosis (CAD) is a helpful addition to laborious visual inspection for preselection of suspected colonic polyps in virtual colonoscopy. Most of the previous work on automatic polyp detection makes use of indicators based on the scalar curvature of the colon wall and can result in many false-positive detections. Our work tries to reduce the number of false-positive detections in the preselection of polyp candidates. Polyp surface shape can be characterized and visualized using lines of curvature. In this paper, we describe techniques for generating and rendering lines of curvature on surfaces and we show that these lines can be used as part of a polyp detection approach. We have adapted existing approaches on explicit triangular surface meshes, and developed a new algorithm on implicit surfaces embedded in 3D volume data. The visualization of shaded colonic surfaces can be enhanced by rendering the derived lines of curvature on these surfaces. Features strongly correlated with true-positive detections were calculated on lines of curvature and used for the polyp candidate selection. We studied the performance of these features on 5 data sets that included 331 pre-detected candidates, of which 50 sites were true polyps. The winding angle had a significant discriminating power for true-positive detections, which was demonstrated by a Wilcoxon rank sum test with p < 0.001. The median winding angle and inter-quartile range (IQR) for true polyps were 7.817 and 6.770 - 9.288 compared to 2.954 and 1.995 - 3.749 for false-positive detections

Finding a correct a priori back-to-front (BTF) visibility order- ing for the perspective projection of the voxels of a rectangular volume poses interesting problems. The BTF ordering presented by Frieder et al. (6) and the permuted BTF presented by Westover (14) are correct for parallel projection but not for perspective projection (12). Swan presented a constructive proof for the correctness of the perspective BTF (PBTF) ordering (12). This was a significant improvement on the existing orderings. However, his proof assumes that voxel projections are not larger than a pixel, i.e. voxel projections do not overlap in screen space. Very often the voxel projec- tions do overlap, e.g. with splatting algorithms. In these cases, the PBTF ordering results in highly visible and characteristic rendering artefacts. In this paper we analyse the PBTF and show why it yields these ren- dering artefacts. We then present an improved visibility ordering that remedies the artefacts. Our new ordering is as good as the PBTF, but it is also valid for cases where voxel projections are larger than a single pixel, i.e. when voxel pro- jections overlap in screen space. We demonstrate why and how our ordering works at fundamental and implementation levels.

Illustrative visualization techniques deviate from physically-based rendering. They were developed to create renditions that consider perceptual capabilities of humans. As an example, humans infer information about shape not only from realistic shading but also from hatching and from outlines that support the mental separation of nearby objects rendered in similar colors. Illustrative techniques are also motivated by human cognitive processes, such as attention and situation awareness. Thus, the additional degrees of freedom of illustrative visualization should be used to emphasize what the author the visualization considers important. Illustrative visualization techniques are inspired by the work of professional scientific and medical illustrators, who developed an impressive set of techniques that depict objects and relations in an abstract manner. Boundary emphasis, non-realistic shading models, and feature lines, such as silhouettes are employed to emphasize essential features of anatomical structures. An essential prerequisite for most of these techniques is curvature estimation. Thus, we introduce curvature-related measures and describe curvature estimation techniques for polygonal surface models. Curvature-related measures, such as Gaussian curvature, are frequently used to filter feature and to restrict the display to regions with high curvature. We explain view-dependent and view-independent feature lines as well as feature lines that emphasize discontinuities of the illumination. Specific examples include ridge and valley lines, suggestive contours, apparent ridges, and photonic extremum lines. In addition to feature lines, stippling, and hatching techniques are introduced, since feature lines are often not sufficient to convey surface shapes. Efficient image generation and frame coherence are essential aspects of computer-generated hatching and stippling. Smart visibility techniques, such as cutaway views, breakaway views, and ghosted views, adapt the transparency of objects to make important structures visible. As an example, the surface of an organ may be displayed semitransparently in regions where a tumor is behind. However, in other regions it may be rendered fully opaque.

This chapter provides an introduction into the most essential medical image acquisition processes with a focus on tomographic 3D image data. As a prerequisite sampling processes and interpolation is explained. We briefly discuss major requirements that guide the selection of imaging modalities in practice, e.g., to provide a sufficient spatial resolution for the diagnostic question, to limit the exposure to radiation and to achieve a sufficient image quality in terms of contrast and signal-to-noise ratio (SNR). Readers should understand that trade-offs between these conflicting goals are necessary. Thus, neither optimum spatial resolution nor optimum image quality are relevant goals in clinical practice. In this chapter we focus on tomographic imaging modalities, in particular on CT and MRI data. Typical artifacts that hamper image interpretation, such as inhomogeneity, partial volume effects, and streaking are explained. We consider recent developments, e.g., dual source and dual energy CT and ultra highfield MRI. Hybrid PET/CT and PET/MRI scanners are an exciting development of the last decade. We discuss them as examples for the potential of the complementary use of imaging data and the necessity to fuse the resulting information. Specific examples of diagnostic questions that benefit from certain imaging techniques are given throughout the chapter. Finally, we consider imaging techniques for intraoperative monitoring, such as fluoroscopy and intravascular ultrasound.

Labeling and measurements enhance medical visualizations with useful information for documentation, follow-up examinations and educational purposes. Labeling and measuring may be performed interactively, e.g., by dragging an annotation line or selecting points for distance measurements. An alternative is the automatic placement of labels, where labels and annotation lines are placed such that they do not overlap with each other. Automatic labeling is essential if several anatomical structures, e.g., five liver metastases should be labeled. We explain labeling techniques for 2D slice-based visualizations as well as for 3D renderings. Tumor staging, the assessment of the severity of a vascular disease, the respectability of a patient, and many other essential tasks depend on a quantitative assessment of pathological structures and the spatial relations between them and adjacent anatomical structures. The diameter and length of a stenosis and the angle between bones after fracture are just a few examples of measures directly relevant to treatment decisions. In treatment monitoring, the essential goal is to assess whether a pathology shrinked down, stayed constant or grew despite treatment. Distance, angle, cross-sectional area and volume measurements are performed to support such tasks. Therefore, we discuss algorithms for efficiently and precisely determining minimum distances between anatomical structures as well as an algorithm to determine the longest diameter of a segmented object. We also discuss different variants of integrating these measures in 3D visualizations. Measurements always require a discussion of their accuracy, since unreliable measurements may mislead users.

In this chapter, we describe techniques to interactively explore volume data. Exploration in this context refers to the specification of which portions of the data should be visible and how they should be displayed. The techniques we describe are used to modify the brightness, contrast and opacity of voxels, and thus the contrast between different materials. Transfer functions, clipping planes, and more advanced virtual resection techniques are amongst these techniques. A crucial problem is to provide user interfaces and interaction techniques that enable physicians to specify their intent as directly as possible. The intent of physicians is to emphasize certain tissues, material boundaries, or spatial relations between adjacent structures, e.g., around a pathology. Simple user interfaces, however, just provide a transfer function editor, where the user has to spend much effort in order to come up with a visualization that is at least close to the above-mentioned visualization goals. Moreover, physicians think in terms of diagnostic and treatment decisions, e.g., with respect to infiltration and resectability. The raw parameters of a transfer function are too low level to enable convenient assistance for such questions. While traditional transfer functions are one-dimensional and global, we also discuss more advanced strategies since these allow to discriminate tissues better. Two-dimensional transfer functions, using gradient magnitude or distances to a target structure are specific examples that overcome the restriction of intensity-based 1D transfer functions. We put special emphasis on recently introduced techniques, such as size-based transfer functions, visibility-driven transfer functions, and local transfer functions, e.g., those that employ LH (low/high) histograms. Several of these advanced techniques employ clustering as an analysis technique to identify regions with similar features. Therefore we briefly introduce a number of clustering techniques and discuss their potential and limitations. Volume interaction also comprises techniques to geometrically specify which portions of the data are visible, e.g., with clipping planes or cutting geometries that are moved inside the data. In this way, virtual resections can be performed in order to define which portions of an organ remain after a pathology is removed from a certain area with a certain safety margin. A basic interaction technique is picking where a certain anatomical structure is selected as a prerequisite for further interaction. Picking suffers from ambiguities since several structures are partially visible at each screen space location. We describe contextual picking, as a knowledge-based technique, and perception-based techniques to disambiguate selections.

In medical visualization, there are a number of mapping techniques that are employed to reduce 3D data, either by projection or reformation, to effective 2D visual representations. Straightforward projection, as employed in the standard 3D graphics pipeline, also serves to reduce 3D data to a 2D representation. However, in this chapter we discuss examples that go a step further in terms of complexity and effectiveness. We have classified these reductive techniques broadly into two groups: projections and reformations. In projections, 3D information is accumulated, or flattened, along one or more directions to obtain a 2D image. In reformations, 3D information is sampled along some geometry that can later be flattened, and the sampled information along with it. There are examples that have characteristics of more than one class. The Bull’s Eye Plot, that is frequently used in cardiology, is an example of a projection (data is aggregated through the heart wall), whereas multiplanar reformatting (MPR), is a straightforward example of reformation.

Surface rendering is one of the two major rendering modes for medical volume data. Instead of classifying volume data and mapping it directly to the viewport, surface rendering is based on an indirect surface mesh representation. This mesh is either generated by extracting an isosurface from the original volume data or by transforming a segmentation result. Surface rendering of volume data was primarily motivated by two issues: it was much faster than volume rendering, since a mesh has a significantly lower memory footprint than a volume dataset and it provides clearly recognizable images with depth cues such as those caused by illumination. These original advantages have somehow lost their relevance when GPU rendering became more and more powerful and could be used to support all stages of direct volume rendering. Thus, even advanced volume rendering of typical medical datasets can be performed in real-time. In the last years, three developments occurred that raised the importance of surface extraction and surface rendering: • With the increased use of biophysical simulation, there is a need for surface meshes as a basis for volume grids that are used for simulations. • Interactive 3D visualizations are used in web browsers and mobile settings. Although direct volume rendering is also feasible in such settings, surface renderings are prevailing due to the lower memory requirements. • Surface mesh construction becomes increasingly important for 3D printing. For challenging treatment planning questions, relevant portions of the human anatomy are modeled and printed in 3D to enable an in-depth collaborative discussion of treatment options. In this chapter, we discuss surface extraction with respect to efficiency, accuracy and smoothness. This includes a discussion of GPU-based mesh extraction. We introduce a variety of mesh smoothing algorithms that process a surface mesh and reduce artifacts and staircases. Advanced mesh smoothing algorithms may be locally adapted and control volume shrinkage or other excessive modifications that hamper the accuracy. Mesh simplification algorithms are introduced to support simulation applications as well as web-based use of medical surface representations.

For diagnosis as well as for many therapy planning tasks it is crucial to understand the branching pattern and morphology of tree-like anatomical structures, such as nerves, vascular, and bronchial trees. A slight difference in vascular topology may lead to a strong difference in the treatment strategy. For therapy planning, it is of paramount importance to recognize shape features and morphology of vascular structures as well as spatial relations between vascular and other relevant structures. For a convenient interpretation, the curvature, the depth relations, and the diminution of the diameter toward the periphery should be depicted correctly. For diagnosis it is essential to reliably recognize and assess vessel wall abnormalities, such as narrowings (stenosis) or dilated parts (aneurysms) as well as plaque and thrombus formation. Primarily CT and MR angiography are employed. Digital subtraction angiography (DSA), a variant of X-ray, is also available as a 3D imaging technique. We focus on processing and visualizing information derived from CT and MR angiography data. In particular for diagnosis, 2D views of the original data, derived information, such as oblique reformations and changes of the vessel diameter, need to be integrated carefully with 3D views. We briefly explain image analysis (segmentation and centerline extraction) techniques, which are a prerequisite for high-quality vessel visualization. We continue with surface visualization techniques using the vessel segmentation result. Model-based and illustrative surface visualizations of vascular trees serve primarily for treatment planning. They are based on model assumptions, such as a circular cross section. In contrast, model-free approaches do not enforce any assumptions but display vascular structures, as they have been segmented. To achieve a high visual quality and ease interpretation, again, implicit surface visualization techniques are considered. Adapted volume rendering also has a great potential for visualizing vascular structures in particular for diagnostic purposes. We present strategies for transfer function specification and local volume rendering techniques, using the segmentation result to emphasize certain regions.

In direct volume visualization, the volumetric dataset is directly represented visually without generating an intermediate representation. This is performed by somehow projecting the volumetric information onto the viewing plane. In this chapter we explain the major strategies and variations that can be employed for this projection. We will focus on the theoretical aspects of direct volume visualization, in particular the employed physical model and the mathematical foundation of the resulting volume rendering equation. Volume rendering relies on numerical approximations of this integral. Volume rendering approaches differ in the processing order of the pipeline that evaluates the volume rendering equation and its components such as sampling and compositing. We explain the volume raycasting algorithm, after which we compactly treat the implementation of direct volume raycasting on the GPU. Despite the classical volume rendering that provides essential depth cues, we also introduce compositing variations, such as maximum intensity projection and X-ray projection. Although these methods do provide only limited depth cues when handling data interactively, there are specific use cases, in particular for the maximum intensity projection that is frequently used for the assessment of vascular structures. Efficient rendering, e.g., by using hierarchical data structures or early ray termination (in volume raycasting) are briefly explained. Slab rendering, where instead of the full volume only a limited set of slices is rendered at the same time, is often an appropriate trade-off between slice-based visualization and 3D rendering. PreIntegrated volume rendering is explained since it leads to a particularly high image quality.

In small animal imaging studies, when the locations of the micro-structures of interest are unknown a priori, there is a simultaneous need for full-body coverage and high resolution. In MRI, additional requirements to image contrast and acquisition time will often make it impossible to acquire such images directly. Recently, a resolution enhancing post-processing technique called super-resolution reconstruction (SRR) has been demonstrated to improve visualization and localization of micro-structures in small animal MRI by combining multiple low-resolution acquisitions. However, when the field-of-view is large relative to the desired voxel size, solving the SRR problem becomes very expensive, in terms of both memory requirements and computation time. In this paper we introduce a novel local approach to SRR that aims to overcome the computational problems and allow researchers to efficiently explore both global and local characteristics in whole-body small animal MRI. The method integrates state-of-the-art image processing techniques from the areas of articulated atlas-based segmentation, planar reformation, and SRR. A proof-of-concept is provided with two case studies involving CT, BLI, and MRI data of bone and kidney tumors in a mouse model. We show that local SRR-MRI is a computationally efficient complementary imaging modality for the precise characterization of tumor metastases, and that the method provides a feasible high-resolution alternative to conventional MRI.

Purpose Automated patient-specific image-based segmentation of tissues surrounding aseptically loose hip prostheses is desired. For this we present an automated segmentation pipeline that labels periprosthetic tissues in computed tomography (CT). The intended application of this pipeline is in pre-operative planning. Methods Individual voxels were classified based on a set of automatically extracted image features. Minimum-cost graph cuts were computed on the classification results. The graph-cut step enabled us to enforce geometrical containment constraints, such as cortical bone sheathing the femur’s interior. The solution’s novelty lies in the combination of voxel classification with multilabel graph cuts and in the way label costs were defined to enforce containment constraints. Results The segmentation pipeline was tested on a set of twelve manually segmented clinical CT volumes. The distribution of healthy tissue and bone cement was automatically determined with sensitivities greater than 82% and pathological fibrous interface tissue with a sensitivity exceeding 73%. Specificity exceeded 96% for all tissues. Conclusions The addition of a graph-cut step improved segmentation compared to voxel classification alone. The pipeline described in this paper represents a practical approach to segmenting multitissue regions from CT.

The visualisation of vector fields is essential for many applications in science, engineering and biomedicine. A large number of vector icons has been developed, but little research has been done on their effectiveness, especially when visualising multiple vector fields simultaneously. We apply research in visualisation and cognitive science to identify four classes of post-processing techniques for visualising two 2D vector fields simultaneously: blending, overlay, bump mapping, and masking. We apply these four post-processing methods to Line Integral Convolution (LIC) textures and thus develop several novel multi-field visualisation techniques. We evaluate their effectiveness with a user study requiring participants to locate and classify singularities, and to rate each method on its effectiveness and aesthetic value. The results of the study suggest that blending is the most effective technique to combine multiple vector field visualisation textures, while masking performs worst. There is some evidence that visualisations with smooth colour changes are perceived as visually more attractive, and that aesthetics increases the perceived effectiveness of a visualisation technique.

Super-resolution reconstruction (SRR) is a post-acquisition method for producing a high-resolution (HR) image from a set of low-resolution (LR) images. However, for large volumes of data, this technique is computationally very demanding and time consuming. In this study we focus on the specific case of whole-body mouse data and present a novel, integrated, end-to-end approach to overcome this problem. We combine articulated atlas-based segmentation and planar reformation techniques with state-of-the-art in SRR to produce high resolution, interactively selected, localized isotropic volumes-of-interest in whole-body mouse MRI. With this method we overcome time and memory related limitations when applying the SRR algorithm to the entire dataset, enabling interactive visualization and exploration of anatomical structures of interest in whole-body MRI mouse data on a normal desktop PC.

Multidimensional projection has emerged as an important visualization tool in applications involving the visual analysis of high-dimensional data. However, high precision projection methods are either computationally expensive or not flexible enough to enable feedback from user interaction into the projection process. A built-in mechanism that dynamically adapts the projection based on direct user intervention would make the technique more useful for a larger range of applications and data sets. In this paper we propose the Piecewise Laplacian-based Projection (PLP), a novel multidimensional projection technique, that, due to the local nature of its formulation, enables a versatile mechanism to interact with projected data and to allow interactive changes to alter the projection map dynamically, a capability unique of this technique. We exploit the flexibility provided by PLP in two interactive projection-based applications, one designed to organize pictures visually and another to build music playlists. These applications illustrate the usefulness of PLP in handling high-dimensional data in a flexible and highly visual way. We also compare PLP with the currently most promising projections in terms of precision and speed, showing that it performs very well also according to these quality criteria.

Nested or intersecting surfaces are proven techniques for visualizing shape differences between static 3D objects (Weigle and Taylor II, IEEE Visualization, Proceedings, pp. 503–510, 2005). In this paper we present an image-based formulation for these techniques that extends their use to dynamic scenarios, in which surfaces can be manipulated or even deformed interactively. The formulation is based on our new layered rendering pipeline, a generic image-based approach for rendering nested surfaces based on depth peeling and deferred shading. We use layered rendering to enhance the intersecting surfaces visualization. In addition to enabling interactive performance, our enhancements address several limitations of the original technique. Contours remove ambiguity regarding the shape of intersections. Local distances between the surfaces can be visualized at any point using either depth fogging or distance fields: Depth fogging is used as a cue for the distance between two surfaces in the viewing direction, whereas closest-point distance measures are visualized interactively by evaluating one surface’s distance field on the other surface. Furthermore, we use these measures to define a three-way surface segmentation, which visualizes regions of growth, shrinkage, and no change of a test surface compared with a reference surface. Finally, we demonstrate an application of our technique in the visualization of statistical shape models. We evaluate our technique based on feedback provided by medical image analysis researchers, who are experts in working with such models.

This paper discusses a number of image analysis challenges emerging from longitudinal small animal molecular imaging studies. Three steps towards a quantitative 3D analysis of follow-up small animal imaging are presented: whole-body articulated registration, change visualization in follow-up data and fusion of optical and 3D structural imaging data. Several application examples are presented in the context of translational cancer research.

Kinematics is the analysis of motions without regarding forces or inertial effects, with the purpose of understanding joint behaviour. Kinematic data of linked joints, for example the upper extremity, i.e. the shoulder and arm joints, contains many related degrees of freedom that complicate numerical analysis. Visualisation techniques enhance the analysis process, thus improving the effectiveness of kinematic experiments. This paper describes a new visualisation system specifically designed for the analysis of multi-joint kinematic data of the upper extremity. The challenge inherent in the data is that the upper extremity is comprised of five cooperating joints with a total of fifteen degrees of freedom. The range of motion may be affected by subtle deficiencies of individual joints that are difficult to pinpoint. To highlight these subtleties our approach combines interactive filtering and multiple visualisation techniques. Our system is further differentiated by the fact that it integrates simultaneous acquisition and visual analysis of biokinematic data. Also, to facilitate complex queries, we have designed a visual query interface with visualisation and interaction elements that are based on the domain-specific anatomical representation of the data. The combination of these techniques form an effective approach specifically tailored for the investigation and comparison of large collections of kinematic data. This claim is supported by an evaluation experiment where the technique was used to inspect the kinematics of the left and right arm of a patient with a healed proximal humerus fracture, i.e. a healed shoulder fracture.

The analysis of multi-timepoint whole-body small animal CT data is greatly complicated by the varying posture of the subject at different timepoints. Due to these variations, correctly relating and comparing corresponding regions of interest is challenging. In addition, occlusion may prevent effective visualization of these regions of interest. To address these problems, we have developed a method that fully automatically maps the data to a standardized layout of sub-volumes, based on an articulated atlas registration. We have dubbed this process articulated planar reformation, or APR. A sub-volume can be interactively selected for closer inspection and can be compared with the corresponding sub-volume at the other timepoints, employing a number of different comparative visualization approaches. We provide an additional tool that highlights possibly interesting areas based on the change of bone density between timepoints. Furthermore we allow visualization of the local registration error, to give an indication of the accuracy of the registration. We have evaluated our approach on a case that exhibits cancer-induced bone resorption.

Purpose Segmentation of rheumatoid joints from CT images is a complicated task. The pathological state of the joint results in a non-uniform density of the bone tissue, with holes and irregularities complicating the segmentation process. For the specific case of the shoulder joint, existing segmentation techniques often fail and lead to poor results. This paper describes a novel method for the segmentation of these joints. Methods Given a rough surface model of the shoulder, a loop that encircles the joint is extracted by calculating the minimum curvature of the surface model. The intersection points of this loop with the separate CT-slices are connected by means of a path search algorithm. Inaccurate sections are corrected by iteratively applying a Hough transform to the segmentation result. Results As a qualitative measure we calculated the Dice coefficient and Hausdorff distances of the automatic segmentations and expert manual segmentations of CT-scans of ten severely deteriorated shoulder joints. For the humerus and scapula the median Dice coefficient was 98.9% with an interquartile range (IQR) of 95.8-99.4 and 98.5% (IQR 98.3-99.2%), respectively. The median Hausdorff distances were 3.06 mm (IQR 2.30-4.14) and 3.92 mm (IQR 1.96 5.92 mm), respectively. Conclusion The routine satisfies the criterion of our particular application to accurately segment the shoulder joint in under 2 min. We conclude that combining surface curvature, limited user interaction and iterative refinement via a Hough transform forms a satisfactory approach for the segmentation of severely damaged arthritic shoulder joints.

CT colonography is a screening technique for adenomatous colorectal polyps, which are important precursors to colon cancer. Computer aided detection (CAD) systems are developed to assist radiologists. We present a CAD system that substantially reduces false positives while keeping the sensitivity high. Hereto, we combine protrusion measures derived from the solution of a non-linear partial differential equation (PDE) applied to both an explicit mesh and an implicit volumetric representation of the colon wall. The shape of the protruding elements is efficiently described via a technique from data visualization based on curvature streamlines. A low-complex pattern recognition system based on an intuitive feature from the aforementioned representations improves performance to less than 1.6 false positives per scan at 92% sensitivity per polyp.

Quantitative analysis on intracoronary optical coherence tomography (OCT) image data (e.g. QOCT) is currently performed by a time-consuming manual contour tracing process in many individual OCT images acquired during a pullback procedure (frame-based method). In order to get a more efficient quantitative analysis process and to investigate the possibilities to use this contour information for development of an image based retrospective OCT-gating method, as a first step a novel approach has been developed that exploits a full automated contour tracing method for coronary lumens in OCT images without the need for intensive observer related actions. This study presents this new method, which was tested on in-vivo acquired human coronary OCT image data of 10 randomly selected patients.

Many computer aided diagnosis (CAD) schemes have been developed for colon cancer detection using Virtual Colonoscopy (VC). In earlier work, we developed an automatic polyp detection method integrating flow visualization techniques, that forms part of the CAD functionality of an existing Virtual Colonoscopy pipeline. Curvature streamlines were used to characterize polyp surface shape. Features derived from curvature streamlines correlated highly with true polyp detections. During testing with a large number of patient data sets, we found that the correlation between streamline features and true polyps could be affected by noise and our streamline generation technique. The seeding and spacing constraints and CT noise could lead to streamline fragmentation, which reduced the discriminating power of our streamline features. In this paper, we present two major improvements of our curvature streamline generation. First, we adapted our streamline seeding strategy to the local surface properties and made the streamline generation faster. It generates a significantly smaller number of seeds but still results in a comparable and suitable streamline distribution. Second, based on our observation that longer streamlines are better surface shape descriptors, we improved our streamline tracing algorithm to produce longer streamlines. Our improved techniques are more effcient and also guide the streamline geometry to correspond better to colonic surface shape. These two adaptations support a robust and high correlation between our streamline features and true positive detections and lead to better polyp detection results.

Orbital fat, or the fat behind the eye, plays an important role in eye movements. In order to gain a better understanding of orbital fat mobility during eye motion, MRI datasets of the eyes of two healthy subjects were acquired respectively in seven and fourteen different directions of gaze. After semi-automatic rigid registration, the Demons deformable registration algorithm was used to derive time-dependent three-dimensional deformation vector fields from these datasets. Visualisation techniques were applied to these datasets in order to investigate fat mobility in specific regions of interest in the first subject. A qualitative analysis of the first subject showed that in two of the three regions of interest, fat moved half as much as the embedded structures. In other words, when the muscles and the optic nerve that are embedded in the fat move, the fat partly moves along with these structures and partly flows around them. In the second subject, a quantitative analysis was performed which showed a relation between the distance behind the sciera and the extent to which fat moves along with the optic nerve.

This paper explores new methods to visualize and fuse multi-2D bioluminescence imaging (BLI) data with structural imaging modalities such as micro CT and MR. A geometric, back-projection-based 3D reconstruction for superficial lesions from multi-2D BLI data is presented, enabling a coarse estimate of the 3D source envelopes from the multi-2D BLI data. Also, an intuitive 3D landmark selection is developed to enable fast BLI / CT registration. Three modes of fused BLI / CT visualization were developed: slice visualization, carousel visualization and 3D surface visualization. The added value of the fused visualization is demonstrated in three small-animal experiments, where the sensitivity of BLI to detect cell clusters is combined with anatomical detail from micro-CT imaging.

Diffusion tensor imaging (DTI) is an MRI-based technique for quantifying water diffusion in living tissue. In the white matter of the brain, water diffuses more rapidly along the neuronal axons than in the perpendicular direction. By exploiting this phenomenon, DTI can be used to determine trajectories of fiber bundles, or neuronal connections between regions, in the brain. The resulting bundles can be visualized. However, the resulting visualizations can be complex and difficult to interpret. An effective approach is to pre-determine trajectories from a large number of positions throughout the white matter (full brain fiber tracking) and to offer facilities to aid the user in selecting fiber bundles of interest. Two factors are crucial for the use and acceptance of this technique in clinical studies: firstly, the selection of the bundles by brain experts should be interactive, supported by real-time visualization of the trajectories registered with anatomical MRI scans. Secondly, the fiber selections should be reproducible, so that different experts will achieve the same results. In this paper we present a practical technique for the interactive selection of fiber-bundles using multiple convex objects that is an order of magnitude faster than similar techniques published earlier. We also present the results of a clinical study with ten subjects that show that our selection approach is highly reproducible for fractional anisotropy (FA) calculated over the selected fiber bundles.

For the visualisation of time-dependent data sets, interactive isosurface extraction and rendering is desirable. It allows the user to study the development of a surface shape in time, such as a moving front or an evolving object shape. For this purpose, the user must be able to interactively specify an isovalue, and a sequence of isosurfaces must be visualised, starting from any time step, in forward or backward direction in time. In this paper, we describe efficient and tightly coupled techniques for time-dependent isosurface extraction and rendering at interactive frame rates. In preprocessing, we create data structures from a time-dependent data set that allow real-time extraction of all isovalue-spanning cells, achieving rates of several hundreds of frames per second. These isovalued cells are then passed to a fast hardware-assisted direct point rendering algorithm for display, thus avoiding time expensive surface construction by triangulation. This algorithm makes effective use of the available graphics hardware.

Direct volume rendering (DVR) is a very useful visualisation technique. However, the difficulty in specifying a suitable transfer function often discourages its application to visualisation problems. This work presents a technique for providing meaningful and fast visual feedback that greatly facilitates the transfer function specification process. The feedback is based on a slice-based preview of the actual volume rendering that can be calculated in real-time and can be superimposed on a slice-based view of the data that is being rendered.

For successful ball-joint replacement surgery, it is important to maintain the joint's geometric center. Pre-operative detection of this center is achieved by detecting the sphere that fits onto the articular surfaces in CT or MRI images. We have developed a novel technique to automatically determine the sub-voxel position and size of a sphere in unsegmented 3D images. The method is invariant to size and robust to noise. It only needs one fourth of a sphere to detect the center. Isotropically as well as anisotropically sampled images can be used. As no segmentation is required, it can be applied directly to clinical images.