Skip to main content

Über dieses Buch

The visualization of human anatomy for diagnostic, therapeutic, and educational pur­ poses has long been a challenge for scientists and artists. In vivo medical imaging could not be introduced until the discovery of X-rays by Wilhelm Conrad ROntgen in 1895. With the early medical imaging techniques which are still in use today, the three-dimensional reality of the human body can only be visualized in two-dimensional projections or cross-sections. Recently, biomedical engineering and computer science have begun to offer the potential of producing natural three-dimensional views of the human anatomy of living subjects. For a broad application of such technology, many scientific and engineering problems still have to be solved. In order to stimulate progress, the NATO Advanced Research Workshop in Travemiinde, West Germany, from June 25 to 29 was organized. It brought together approximately 50 experts in 3D-medical imaging from allover the world. Among the list of topics image acquisition was addressed first, since its quality decisively influences the quality of the 3D-images. For 3D-image generation - in distinction to 2D­ imaging - a decision has to be made as to which objects contained in the data set are to be visualized. Therefore special emphasis was laid on methods of object definition. For the final visualization of the segmented objects a large variety of visualization algorithms have been proposed in the past. The meeting assessed these techniques.



Image Acquisition


Magnetic Resonance Imaging

Magnetic Resonance Imaging within the past 10 years has become the primary medical imaging modality for many neurological and orthopedic applications. Starting from the basic phenomena of Nuclear Magnetic Resonance, the principles of 2D and 3D Magnetic Resonance Imaging are described. As a new and possibly important future application the basics of Magnetic Resonance Angiography allowing the display of vascular structures without the use of contrast agents is being elaborated.
Wilfried K. Loeffler

3D Echography: Status and Perspective

3D echography is still in its infancy but recent preliminary clinical results reveal exciting potentialities. Echography presents specific characteristics which have to be exploited when designing a 3D ultrasound system. After reviewing the various schemes which have been followed to reliably acquire 3D data, the possible exploitation of these volumic data will be described, insisting on the advantages / limitations of the various possibilities. Finally, a survey of the still meager clinical investigations will be made.
F. Hottier, A. Collet Billon

Object Definition


Network Representation of 2-D and 3-D Images

Much of the data and knowledge representation in image interpretation needs to be in the form of networks. We propose that a network representation of the grey level changes in an image should be constructed as early as possible during bottom-up processing. We refer to this low level description as the image structure representation. The image is treated as a continuous surface made up of triangular facets (in 2-D) or as a continuous volume of tetrahedrons (in 3-D). The image is completely segmented into nonoverlapping slope districts. Each slope district is anchored between one peak and one pit and (usually) two or more saddle points. Slope districts can easily be grouped in several ways to form more meaningful entities such as edge support regions, ridges, convex corners etc.
A. C. F. Colchester

Segmentation and Analysis of Multidimensional Data-Sets in Medicine

Segmentation and analysis of 2-D and 3-D CT and MR data sets is illustrated on examples covering renal function studies in MR time sequences, discrimination of brain matter in multispectral 3-D MRI, extraction of brain tumor by structure segmentation methods, and model guided analysis of hip-joint and pelvic anatomy. It is argued in favor of a tightly integrated analysis sequence which exploits the characteristics of the original measurements and all available anatomic information. Knowledge-based analysis has produced some preliminary results but must be considered still in its infancy. Novel ideas for describing and representing natural structures in conjunction with AI tools are needed for generally applicable analysis methods.
Olaf Kübler, Guido Gerig

Toward Interactive Object Definition in 3D Scalar Images

We present a method of object definition designed to allow fast interactive definition of object regions in 2D and 3D image data by a human user based on an automatically computed image description of sensible image regions. The image description is a quasi-hierarchy of ridges (or courses) and subridges (or subcourses) in the intensity surface corresponding to the image. Two methods of ridge computation are presented, one based on the intensity axis of symmetry and another based on flow lines in the intensity surface. A system for interactive object definition using this approach is described, and the use of this approach on a variety of medical images is evaluated. Generalizations of these descriptions and the interactive object definition tool to 3D are discussed.
Stephen M. Pizer, Timothy J. Cullip, Robin E. Fredericksen

Steps Toward the Automatic Interpretation of 3D Images

We describe in this paper part of the research being performed at Inria on the automatic interpretation of three-dimensional images. We identify three common key problems which we call segmentation, representation, and matching of 3D regions. We describe our approach for solving these problems, our current results on 3D medical images, and give the trends of our future work.
N. Ayache, J. D. Boissonnat, L. Cohen, B. Geiger, J. Levy-Vehel, O. Monga, P. Sander

Image Processing of Routine Spin-Echo MR Images to Enhance Vascular Structures: Comparison with MR Angiography

Magnetic Resonance Imaging (MRI) is a highly flexible diagnostic imaging technique providing complex information about the morphology of various normal and abnormal tissues. Besides this clinically valuable anatomic and pathologic information, there are some unique functional features which can be extracted from MR images. The influence of macroscopic and microscopic motion on MR images as revealed by more or less specialized pulse sequences may demonstrate the presence of physiologically important processes such as blood or cerebrospinal fluid (CSF) flow, tissue perfusion and diffusion [Axel84, Demoulin87, Demoulin89, Haacke89, Wehrli87]. Recent progress in the implementation of these MRI methods shows that while these applications have not yet been fully exploited, their use in clinical practice is not far in the future.
The use of image processing techniques can definitely improve the visualization, display and interpretation of MR images. The computerized post-processing techniques may also have a very significant role in accessing the physiologic information available from MRI. In this work we applied image processing techniques to both routine, standard spin-echo MR acquisitions and to MR angiograms (MRA) in order to extract and display vascular structures with intraluminal flow. We demonstrated that it is possible to segment a standard spin-echo data set into brain parenchyma (white and gray matter), cerebrospinal fluid (CSF) and vascular structures. The segmented images can be displayed using 3D surface rendering and selective clipping. The extracted vascular structures obtained with the MRI and MRA methods were compared.
While the vessels were better delineated by the MRA method, especially after image processing, the major morphologic features were also accessible after segmentation of the standard spin-echo images.
Guido Gerig, Ron Kikinis, Ferenc A. Jolesz

A Multispectral Pattern Recognition System for the Noninvasive Evaluation of Atherosclerosis Utilizing MRI

This paper describes an image processing, pattern recognition and computer graphics system for the noninvasive identification and evaluation of atherosclerosis using multidimensional Magnetic Resonance Imaging (MRI). Particular emphasis has been placed on the problem of developing a pattern recognition system for noninvasively identifying the different plaque classes involved in atherosclerosis using minimal a priori information. This pattern recognition technique involves an extension of the ISODATA clustering algorithm to include an information theoretic criterion (Consistent Akaike Information Criterion) to provide a measure of the fit of the cluster composition at a particular iteration to the actual data. A rapid 3-D display system is also described for the simultaneous display of multiple data classes resulting from the tissue identification process. This work demonstrates the feasibility of developing a “high information content” display which will aid in the diagnosis and analysis of the atherosclerotic disease process. Such capability will permit detailed and quantitative studies to assess the effectiveness of therapies, such as drug, exercise and dietary regimens.
Michael B. Merickel, Theodore Jackson, Charles Carman, James R. Brookeman, Carlos R. Ayers

3D Reconstruction of High Contrast Objects Using a Multi-scale Detection / Estimation Scheme

Reconstructing a three-dimensional (3D) volume from a set of twodimensional X-ray projections raises theoretical, instrumental and computational difficulties. Focused on high contrast objects, solutions are proposed for the successive steps of a 3D reconstruction procedure, from the raw measurements on an image intensifier up to the reconstruction algorithm based on a multi-scale detection / estimation scheme.
Didier Saint-Félix, Yves Trousset, Catherine Picard, Anne Rougée

Matching Free-form Primitives with 3D Medical Data to Represent Organs and Anatomical Structures

Three-dimensional representations of organs and anatomical structures can be derived from sets of two-dimensional data (images) in different ways. Most approaches are based on the use of voxels (to provide digital representations of volumes) or triangles (to sample surfaces that interpolate sets of contours). The approach we have developed is different: it is basically interactive and thus it takes advantage of user’s expertise to define specific and highly-structured geometrical models of organs. The first problem is to define free-form primitives as basic components for the creation of three-dimensional models. The second problem is to make these primitives match as well as possible with actual data: an interactive procedure enables the user to roughly position the primitive onto the data; then, an automatic process selects the relevant data and optimizes the shape of the primitive. In this paper, we describe this interactive approach and the problems it raises; then, we focus on the description of an original solution to optimize the shape of free-form primitives through actual data.
Jean Sequeira, Franck Pinson



A Survey of 3D Display Techniques to Render Medical Data

This paper gives a brief survey of 3D display techniques which were developped during the last fifteen years to render medical data. These methods are classified in two families surface rendering and volume rendering. Emphasis is done on the two most recent methodologies which become widely used to render medical volumetric data sets: Ray-Tracing and Octree Encoding.
Jean-Louis Coatrieux, Christian Barillot

Rendering Tomographic Volume Data: Adequacy of Methods for Different Modalities and Organs

A large variety of different algorithms for rendering radiological volume data has been described in the past. It has turned out that not every method is equally well suited for all kinds of modalities and organs. We therefore compare algorithms for different applications. The algorithms include Z-buffer gradient shading, gray level gradient shading, transparent gray level gradient shading, and two mutations of the marching cubes algorithm. The object classes are bone (CT), head (MRI), heart (MRI), vessels (MRI), ultrasound, and isodoses in radiotherapy. Properties of the algorithms, such as the achieved realism, the flexibility, and the computational cost are discussed. As a general result we found that the best visualization is achieved when different algorithms appropriate to the objects can be applied within one image.
Karl Heinz Höhne, Michael Bomans, Andreas Pommert, Martin Riemer, Ulf Tiede, Gunnar Wiebecke

Intermixing Surface and Volume Rendering

Two complementary methodologies for volume visualization, surface rendering and volume rendering, can be employed constructively together. In many situations, as in 3D medical imaging, the voxel-based sampled image need to be visualized together with synthetic surface-based objects such as surgical cuts, prosthesis, scalpels, and radiation beams. Four approaches for intermixing geometric models with sampled 3D medical data are presented. Details are provided on a hybrid approach that employs a Z-merging algorithm.
Arie Kaufman, Roni Yagel, Daniel Cohen

Combined 3D-Display of Cerebral Vasculature and Neuroanatomic Structures in MRI

Neuroradiologists and neurosurgeons in many clinical situations take a vital interest in images revealing not only intracranial vasculature, but also other neuroanatomic structures. With MRI, high-quality volume datasets may be acquired containing the necessary anatomic information. 3D-postprocessing methods are required in order to generate integrated displays of vessels and brain structures. With respect to this issue in this paper a 3D-visualization approach is proposed which has been designed according to the principles of robustness, user-friendliness and clinical practicality. It includes measurement techniques for the acquisition of flow-compensated angiographic datasets and T1-weighted tissue volumes as well as image processing methods operating on the ray-tracing principle.
Hans-Heino Ehricke, Gerhard Laub

Preliminary Work on the Interpretation of SPECT Images with the Aid of Registered MR Images and an MR Derived 3D Neuro-anatomical Atlas

This paper describes two methods to aid interpretation and quantification of SPECT or PET images. In the first method 3D SPECT or PET data sets are aligned and scaled to a 3D MRI data set of the same patient using 4 skin markers visible on each modality. Three display schemes have been implemented for viewing the aligned slices. Examples of these displays are provided. The second method uses a labelled 3D MRI reference data set from a volunteer to identify major anatomical structures. The MR reference data set is aligned with the isotope image using the same 4 markers plus a marker on the vertex of the skull. The reference data set is segmented approximately into the major tissue types - cerebrospinal fluid (CSF) and grey and white matter. Major structures are identified via labels in the 3D data set. A linked cursor aids delineation of anatomical regions on the isotope image using the outline of structures on the reference data set as a template. Directions for future research in the generation of complete digital anatomical atlases, which include inter-individual variations, are outlined.
David J. Hawkes, Derek L. G. Hill, Eldon D. Lehmann, Glynn P. Robinson, Michael N. Maisey, Alan C. F. Colchester

Volume Visualization of 3D Tomographies

With the aid of computer-supported 3D visualization procedures, such as triangulation [7, 3], Cuberille method [9, 10, 11, 4] or volume rendering approaches [5, 18, 20, 26, 13, 12, 8] 3D images can be generated from spatially associated tomography image series. These 3D images assist the physician in diagnosing, therapy planning and therapy control.
In this paper, a visualizing procedure is introduced which is based on a ray tracing approach by Kajiya [15, 16]. This model was modified and simplified according to medical requirements so that expressive images can be produced at a relatively low expense. Next to the presentation of surfaces, this visualization procedure also permits the presentation of tissues in a translucent form. Thus, organs may be shown in their environment in context.
The visualization of organs in a volume of data implies the association of the voxels with organs. In medical applications, it is useful to apply soft classifiers, e.g., the selflearning, selforganizing topological map [17, 1, 21]. It is, furthermore, necessary to integrate anatomical and morphological knowledge by AI methods.
H. P. Meinzer, U. Engelmann, D. Scheppelmann, R. Schäfer

Echocardiographic Three-Dimensional Visualization of the Heart

To perform three-dimensional (3-D) reconstruction of the heart by ultrasound, we developed a novel rotating echocardiographic probe which, with computer assistance, allows “real” 3-D reconstruction of the beating heart from 62 standard fan shaped two-dimensional (2-D) images acquired at 2.903 degree increments of rotation around its central axis. To reconstruct 3-D images of the beating heart, an entire cardiac cycle was recorded from each transducer position with electrocardiographic gating; acquisition time is 75 to 123 seconds in normal sinus rhythm. For each frame of the cardiac cycle, the 62 images digitized in cylindrical coordinates were processed by a scan converter algorithm to reconstruct a 3-D cone of information in cartesian coordinates. From the 3-D matrices stored in the computer, 2-D echocardiographic images in any plane at specified times in the cardiac cycle, or throughout the entire cardiac cycle, can be derived and visualized. A computer workstation-based system was developed to create full 3-D perspective projections of the echocardiographic data based on a technique called ray tracing, and adapted for use in visualizing 3-D scalar fields. The 3-D images obtained in normal volunteers demonstrated that our system permits an accurate reconstruction of the heart with the same spatial and temporal resolution as the original 2-D echocardiograms without cumbersome external reference systems.
Riccardo Pini, Elisabetta Monnini, Leonardo Masotti, Kevin L. Novins, Donald P. Greenberg, Barbara Greppi, Marino Cerofolini, Richard B. Devereux

Object Manipulation and Interaction


Surface Modeling With Medical Imagery

3D biomedical data obtained through tomography provide exceptional views of the interior structure of biological material. While visualization is one of the primary purposes for obtaining these data, other more quantitative and analytic uses are possible. These include modeling of tissue properties and interrelationships, simulation of physical processes, interactive surgical investigation, and analysis of dynamics. In large part these uses require the representation of tissue geometry, and this suggests the use of descriptive formalisms that make this geometric structure explicit. The descriptive formalism we are developing models tissue imaged from tomography as two-dimensional manifolds in three space. The uses demonstrated here include simple versions of surgical simulation, kinematic modeling, and kinematic analysis.
H. Harlyn Baker

Manipulation of Volume Data for Surgical Simulation

Many systems exist for the 3D visualisation of volume data. The manipulation (translation, rotation, Boolean operations) of such data is less well developed. In surgical simulation, the requirement is to have multiple object fragments that can be independently manipulated and merged. This paper discusses an approach using an interactively specified volume mask to dissect the data, and subsequent operations performed on Boolean expressions of masks and data.
Simon R. Arridge

Computer Assisted Medical Interventions

Existing imaging devices can be used to plan complex medical and surgical interventions. Recent advances in Robotics provide the opportunity of assisting the physician or the surgeon in performing the intervention. Assisting both planning and performing of interventions first raises problems of matching of various multimodality data. Then the performance of an intervention with a partially autonomous system puts specific problems which are discussed. A general methodology for Computer Assisted Medical Interventions is proposed, which turns out to be a particular case of the classical loop of Perception - Decision - Action. Stereotactic Neurosurgery is a domain in which this methodology is readily applicable. Results are discussed.
Stéphane Lavallée, Philippe Cinquin



Systems for Display of Three-Dimensional Medical Image Data

This paper briefly describes several research projects at UNC Chapel Hill involved with the display of 3D medical image data. A number of display methods is reviewed, and several new approaches outlined. Despite dramatic progress in display capability in recent years, even the most powerful current systems are inadequate for many daily medical imaging tasks. More powerful display systems in the near future may dramatically increase comprehensibility of complex 3D medical image data by enabling smooth interaction by direct manipulation, display without noticeable lag, and presentation of more powerful 3D depth cues such as head-motion parallax and stereopsis. These new systems will consist not only of a more powerful display engine, but also of an improved user interface with which the user can more easily interact than with current systems. One promising such interface for the future is based on a head-mounted display, one that allows the user to observe synthetic 3D structures superimposed on the world around him, allowing him to walk about these structures and modify them by direct manipulation using hand-held (simulated) tools. The poor quality of current miniature video displays prevents this approach from becoming a useful display method for 3D medical image data.
Henry Fuchs

A Software System for Interactive and Quantitative Analysis of Biomedical Images

A comprehensive software system called ANALYZE has been developed which permits detailed investigation and structured evaluation of 3-D and 4-D biomedical images. This software system can be used with any 3-D imaging modality, including x-ray computed tomography, radionuclide emission tomography, ultrasound tomography, magnetic resonance imaging, and both light and electron microscopy. The system is a synergistic integration of fully interactive modules for direct display, manipulation and measurement of multidimensional image data. Several original algorithms have been developed to optimize the tradeoffs between image display efficiency and quality. One of these is a versatile, interactive volume rendering algorithm. The inclusion of a variety of semiautomatic segmentation, quantitative mensuration, and process design tools (macros) significantly extends the usefulness of the software. It can be used as a “visualization workshop” to prototype custom applications. ANALYZE runs on standard UNIX computers without special-purpose hardware, which has facilitated its implementation on a variety of popular workstations, in both standalone and distributed network configurations.
Richard A. Robb

CARVUPP: Computer Assisted Radiological Visualisation Using Parallel Processing

A parallelised volume rendering algorithm has been constructed and implemented on a network of INMOS´ transputers. The technique offers great flexibility in terms of expansion to any number of processors available to the user and achieves a good speed-up factor. It is well suited to objects represented by a number of 2-dimensional slices and our main use for it has been Computerised Tomography (CT) data of human anatomical structures. At present using 16 T8 transputers we can, on average, render and display a volume of 6 MBytes of data in about 12 seconds with full shading. The original voxel densities are retained throughout the process so that different features within the volume of data can be identified, rotated in 3-d and displayed.
In this paper we describe the basic principles and performance of CARVUPP and its potential as a clinically usable tool. The system architecture and the parallelised algorithm we have developed will be outlined in detail.
Farzad E. Yazdy, Jon Tyrrell, Mark Riley, Norman Winterbottom



A Volume-Rendering Technique for Integrated Three-Dimensional display of MR and PET Data

Methods have recently been developed for using MR data to create 3-D computer models of the gyral anatomy of the brain. This paper describes image processing and computer graphic techniques for using PET images to produce 3-D models of brain surface metabolism. Then, an existing algorithm for retrospective image registration is used to fuse the two constructs into a unified 3-D model of brain structure and function. Neurosurgical planning can benefit from this display technique which localizes PET-detected metablolic abnormalities with respect to MR-defined gyral anatomy.
Xiaoping Hu, Kim K. Tan, David N. Levin, Charles A. Pelizzari, George T. Y. Chen

Computer Assisted Radiation Therapy Planning

The recent advances in 3D medical imaging are the basis to develop a new generation of computer assisted tools for radiotherapy planning. The precise extraction of therapy relevant information from 3D images, 3D display of anatomical structures, more accurate placement of beams and the calculation and display of dose disributions in three dimensions are expected to contribute to enhanced local control of tumours and to help to reduce side effects. The paper reports the current state of the art in radiotherapy planning and gives recommendations for future research.
W. Schlegel

CAS — a Navigation Support for Surgery

Computer Assisted Surgery (CAS) is a new navigation support for skull base surgeons. The combination of 3D coordinate measurement techniques, voxel processing methods and pseudo-3D image presentation supports preoperative planning of therapy, pathfinding during the operation itself and postoperative therapy control. For this purpose, the surgeon employs a hand-guided electro-mechanical 3D coordinate digitizer to locate points of interest within the operative field. The coordinates measured this way are correlated with a voxel model of the object gained by a preceding CT examination. With a prototype system the accuracy of this method has proven to be better than ±1 mm. The system has been successfully applied in more than 100 ENT operations and ten neurosurgical procedures. A similiar system was tested two times in the field of radiotherapy for the computer assisted placement of afterloading probes.
Ludwig Adams, Joachim M. Gilsbach, Werner Krybus, Dietrich Meyer-Ebrecht, Ralph Mösges, Georg Schlöndorff

Three-Dimensional Imaging: Clinical Applications in Orthopedics

Advances in three-dimensional imaging now provide the radiologist and referring physician with the ability to create accurate three-dimensional models of any part of the human body. The major area of clinical applications of three-dimensional imaging has been in orthopedics. This paper presents a review of the basic applications of 3D imaging in such areas as trauma, oncology, degenerative arthritis, and congenital disease. The techniques used for the generation of three-dimensional images is presented as well as a detailed description of some of the basic clinical applications of three-dimensional imaging. A preview of some of the more advanced and newer applications of three-dimensional imaging in regard to orthopedics is also reviewed.
Elliot K. Fishman, Derek R. Ney, Donna Magid

3D Morphometric and Morphologic Information Derived From Clinical Brain MR Images

Data from conventional clinical MR brain images were processed using multi-step computerized segmentation as well as 3D analysis and rendering techniques. The usefulness of so obtained morphometric information and morphologic display for the development of new concepts for diagnosis and follow up of diseases was demonstrated with data sets from patients with Alzheimer’s disease, normal pressure hydrocephalus, multiple sclerosis and brain tumors.
Ron Kikinis, Ferenc A. Jolesz, Guido Gerig, Tamas Sandor, Harvey E. Cline, William E. Lorensen, Michael Halle, Stephen A. Benton


Weitere Informationen