An interface metaphor is proposed for developing a 3D cursor for volume rendering applications. The metaphor, Luminous Gel, is a material for moulding 3D cursor. We have successfully demonstrated that such a cursor's location and spatial relationships with volume- rendered 3D objects are readily perceivable. Editing tools made in Luminous Gel have also been developed, which are useful as effective positioning tools for functional operations and inspecting devices for features exploration and scene understanding by enabling immediate depth perception and object highlighting. The editing tools have also been further elaborated for performing planar and curved sectioning directly and naturally.

A computer software package named LEGO was designed and implemented to enable medical personnel to explore and manipulate laser scanned 3D geometry obtained from a Cyberware 4020PS scanner. This type of scanner reconstructs a real world object into a mathematical computer model by collecting thousands of depth measurement using a low powered laser. LEGO consists of a collection of tools that can be interactively combined to accomplish complex tasks. Tools fall into five major categories: viewing, simple, quantitative, manipulative, and miscellaneous. This paper is based on a masters thesis obtained from the University of Illinois at Chicago.

3D visualization algorithms such as Surface Tracking, Marching Cubes and ray causing can reconstruct 3D images from volume data such as CT and MRI and have been used for medical applications. In this paper we present some of our work to further improve these popular algorithms so that the resultant algorithms are more efficient.

This research is in support of the development of an image processing system which is capable of detecting and tracking blood vessels in photographs or video images of the human microcirculation system. We describe a model which replicates the illumination processes contributing to a film or video image of the microvessels of the human bulbar conjunctiva. The model provides a foundation for microvessel detection algorithms, for measurement of vessel parameters, for determining relative depth of blood vessels, and for separating neighboring vessels in complex images. The model is based on a cylindrical vessel embedded in a diffuse medium which is on a reflecting background. A light source illuminating the scene is reflected by it's components and passes through a pinhole to an image plane, which records these reflections as intensity values at discrete pixel locations. Fundamental physical principles which include Lambert's cosine law, isotropic spreading, Fresnel's law and Beer's law are systematically applied to the model. A video apparatus and a phantom were constructed to analyze different illumination conditions and to verify the model. A simulation based on the model compared favorably with data taken from phantom images.

3DVIEWNIX, is a data-, machine-, and application-independent software system, developed and maintained on an ongoing basis by the Medical Image Processing Group. It is aimed at serving the needs of biomedical visualization researchers as well as biomedical end users. 3DVIEWNIX is not designed around a fixed methodology or a set of methods packaged in a fixed fashion for a fixed application. Instead, we have identified and incorporated in 3DVIEWNIX a set of basic imaging transforms that are required in most visualization, manipulation, and analysis methods. The result is a powerful exploratory environment that provides not only the commonly used standard tools but also an immense variety of others. In addition to visualization, it incorporates a variety of multidimensional structure manipulation and analysis methods. We have tried to make its design as much as possible image- dimensionality-independent to make it just as convenient to process 2D and 3D data as it is to process 4D data. It is based on UNIX, C, X-Window and our own multidimensional generalization of the 2D ACR-NEMA standards for image data representation.

We describe a refined method for estimating the 3-D geometry of cerebral structures of a patient's brain from magnetic resonance (MR) images by adapting a 3-D atlas to the images. The 3-D atlas represents the figures of anatomical subdivisions of deep cerebral structures as series of contours reconstructed from a stereotactic printed atlas. The method correlates corresponding points and curve segments that are recognizable in both the atlas and the image, by elastically deforming the atlas two-dimensionally, while maintaining the point-to-point and contour-to-contour correspondence, until equilibrium is reached. We have used the method experimentally for a patient with Parkinson's disease, and successfully estimated the substructures of the thalamus to be treated.

We have developed a technique called Interactive, Image-Guided Surgery (IIG) which tracks the position of a surgical instrument during surgery and displays that position on preoperatively-obtained tomograms. Recently we have incorporated real-time intra-operative video images: endoscopes and ultrasound into the data stream. The endoscope or ultrasound transducer's position may be tracked during surgery, allowing guidance and the correlation of intraoperatively perceived structures with preoperatively obtained tomograms.

We describe a neurosurgical planning support system that we developed on a workstation. It enables a surgeon to generate useful 3-D images interactively and to simulate surgery on the computed images pre-surgically to confirm the optimum parameters for the instruments used in stereotactic surgery. In particular, we introduce two techniques for implementing indispensable functions in systems like this. One is an algorithm that can detect boundaries in medical images, and the other is an algorithm that can reduce the volume of polygonal data for surface-rendered images. These techniques move our system a step toward clinical use.

A new system has been designed and built to validate the concept of 3D Computerized Angiography (3D CA). It addresses all the difficulties met from the acquisition of raw data on a patient to the display of the reconstructed volume. The main characteristics of the system and its operating modes are described. Special attention is paid to data processing aspects. The first in vivo results obtained with this system are presented.

We have constructed the 3D liver model consisting of numerous cubic compartments from slices of structural data measured using x-ray CT and simulated processes of liver reproductions after hepatic lobectomies, which are operations to remove intrahepatic tumors, considering structural and functional characters of livers. We took account of some factors affecting reproductions; 3D distributions of intrahepatic functional conditions, accelerations of reproducing activities around the excised surfaces, and functional compensations between lobes. The effect of these factors in each compartment were calculated from structural data according to the relative position of the compartment in the model. Compartments were duplicated in proportion to the parameters. Reproducing processes of liver tissues and changes of the appearance were simulated. The same excision as a certain clinical case was assigned to the model. Simulations were passed on changing values of parameters to modify the model accordingly. The result was compared with the clinical case.

Consistent image presentation is crucial to the diagnostic and commercial success of medical imaging networks. The issues and factors involved with insuring the integrity of the tonal presentation in a networked environment are discussed, and a solution is proposed incorporating the ACR-NEMA 3.0 data structure. An example of a representative system configuration is given.

A Quality Control (QC) test suitable for routine daily use has been developed for video based on-line portal imaging systems. It is designed to test for acceptable performance in high contrast spatial resolution. A phantom consisting of five sets of high-contrast rectangular bars with spatial frequencies of 0.1, 0.2, 0.25, 0.4, and 0.75 lp/mm is used for the test. A critical frequency f_c for which the square wave MTF is 50% is used as a 'go/no go' decision level for passing the test. Data obtained during a one month period were used to determine f_c. The test is shown to provide an automatic and objective measure of the system's imaging performance and is a useful QC tool for routine clinical work.

The computed radiography images of 100 randomly selected traumatic cervical spine series were evaluated. The studies were reviewed on the laser printed hardcopy and 2K monitor soft copy images. In addition to the cervical vertebrae, the cervico-thoracic vertebral body interface must be recognized for a lateral c-spine image to be acceptable. The level of visualization of the spine was on average, 1/2 vertebral body better on the monitor than the hardcopy image. In 8% of cases, this improve visualization allowed clearance of the lateral cervical spine thereby expediting patient care in this critical area. This presentation will cover the quality of images and techniques to improve the success rate for clearing the cervical spine.

In this study, we used a DuPont Linx CR system to generate film images of the AP & lateral views of a foot phantom. Images were printed using a 2 on 1 CR image output format. Image display parameters for one of these two images was varied in a systematic fashion while the second image remained at the constant default settings recommended by the manufacturer. Twelve radiologists evaluated the relative image quality of each pair of images for four anatomic areas: osseous detail in the fore foot, mid foot and hind foot, and soft tissue detail. The resultant data were analyzed to investigate the influence of the display parameters on radiologist preference. The results of this study demonstrate that significant improvements are achievable for the CR display parameters when compared to the default values. This improvement in radiologist preference depends on the anatomical regions under consideration.

Digital storage and display of mammograms could significantly improve operations in breast cancer screening programs. If mammographic films are scanned with high performance instruments, enhanced display and/or computer-aided diagnosis may improve diagnostic performance. We are studying observer performance with digital mammograms printed with 1.9X magnification, skin line equalization, and background masking. The purpose of the study is to establish that mammograms can be digitized with no loss in quality and displayed so as to improve observer performance. Performance measures of interest include diagnostic accuracy and interpretation time.

The methodology reported here enables us to mathematically model and quantify the motion of each component bone, relative motion of bones, the contact surface of bones and their change during motion for complex joints, from a time sequence of MR image volumes acquired in vivo. Additionally, since we model the bone surfaces, we are able to display in vivo joint motion. Through a variety of new rendering techniques we are able to create realistic displays of bones from MR images and to combine these displays with the motion parameters. This approach opens an entirely new avenue for understanding true 3D joint kinetics and for the potential utilization of kinematic information for surgical planning.

Reconstructing a 3D structure from a set of its 2D X-ray projections requires that the source position and image plane orientation in 3D space be obtained with high accuracy for each of the imaging chain positions. This knowledge is generally obtained through a geometrical calibration of the data acquisition system. In this paper, we present a fully automatic method for such a geometrical calibration, well suited to a 3D X-ray imaging system which acquires sets of 2D projections during a rotation of the imaging chain. This method is based on (1) the use of a dedicated calibration phantom with reference points, (2) an automatic algorithm to detect and identify the reference points on the phantom's 2D X-Ray projections, and (3) the estimation of the imaging chain geometrical parameters by minimizing the reprojection mean quadratic error measured on these reference points. Results obtained both from simulation data and from data acquired on an experimental bench are presented.

Fabricated anatomic models are used in presurgical planning and in the creation of custom cranial prostheses. To more carefully examine the differences between the original bone geometry and the fabricated model, we created a synthetic digital image phantom composed of hemi-spherical and hemi- parallelpiped solids rendered into a multi-image CT volume dataset. The computer program that created the phantom allows the dataset to contain an unlimited number of these solids either isolated or overlapping. Because of the image element's simple geometry, we can calculate the theoretical volume and measure the rasterized (sampled) volume during 3-D image synthesis. Subsequent reconstruction and measurement provides a means to determine the 3- D reconstruction volume, surface area, and linear measurement error. The phantom dataset (28 slices of a three hemispheres and one hemi-parallelpiped with .5 X .5 X 1.5 millimeter voxels) yielded 212, 342 surface triangles when subjected to 3-D segmentation and surface reconstruction. A rapid prototyping model was created and measurements taken to compare the accuracy and precision of the model process. These measurements are compared with an actual patient skull bone model and a scan of a dried mandible to obtain a partitioning of the error. All tests showed an accuracy of less than .55 millimeter (< 1%).

A new scatter rejection technique using the modified air gap method is presented. Two x-ray cassettes are used per exposure. One cassette (front cassette) was placed immediately behind the object. The other (back cassette) was positioned 60 cm behind the object. The front radiograph has good spatial resolution but includes a relatively large amount of scatter. The back radiograph has poor high frequency resolution but little scatter due to the air gap. The spectral composition technique extracts high frequency information from the front radiograph and combines it with the low frequency information of the back radiograph to produce an improved composite synthesized image. No artifacts due to the spectral composition technique were observed. In the lung and heart regions, the modified air gap method resulted in comparable or superior scatter rejection performance as compared to the conventional grid technique.

The advent of Photostimulable Phosphor Computed Radiograph (PPCR) makes it now possible to acquire X-ray image data in digital format in the clinical department. This has led to alternative ways of viewing radiographic images, including softcopy review from a CRT device. However due to differing response characteristics of hardcopy film and display monitors it was necessary to investigate these systems and develop a technique that could allow radiographs to be optimally displayed on a variety of differing devices.

The appearance and diagnostic quality of images from Fuji-based computed radiography equipment is dependent upon a series of steps involving both procedural criteria and machine parameters. These steps can be divided into the three general categories of image acquisition, image digitization, and image display. The implementation of an effective quality control program for computed radiography requires an understanding of the interdependence of these three stages of image production, and the development of methods to assess both operator and machine deviations from required performance. Our experience thus far with the implementation of CR in a large scale PACS at 3 medical centers underscores the need for dedicated applications training support, and for a systematic approach to parameter adjustment.

We examine the requirements for a high quality image recording medium and present a new digital hardcopy system that is very suitable for the emerging new medical digital technologies. Imaging is performed using high power solid state lasers, and images are realized without chemical processing. The imaging mechanism of the new system is highly deterministic, and produces images of excellent quality. Novel imaging algorithms are employed to generate 8 to 10 bits of discernible shades of gray. This system has a high dynamic range of optical densities ranging between 0.01 and 3.5, and a MTF that exceeds most analog systems.

A mathematically defined standard display function is proposed for black-and-white hard and soft-copy image displays based on perceptual linearization. Since perceptual linearization is related to threshold contrast of targets with specific parameters, perceptually linearized display functions are calculated for a variety of target descriptors to illustrate the range of linearization by the proposed standard. Rogers and Carel's formula and the two-dimensional models of visual system sensitivity of Barten and Daly are used to calculate such display functions. The proposed standard provides perceptual linearization for the peak contrast sensitivity near 4 c/deg. The standard is adaptable to black-and-white display systems independent of maximum luminance and luminance range. Perceptual linearization is maintained largely independent of object size, external noise, observer performance variations, low supra-threshold contrast, target orientation, and ambient light. The standard facilitates similarity between the soft and hard-copy of an image independent of luminance. When using the standard function for displaying images, a minimum number of bits for the D/A converter is required that provides uniform digitization resolution over the entire display range. The standard display function, however, is not a visualization standard for images of specific applications.

This paper is an attempt to correlate CRT noise (measured physically) with contrast sensitivity obtained from the detection of square-wave patterns presented to human observers on a CRT. It presents the results of both physical and psychophysical evaluations of the noise of high resolution CRTs at a luminescence level of 65 ft-L. Temporal as well as spatial noise was physically measured with the aid of a photomultiplier-based evaluation system. Human contrast sensitivity was determined physchophysically using square-wave patterns of frequencies ranging from 0.47 lp/deg to 14.9 lp/deg. Even though the set of data is not very extensive, the results indicate a correlation between physical and psychophysical evaluation: for the luminance levels under consideration, the detection of human observers seems to be limited by the spatial noise (phosphor granularity); the temporal noise plays a role only at low luminance levels.

Models for the prediction error distribution in losslessly encoded grayscale images are explored. The prediction error distribution results from the use of linear predictors but the techniques used in the paper may also be applied to distributions which arise from the use of other methods to encode losslessly digital images. A compact method for representing the prediction error distribution for 12-bit greyscale images is used and the trade-off between space required for a distribution and the use of multiple distributions is investigated. Models considered include zeroth and first order Markov models. The variation of the prediction error distribution over the image is considered and shown to be important in achieving better compression. Choosing the predictor formula adaptively is also investigated.

In Vector Quantization, the codebook generation requires a clustering scheme, where the number of the clusters (the codebook size) is pre-defined. There are several algorithms which resolve this problem. For example, we quote the well known Linde-Buzo-Gray (LBG) algorithm, and the recent finite state algorithm. Most of them are iterative, and thus, time consuming. Our purpose is to look for a fast one, which is necessary non-iterative and represents adequately the image to be coded. After a review of existing algorithms for codebook generation, we propose a Modified Decision-Directed Clustering (MDDC) technique for codebook generation and its application in radiological image. The Convergence of the (MDDC) algorithm to a globally sub-optimal codebook in finite time is proved.

A method for quantifying the magnitude of block artifacts caused by JPEG type of image compression has been developed. The behavior of block artifacts as a function of compression ratio and edge enhancement operations can now be mathematically tracked. This mathematical technique is found to be useful for development of image compression techniques with regard to eliminating block artifacts.

The transmission of digitized medical images over the existing telecommunications infrastructure presents a formidable challenge. To achieve delivery times on the order of seconds--as opposed to minutes or hours--for large-format high-resolution images, data compression on the order of 25:1 is necessary. This degree of data compression cannot be reached with lossless techniques. This paper reports on an adaptation of a standardized technique for lossy image compression (the JPEG approach), which provides high compression ratios for radiographic images with minimal apparent loss of diagnostic quality.

In this paper, a lossy image compression algorithm based on a prediction and classification scheme is discussed. The algorithm decomposes an image into four subimages by subsampling pixels at even and odd row and column locations. Since the four subimages have strong correlations to one another, one of them is used in predicting all the others and the resulting differences between the predicted subimages and the original subimages are encoded. Estimated differences tend to be large in a region where pixel values change rapidly, while the differences are small in a monotonous region. This redundancy is explored by dividing the estimated differences into subsets based on the slope of pixel changes, the basis for which is found in some human perception models used to measure the visibility of distortion. The resulting classified estimated differences having different visibilities are encoded with classified vector quantizers.

In our previous studies we found that excessive spatial resolution can be partially recovered by error-free compression. However, the excessive gray dynamic resolution is not reducible as far as information content is concerned. Both information evaluation and preliminary clinical tests indicated that no information existed beyond 9-bit when digitized by a sample size of 180 (mu) with laser film digitization or computed radiography. In this study, we found that a high compression method can only achieve a small fraction of true compression efficacy over an error-free compression for a well-defined digital radiographic imaging system. This implies that the following two procedures contain similar digital information: (a) digitization of a 12- bit image and processing by a moderate irreversible compression (e.g., DCT type compression) and (b) digitization of the same image 8-bit followed by an error-free compression method (e.g., DPCM/arithmetic coding). Images processed by the above methods require about the same digital storage. The image quality of 12-bit with 5:1 irreversible compression is very close to that of 8-bit with 3:1 error-free compression. Higher compression efficiency (e.g., 0.5 bit/pixel) using procedure (a) would degrade the image quality particularly in edges and small structures. This is because the quantization procedure acts as a filter in the DCT compression. Without the interference of noise, the compression efficiency of using irreversible and reversible compression techniques are comparable as far as information is concerned.

An adaptive image coding scheme based on Discrete Cosine Transform (DCT) is considered. A set of 90 features in the spatial and spectral domain leads to a subset of features which is used to automatically classify subimages, taken from a multimodal medical image data base. The classifier, based on a binary decision tree, discriminates 13 classes. In the DCT domain, a normalization matrix for each class is generated using the features computed on subimages. This matrix allows to select the significant DCT coefficients associated to a class. This method leads to a performant adaptativity for the coding scheme. The classifier is very simple and cheap in computing time. A given subimage is classified, transformed with DCT, normalized by the matrix associated to its class, quantized and coded with Huffman tables.

We have developed a software module which performs 2-D and 3-D image compression based on discrete wavelet transform/subband coding techniques. The software allows the user to interactively determine the tradeoff between compression ratio and fidelity (by viewing the results) and to interactively define specific regions of the image--of any size and shape--that are to be preserved with full fidelity while the rest of the image is compressed (again, viewing the results). The compression achieved is superior to the JPEG standard algorithm. We present sample results on a variety of medical images and direct comparisons with JPEG results. We also show examples of the improvements gained by true 3-D compression of a 3-D image (as opposed to 2-D compression of each slice), discuss human visual system response issues, and describe extensions of the current approach to still more efficient compression schemes.

We evaluated a prototype free-space pointing device with a medical image display workstation. The target environment is the Intensive Care Unit (ICU), where there is very little counter space available, and image workstations are used intermittently and for short periods of time. Managers of the typical ICU do not want to dedicate space to PACS, but would rather mount the image monitors through the wall at eye level, so they can be viewed from the hallway. The hallway image viewing location allows use by a large number of people, as when making morning ward rounds or teaching rounds. Because many physicians are accustomed to graphical user interfaces and pointing devices, the transition to the free- space mouse is an easy and natural one. The use of a free-space mouse allows a very flexible interaction and intuitive graphical user interface, but does not require a horizontal surface, and is easily operated with one hand from the standing position.

We have developed a technique for visualizing medical image data sets that have multiple values at each physical location. These data sets are increasingly common as physicians attempt to correlate between modalities (for instance, CT and nuclear medicine, MR and PET, CT and MR) as well as within modalities (for instance, MR metabolic and anatomical scans). Our technique, termed 'isoluminance', is designed primarily for displaying sets that have two scalar values associated with each physical (x,y) location on a two dimensional scan. A perceptually uniform luminance scale is used to encode one dimension. At each step of the luminance scale a set of isoluminant hue steps are used to encode the other dimension. The hue scale is chosen to be perceptually uniform and as 'natural' as possible. The resulting data set can then be displayed as a single image on a color display. We have found observers using our technique are able to comprehend both of the data sets, to understand relationships between the data sets, and, when using interactive manipulation techniques, are able to select or label specific features of the data set.

The image viewing workstation is an all-important link in the PACS (Picture Archiving and Communications System) chain since it represents the interface between the system and the user. For PACS to function, the working environment and transfer of information to the user must be the same or better than the traditional film-based system. The important characteristics of a workstation from a clinical standpoint are acceptable image quality, rapid response time, a friendly user interface, and a well-integrated, highly-reliable, fault-tolerant system which provides the user ample functions to complete his tasks successfully. Since early 1992, the MDIS (Medical Diagnostic Imaging Support) system's diagnostic and clinical workstations have been installed at Madigan Army Medical Center. Various functionalities and performance characteristics of the MDIS workstations such as image display, response time, database, and ergonomics will be presented. User comments and early experience with the workstations as well as new functionality recommended for the future will be discussed.

Previous studies on the effectiveness of workstations using X window applications as PACS display and processing stations have revealed some difficulties owing to X's limited ability to support image processing. In response to studies of this kind in medicine as well as other imaging application domains, an extension to X windows has been proposed called the X window Imaging Extension (XIE). XIE provides mechanisms for efficient transport, storage, display, and processing of images on any X-capable hardware. This paper attempts to highlight important aspects of this extension as they apply to PACS and medical imaging. XIE's computational model is reviewed and methods of using it to support image review operations are discussed.

The increasing computational demands of medical imaging will exceed the capacity of standard microprocessors. For the most computationally intense problems, such as real-time scanning, parallel processing will be required. We evaluate the performance of a master-slave model of coarse-grained parallel processing on examples of reconstruction and postprocessing problems. We use a commercially available multicomputer system in configurations of from one through eight processors with distributed, shared memory. We examine a variety of 2D medical imaging problems ranging from pointwise operations, such as window-level, to global operations, such as 2D FFT. Parallel processing with the master-slave model is most efficient when data transfer among processors is minimized. This can be done by a combination of high-performance computer architecture and well-designed processing algorithms.

The Univ. of Washington's Image Computing Systems Lab. (ICSL) has been involved in research into the development of a series of PACS workstations since the middle 1980's. The most recent research, a joint UW-IBM project, attempted to create a diagnostic radiology workstation using an IBM RISC System 6000 (RS6000) computer workstation and the X-Window system. While the results are encouraging, there are inherent limitations in the workstation hardware which prevent it from providing an acceptable level of functionality for diagnostic radiology. Realizing the RS6000 workstation's limitations, a parallel effort was initiated to design a workstation, UWGSP6 (Univ. of Washington Graphics System Processor #6), that provides the required functionality. This paper documents the design of UWGSP6, which not only addresses the requirements for a diagnostic radiology workstation in terms of display resolution, response time, etc., but also includes the processing performance necessary to support key functions needed in the implementation of algorithms for computer-aided diagnosis. The paper includes a description of the workstation architecture, and specifically its image processing subsystem. Verification of the design through hardware simulation is then discussed, and finally, performance of selected algorithms based on detailed simulation is provided.

To obtain successful results in the therapeutic application of lasers, understanding laser beam distribution in lased tissue is essential. However there are few practical available techniques to observe the distribution of near-infrared lasers. Charge-coupled device (CCD) image sensors have an excellent sensitivity to near-infrared ranges, and so the authors have studied an imaging technique for near-infrared laser beam distribution in tissues using a video camera with a CCD image sensor. As the result of this experiment, it has been confirmed that the near-infrared laser beam distribution in tissues was easily captured as a visible pattern by a CCD camera on a CRT frame. Since the CCD camera can visualize the near-infrared laser beam distribution in tissues directly on the image of a subject, every part of the near-infrared laser beam distribution can be exactly corresponded to the subject. And so this technique is very useful for the clinical application of near-infrared lasers. The greatest advantage of this technique is the applicability to monitor the laser distribution in the tissue during an operation with near-infrared lasers. The results of this experiment suggested the possibility to lead to 'laser-intensitography', which was also experimented.

This paper describes the utilization of object oriented programming concepts to address the problem of medical imaging display and review application development. The outcome is a set of object class libraries implemented in C++, collectively called Perfect Vision.

Rigid requirements of perfect network transfer (i.e., without any loss of data) may be delaying the deployment of picture archiving and communication systems (PACS) and teleradiology. By using a clever and fast packet encoding mechanism to transmit images and then using a four- neighbor interpolation recovery scheme to 'fill in' lost pixels, some packet loss during network transmission may be affordable without affecting the diagnostic quality of the image or influencing the radiologist's diagnostic performance. To test this, radiologists viewed mammographic images with 0%, 15% and 25% transmission loss and reported on the presence or absence of microcalcifications. Observer performance was measured using receiver operating characteristic (ROC) techniques. Diagnostic performance in the 15% loss condition did not differ significantly from performance in the 0% loss condition. 25% transmission loss resulted in a decrease in performance. Thus, up to 15% loss can be tolerated without affecting diagnostic performance. The utilization of loss/performance curves may allow flexibility in network transmission performance requirements, which could ease PACS and teleradiology implementation using current technology.

This paper presents the data examples for clinical application of the authors' original thermographic techniques named developmental thermography (DT); (1) triple aspect thermography (TAT), (2) multiple aspect thermography (MAT) and (3) panoramic thermography (PT). Appropriate examples for the advantageous application for each technique of DT are selected and the specific advantages are discussed. The greatest advantage in the clinical application of DT is the capability of simultaneous display of the thermogram over even the entire aspect of a subject, and so DT has a lot of advantages in clinical application. Error free and wide coverage in DT are especially useful for the diagnosis of cancer. DT is expected to contribute to future advance of medical thermology.

Today's concern over rising medical costs demands that medical diagnostic systems have cost effective life cycles. The design of a Medical Image Capture, Formatting, and Display (MICFD) system must be based on an architectural approach utilizing new technologies for upgrades and modifications. The need to maintain a cost effective operation dictates flexible, easily upgradeable architectures. In the past, Image Capture, Formatting, and Display capabilities have been contained within a single medical diagnostic system and embodied an architectural approach that limited significant performance upgrades due to tight coupling between software and a specific vendor's hardware. Significant capability enhancements to these systems could in the past only be accomplished by replacement of the entire system. The MICFD system described in this paper was specifically tailored to meet the needs of a Medical Diagnostic Center to monitor and analyze a diagnostic procedure through the use of state-of- the-art image capture, formatting, and display technologies. Further, the architecture is such that incremental enhancements can be made to strengthen budget profiles. A review of the requirements for a MICFD system that will support multiple diagnostic systems and provide a method for minimizing life cycle cost is presented in this paper. To fulfill these requirements, SPARTAC (SPArta Real Time Analysis, Computation and Control Center) architecture and design concepts have been used.

The methods currently used at the University of Ottawa Heart Institute for storing, viewing and manipulating echocardiographic and cardiac angiographic examinations, are aimed at helping specialists in performing their diagnoses. It is cumbersome for attending cardiologists to review these studies and in particular, it is next to impossible for them to simultaneously view and compare studies, even between different views from the same study. This paper details our efforts in designing, constructing and evaluating a computer-based workstation to give the attending physician easy access to the recently acquired studies, and to allow them to do comparisons between studies, even of different modalities.

This paper introduces an interface facility (IF) developed within the overall framework of RACE research project. Due to the nature of the project which it has been focused in the Remote Medical Expert Consultation, the involvement of distances, the versatile user advocation and familiarity with newly introduced methods of medical diagnosis, considerable deficiencies can arise. The aim was to intelligently assist the user/physician by providing an ergonomic environment which would contain operational and functional deficiencies to the lowest possible levels. IF, energizes and activates system and application level commands and procedures along with the necessary exemplified and instructional help facilities, in order to concisely allow the user to interact with the system safely and easily at all levels.

The present paper deals with the development and evaluation of a new compression scheme, for angiocardiography images. This scheme provides considerable compression of the medical data file, through two different stages. The first stage obliterates the redundancy inside a single frame domain since the second stage obliterates the redundancy among the sequential frames. Within these stages the employed data compression ratio can be easily adjusted according to the needs of the angiocardiography applications, where still or moving (in slow or full motion) images are hauled. The developed scheme has been tailored on the real needs of the diagnosis oriented conferencing-teleworking processes, where Unified Image Viewing facilities are required.

Tests were performed to determine the color characteristics that would permit the information contained on labels placed on x-ray films to be captured by laser scanner film digitization and by ultraviolet light film copying. Tests of laser film digitization demonstrated that the colors allowing transmission of the He-Ne laser beam were red, orange, pink, yellow, and white. The ink colors absorbing the He-Ne laser beam were green, blue, and black. Labels on films should be placed in areas of the film that are clear or lightly exposed. Experiments performed to determine the label requirements for film copying demonstrated that white, clear and pink labels transmitted the ultraviolet light used in our film copier and thus proved to be good background colors for the paper. Black and green ink and black crayon were the most opaque colors and would therefore prove to be the best lettering colors. Because of the relatively low intensity of the ultraviolet light source, the paper stock used for these labels should be relatively thin and should be placed on areas of the film that are clear or lightly exposed.

It is possible to use storage phosphor radiography (SR) devices in a manner that results in excess exposure to the patient without the operators knowledge. Because these SR systems have an automatic correction for the final optical density (OD) of the image, the technologist and radiologist will not be able to use excessive blackness of the image as a sign of overexposure. Tests reported here demonstrate that it is possible to obtain images of a chest phantom that appear acceptable with a 32 times difference in exposure (maximal exposure .86 R). It is possible to obtain exposures of a pelvis phantom that appear acceptable up to the tube limit of our machine (4.8 R). Tests of the Fuji AC-1 demonstrate that it will accept a much wider range of exposures than the AGFA ADC prototype which permits only a 7 times difference in exposure before the image is degraded.

The Fuji AC-1 is a popular system for digital radiography with great potential, but for many the initial implementation of the system in a radiology service is frustrating experience for lack of effective support and a clear operational manual. Adjusting the image processing settings for site specific image optimization has been, for most users, difficult and confusing. The effect of each of the image processing parameters available on the Fuji AC-1 is not easily understood by users of the machine. The application manual that comes with the machine briefly defines the parameters and gives recommended settings for different types of examinations, but the manual does not provide the user with adequate information or a suggested method to allow the user to easily optimize the image for that user's site. This paper defines in more precise terms the meaning of the various parameters and suggests a method of procedure to optimize the parameters to each radiologists preference. The results are based on experiments and measurements we performed with a Fuji AC-1 +. This information should assist the radiologist to use a more systematic method of image quality improvement rather than the somewhat random trial and error process that current users have described.

An improved version of an efficient method for the reversible compression of digitized medical images is described. The modifications made in this version are aimed at reducing some of the limitations imposed by the original method. As in the original method, the improved method uses the well-known linear prediction technique to decorrelate a given image. A statistical source model with multiple contexts is employed to model the sequence of decorrelated image pixels. The selection of contexts for the source model is based on the horizontal and vertical components of the gradient in the given image as well as the predicted gray-level value of a pixel. The selection procedure is however entirely adaptive in the improved method, whereas it is only partially adaptive in the original method. The source model statistics are also calculated adaptively. The decorrelated image pixels are encoded using the appropriate contextual statistics with the arithmetic coding technique. Experiments on three groups of medical images show that the improved method achieves satisfactory compression performance.

Digitization of screen film mammography is a necessary first step in the application of computer and communication technologies to breast cancer screening. A Georgetown University Medical Center team working together with DBA Systems, Inc. has been testing a CCD camera based film digitizer with a focal spot size of 35 microns. It is anticipated that this high resolution digitizer will assist in the development of a clinically acceptable digital mammography system for the purpose of digital feature analysis, transmission and storage.