Views are expressed on desirable characteristics of image data compression schemes for PACS, in particular, that such schemes should not only enable reversible and irreversible compression but also provide user-friendly features and compatibility with PACS requirements. Among the image conversion schemes for data compression, hierarchical picture decomposition meets most of the desirable features and requirements, such as ease of composing a pictorial index of a patient file, adaptation of transmission to the variable resolution requirements in PACS, quasi-instant transmission of low-resolution pictures for teleradiology, and ease of implementing spatial pyramid filters. One of the simplest image conversion schemes which fulfills these requirements is the S-transform. Among the coding methods, the Lempel-Ziv scheme offers features which could be advantageous for PACS applications, such as that the code is always optimal - even though it is derived from a single-pass operation, that it may have fixed length, and that the code table does not need to be transmitted. Results of reversible and irreversible data compression with the S-transform, Lempel-Ziv coding, and quantizer functions are presented for computed radiographs and magnetic resonance images. The results are assessed and compared with S-transform/Huffman coding as well as differential pulse code modulation with Huffman or Lempel-Ziv coding.

The availability of ubiquitous wide-area channels of both modest cost and higher transmission rate than voice-grade lines promises to allow the expansion of electronic radiology services to a larger community. The band-widths of the new services becoming available from the Integrated Services Digital Network (ISDN) are typically limited to 128 Kb/s, almost two orders of magnitude lower than popular LANs can support. Using Discrete Cosine Transform (DCT) techniques, a compressed approximation to an image may be rapidly transmitted. However, intensity or resampling transformations of the reconstructed image may reveal otherwise invisible artifacts of the approximate encoding. A progressive transmission scheme reported in ISO Working Paper N800 offers an attractive solution to this problem by rapidly reconstructing an apparently undistorted image from the DCT coefficients and then subse-quently transmitting the error image corresponding to the difference between the original and the reconstructed images. This approach achieves an error-free transmission without sacrificing the perception of rapid image delivery. Furthermore, subsequent intensity and resampling manipulations can be carried out with confidence. DCT coefficient precision affects the amount of error information that must be transmitted and, hence the delivery speed of error-free images. This study calculates the overall information coding rate for six radiographic images as a function of DCT coefficient precision. The results demonstrate that a minimum occurs for each of the six images at an average coefficient precision of between 0.5 and 1.0 bits per pixel (b/p). Apparently undistorted versions of these six images can be transmitted with a coding rate of between 0.25 and 0.75 b/p while error-free versions can be transmitted with an overall coding rate between 4.5 and 6.5 b/p.

An imaging system to be used interactively in the operating room for stereotaxic neurosurgery is being developed at the University of Washington. System features include multiple windows displaying registered planar images from X-ray Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) scanners, database capabilities, and interactive image enhancement and processing capabilities. Patients are preoperatively scanned in CT and MRI using a compatible stereotaxic localization system mounted on a patient head ring. The images are downloaded onto magnetic tape and transferred into the imaging system. The current system is an IBM PC/AT personal computer with the XENIX operating system and a custom image processing subsystem designed at the University of Washington. This system is being upgraded to a SUN 4/260 engineering workstation with a TA AC-1 image processing and graphics accelerator. Image registration within the 3-dimensional volume is done automatically by using information from the image data header generated by the source modality and from the stereotaxic frames's fiducial markers in the images. After the images have been registered, a mouse-driven cursor is made active in one window and roamed on the active plane. The image in each of the other windows is updated in real time to show the slice at the intersection of the cursor as the active slice is roamed. The active window can be changed by the operator to allow roaming in the 3-dimensional stereotaxic volume. Once the target and entry points have been chosen, stereotaxic target coordinates and guidance parameters are generated and the needle track to the target is displayed. This paper presents our algorithms, hardware and software details of the system, and the preliminary clinical results.

A compression server in a PACS environment has to deal with images of different types and sizes. The images will flow into compression at different rates, ranging from 1-2 Mbytes/sec to 9.6 Kbytes/sec. Additionally, the pattern of the flow can vary. Some images will enter compression in one block. Other images, especially large images (i.e., digitized film images) will enter compression as a series of blocks. Interleaving of blocks can also occur. For example, the first two blocks from a digitized X-ray film may enter compression followed by a block from a CT image followed by more blocks from the X-ray film. In order to process incoming images rapidly, the compression service must compress blocks as they arrive. The compression of a CT image should not have to wait for the final arrival of a slowly transmitted large digital X-ray image. On the other hand, temporarily buffering large images or adding extra compression hardware may make the compression service too expensive. These PACS network considerations argue for compression that operates on images locally. That is, the compression algorithm should not have to know the statistics of the entire image to be effective. Any transforms should operate on local blocks within the image independent of the results on antecedent or subsequent blocks. In addition, since the network may present relatively small blocks to compression (as small as 64 Kbytes), the compression technique should not add a large amount of overhead to the compressed data in the form of tables, descriptors, and so forth. In this paper, the effects of breaking data into fragments was tested using a simulation tool dubbed PAW (Performance Analysis Workstation.) Two cases were considered. In the first case, large images were compressed in their entirety. In the second case, large images were broken into fragments and the fragments were compressed separately. The results show that processing the data in fragments is desirable.

In this paper, we present a technique which employs the mini-max error criterion for image compression using adaptive vector quantization. In vector quantization (VQ), the image vectors are usually coded using an "universal codebook" generated from a set of training images. The coding performance using this codebook is potentially poor for images outside the training set. A number of inter and intra-image techniques have been proposed to adapt the codewords to the input image. However, these techniques do not guarantee the closest codewords to be within a prespecified bound of the input vectors. This can result in large errors which give rise to artifacts. We propose an intra-image adaptive technique which employs a criteria that minimizes the maximum error. Here, the codebook is generated on the fly from the input vectors to be coded. A primary codebook of size, 8 or 16 is typically used to store the frequently used codewords. A larger secondary codebook is used to store the less frequently used codewords. Both the transmitter and receiver maintain identical codebooks and hence keep track of any changes without any overhead information. As it is a single-pass technique, real-time implementation is possible.

A new orthogonal transform coding scheme by Zernike polynomials for CT image archiving is discussed in this paper. By considering the image formation process, it can be shown that all of the meaningful information about object is contained in the set of lower order transform coefficients, which maximum order is determined by the measuring condition of projection data. Thus, the practical CT images can be exactly represented by a finite number of transform coefficients. In the proposed coding scheme, these coefficients are extracted and quantized according to the bit allocation rule which is prescribed by the statistical nature of CT images. Since both structual and statistical redundancies are reduced, an information lossless reversible data compression with a high compression ratio can be achieved. The experiments were carried out to confirm the coding characteristics.

In the last decade medical imaging has seen many changes. Various new and improved devices have been widely accepted allowing new noninvasive approaches and reduced radiation dosage for diagnostic imaging. The need then arises to interconnect different devices from a variety of manufacturers. These devices must also be able to communicate with disks, processing units and displays. For such Medical Picture Archiving and Communication Systems (PACS) to become feasible, a communication standard is required. Interfaces conforming to this standard should be provided by all manufacturers.

Image compression algorithms based on full-frame discrete transforms completely eliminate the artifacts generated by block quantization techniques. Previously, we have shown that full-frame compression using an adaptive two-dimensional bit-allocation table is capable of achieving compression ratios as high as 10:1 to 20:1, while maintaining excellent image fidelity, when applied to computed radiography images, and film radiographs digitized with 2K x 2K resolution. Cross-sectional images such as computed tomography and magnetic resonance have spectral properties qualitatively different from those of projection radiographs. We report results of a study on the efficiency of our algorithm applied to images from these digital modalities. We find that we can achieve excellent results with compression ratios in the range 3:1 to 7:1. For CT images of the head, previous attempts at using the full-frame technique had required a hybrid method that used separation of skull and soft tissue. This is unnecessary with our new two-dimensional bit-allocation table. The study was performed on a hardware implementation of our algorithm which can compress a 512 x 512 x 12-bit image in less than 1.5 seconds.

An ultrasonic imaging system has been integrated into an existing Picture Archiving and Communication System (PACS) module in the Pediatric Radiology Section at UCLA. A microcomputer (IBM-PC/XT) manages the acquisition and transfer of ultrasonic images through an Ethernet network to the host computer. The microcomputer is configured with: (1) a video frame grabber that can store four ultrasound (512x512) images, (2) a network interface board employing TCP/IP protocol, (3) a 32 megabyte winchester disk, and (4) a software system which coordinates acquisition, temporary storage, and PACS transfer.

Pixel subsampling is a simple and straight-forward method for digital image compression. Adjacent pixels in chest radiographs are highly correlated. Therefore, one may reduce the heavy burden of image data storage by subsampling pixels and reconstructing by interpolation whenever viewing is required. We performed a study on compression of chest radiographs using: (1) pixel subsampling followed by interpolation; and (2) a bit-allocation technique based on the full-frame discrete cosine transform (DCT). To obtain a compression ratio of 16:1, we subsample one pixel from each 4x4 pixel matrix. Bilinear and cubic spline interpolation are the two methods used in this study since they are simple and smooth interpolation methods. The more complex process of full-frame DCT yields optimal performance when the correlation of adjacent pixels are high [6]. By adjusting two quantization parameters in the full-frame DCT method, one can also achieve a 16:1 compression ratio. Our evaluation of pixel subsampling versus full-frame DCT uses: (1) statistical comparison of root mean-square errors (RMSE); (2) comparison of local structure fidelity; and (3) comparison of differences in subtraction images (original minus reconstructed).

Digital Video Fluorography (DVF) systems [1] provide digital radiography and digital fluoroscopy for the conventional radiographic and fluoroscopy (RR) rooms. During fluoroscopic examinations, DVF system's digital radiographic imaging ability replaces the conventional 100-mm photospot film camera and cassette spot film devices. The acceptance of the DVF system is highly dependent on its image quality. A DVF imaging system (S &S Inficon's FC1000) [2] is designed to produce high quality digital radiographic images. The image quality is dependent on the quality of the imaging chain including the x-ray source, image intensifier, video camera, digital system and hard copy device. This paper will only discuss the digital imaging portion of the imaging chain which include the video camera, the digital system and the hard copy device. The performance characteristics of each component of the digital imaging chain including the video camera, video digitization, image matrix size, image enhancement algorithms and hard copy device will be discussed.

We have built a versatile Direct Memory Access Interface for IBM PC-AT compatible personal computers which can accept digital image data from a variety of high density slow scan image capture devices. Our first implementation of this design is tailored to the Hamamatsu Photonics Cooled CCD Camera, which is designed to capture very low light level, high resolution images with 12 bit dynamic range. The interface has been designed for camera format flexibility and can easily be configured for larger image arrays and up to 16 bits of dynamic range. The interface loads digital image data to both the PC memory and the on-board RAM of most graphics controllers, allowing realtime image display/update for many non-standard scan rate image capture devices. The interface also provides arithmetic processing for realtime dark current correction and frame integration.

In this work, a new technique for variable rate VQ design based on tree structures is applied to medical images. It is an extension of an algorithm for optimal pruning in tree-structured classification and regression due to Breiman, Friedman, Olshen, and Stone [1]. The algorithm finds subtrees of a given tree-structured vector quantizer (TSVQ), each one optimal in that it has the lowest average distortion of all subtrees with the same or lesser average rate [2]. Since the resulting subtrees have variable height, natural variable rate coders result. Image reproduction at 1.5 bits per pixel is excellent and pathology in brain magnetic resonance images can be diagnosed in images at less than 0.5 bit per pixel. Finally, TSVQ is stored in a format convenient for progressive transmission of images.

Iterative user-interface development is an effective approach for developing a complex computer human interaction (CHI) system such as a radiology workstation. With an iterative design approach the toolsmith [Brooks 76] specifies a problem, produces a task analysis, develops a system design, and implements a prototype. This prototype is then evaluated using observations, verbal protocol, controlled subject experiments, and field studies, all with real users performing real tasks. The results of the evaluations are then used to modify and hopefully improves the problem specification, task analysis, system design, and implementation details.

Medical workstations of the future will support the real-time display and interactive manipulation of 3-D objects derived from CT, MRI and other imaging modalities. As part of such an integrated system for visualizing 3-D volumes, we have developed an highly interactive, flexible, and portable program for the 2D display of image slices and contours outlining anatomical objects. The editing of these contours as well as their automatic creation by thresholding and edge-tracking is supported. The contours may later be used to generate 3D surfaces for shaded-graphics rendering, or to mask out regions of interest in the image for volume rendering. This "image executive" program, or Imex, is designed to run in a windowing environment (i.e., the X Window System). The user-interface model, which may be described as a "Macintosh 1 for images", associates one movable and resizable window with each displayed view of a 2D image slice or of an indexed array of slices. Any number of views into one slice or into a subset of an array of slices may be present. Natural interaction is achieved by providing immediate response and by using the mouse to effect navigation and viewing functions, for example grabbing and dragging a slice onto another to copy its field-of-view or other attributes.

Usability testing with radiologists is an integral part of the development of a PACS workstation'. This paper describes a specific example of usability testing conducted to evaluate a prototype of an image organization and navigation system developed as part of the IRIS project2. The requirements for a computerized image organization and navigation system are specified; then the prototype navigation system is described. Finally, the usability testing method and results are presented.

Adaptive histogram equalization (ahe) is a method for adaptive contrast enhancement of digital images proposed by Pizer et al.. It has the properties that it is an automatic, reproducible method for the simultaneous viewing of contrast within a digital image with a large dynamic range. Recent experiments have shown that in specific cases, there is no significant difference in the ability of ahe and linear intensity windowing to display grey-scale contrast. More recently, Pizer et al. have proposed a variant of ahe which limits the allowed contrast enhancement of the image. This contrast-limited adaptive histogram equalization (clahe) produces images in which the noise content of an image is not excessively enhanced, but in which sufficient contrast is provided for the visualization of structures within the image. Images processed with clahe have a more natural appearance and facilitate the comparison of different areas of an image. However, the reduced contrast enhancement of clahe may hinder the ability of an observer to detect the presence of some significant grey-scale contrast. In this work, a psychophysical observer experiment was performed to determine if there is a significant difference in the ability of ahe and clahe to depict grey-scale contrast. Observers were presented with CT images of the chest processed with ahe and clahe into some of which subtle artificial lesions were introduced. The observers were asked to rate their confidence regarding the presence of the lesions; this rating-scale data was analyzed using Receiver Operating Characteristic curve techniques. These ROC curves were compared for significant differences in the observers' performances. In this study, no difference was found in the abilities of ahe and clahe to depict contrast information.

Receiver operating characteristic is used to compare the image quality of films obtained digitally using computed radiography (CR) and conventionally using analog film following fluoroscopic examination. Similar radiological views were obtained by both modalities. Twenty-four cases, some with a solitary noncalcified nodule and/or pneumothorax, were collected. Ten radiologists have been tested viewing analog and CR digital films separately. Final results indicate that there is no statistically significant difference in the ability to detect either a pneumothorax or a solitary noncalcified nodule when comparing CR digital film with conventional analog film. However, there is a trend that indicated the area under the ROC curves for detection of either a pneumothorax or solitary noncalcified nodule were greater for the analog film than for the digital film.

Despite the rapid progress made in the electronic design of imaging workstations for medicine, much less effort has gone into the design of environments in which such systems will be used. Based on studies of radiologist film reading sessions, considerable time will be spent working at such viewing systems. If the rooms in which the workstations are placed are not conducive to comfortable work, it will certainly not favor electronic viewing over film reading. In examining existing reading environments, it is also apparent that they are not optimum, even for film. Since some of the problems for film and electronic viewing overlap, such as heat generation (by the alternators, viewboxes, or workstation electronics) and glare from light sources, it should be possible to develop solutions which are applicable to both environments or to rooms which will feature both viewing systems. This paper will discuss some of the approaches to designing environments in which viewing of images is supported by the room architecture and engineering and not degraded by it. To illustrate these points, a design based on the constraint of a real room size and available architectural materials will be developed.

A multimodality picture archiving and communication system (PACS) is in routine clinical use in the UCLA Radiology Department. Several types workstations are currently implemented for this PACS. Among them, the Apple Macintosh II personal computer was recently chosen to serve as a desktop workstation for display and analysis of radiological images. This personal computer was selected mainly because of its extremely friendly user-interface, its popularity among the academic and medical community and its low cost. In comparison to other microcomputer-based systems the Macintosh II offers the following advantages: the extreme standardization of its user interface, file system and networking, and the availability of a very large variety of commercial software packages. In the current configuration the Macintosh II operates as a stand-alone workstation where images are imported from a centralized PACS server through an Ethernet network using a standard TCP-IP protocol, and stored locally on magnetic disk. The use of high resolution screens (1024x768 pixels x 8bits) offer sufficient performance for image display and analysis. We focused our project on the design and implementation of a variety of image analysis algorithms ranging from automated structure and edge detection to sophisticated dynamic analysis of sequential images. Specific analysis programs were developed for ultrasound images, digitized angiograms, MRI and CT tomographic images and scintigraphic images.

Multimedia radiological reports tie together radiographic images and operations on these images with text, voice or graphical annotation. The result is an animated presentation of the radiological examination report. A temporal representation for a multimedia radiological report is introduced which describes the temporal behaviour of the animated report. The temporal behaviour specifies the synchronization requirements for the coordinated presentation of the multimedia report components. The temporal behaviour is encoded in a temporal semantic network which is a network of temporally ordered nodes, where each node represents some primitive behaviour. Automatic creation of the multimedia radiological report consists of the capture, interpretation and storage of the temporal semantics of the radiologist's reporting session in terms of a set of primitive behaviours. Automatic playback of the multimedia radiological report is the orderly traversal of the temporal semantic network, invoking the behaviour associated with each of the nodes in the temporal order described by the interconnecting arcs.

We have compared the volume-based and the surface-based algorithms for 3D rendering of the human skull as applied to planning of craniofacial surgery. Both methods were applied to a selected sample of 12 CT studies exhibiting a variety of clinical problems. The number of slices ranged from 24 to 72 and the slice thickness from 1.5 to 6.2 mm. In general, volume renderings capture the anatomy in greater detail but are more prone to interpolation and thin bone artifacts. The surface based and the volume based techniques have complimentary roles in medical imaging. Surface extraction is needed in applications involving pattern recognition and fabrication of prostheses. Volumetric rendering excel in imaging multiple tissue and in handling diffuse edges.

Currently, optimal images of human brain anatomy are obtained using magnetic resonance imaging (MRI), while optimal images of brain function are obtained using positron emission tomography (PET). This paper describes a method to combine and display the complementary information in these two modalities. High-resolution MR and PET images of the human brain were obtained and correlated to form three-dimensional volumes of structural and functional image data. Slices from each volume were combined to form 2-D images of metabolism "mapped" over anatomic structures. The volumes of data were then rendered using a volume imaging technique. The results of the procedure are used to visualize the three-dimensional relationship between brain function and brain anatomy.

The problem of reconstructing a polyhedral surface from cross-sectional data admits of an exact and automatic solution when the surface is densely sampled as a voxel surface. We regard such a surface as a discrete sampling of a smooth manifold in general position. Using the full adjacency graph of the surface provided by a voxel model, we are able to classify the small set of possible topological changes in the sections of the surface; we then deal with these cases exhaustively, and thereby guide the triangulation process. In this way we are able to carry out triangulation without human interaction. When the sections are sparsely sampled, heuristic methods must suffice for triangulation. These methods work well on segments of "generalized cylinder," i.e., runs of sections containing single loops, but they often fail when attempting to process highly convoluted surfaces. This is because the topology of the sections changes when a critical point of the surface is encountered. We outline a heuristic algorithm based upon the same theoretical considerations as the full (dense-sampling) algorithm. This second algorithm follows a parsing step similar to that of the full algorithm, but locates critical nodes which are saddle points by a least-squares procedure, and uses this fit to supply the information required to complete the triangulation. This method provides a solution which is comparable in quality to the correct solution derived from the first algorithm.

A workstation model using miniaturized image display as a pictorial directory has been developed using a PIXAR Imaging Computer. The prototype was developed as part of an investigation into the appropriate user interface for a multi-modality PACS workstation. We have found that such a display allows rapid image review in order to determine the clinically significant images for further study. The details of the prototype are presented.

Although new imaging modalities routinely use digital imaging processing and display, digitization of conventional radiographic procedures has been slower to gain acceptance. Digital radiology potentially offers greater flexibility in image manipulation and display. It also has potential for high density storage of images and both local and long distance transmission of image data. Because of smaller film size and the potential reduction in the number of repeat examinations performed due to exposure errors because of the larger dynamic range of the imaging plate, reduced examinations costs could ensue. Of particular importance in our area, where we are dealing with a pediatric population, is the potential decrease in radiation dosage associated with computed radiology (CR) systems.

Medical innovations require rigorous testing on a number of levels to determine their applicability to clinical practice. This presentation describes our experiments with phosphor-plate digital radiography to determine whether this technology can replace conventional radiography for selected applications of urologic and chest radiology. While our work predominantly focuses on technical efficacy, there are implications for possible improvements in patient outcomes and cost-effectiveness.

This experiment is the first in a series designed to determine whether radiologists will be able to work as effectively when viewing quickly alternating single images on a single display screen as when viewing two or more images simultaneously from two or more screens. The experiment was conducted with non-radiologists as subjects and non-radiological materials. However, the experimental task and materials were designed to require cognitive and perceptual processing representative of the radiologist's task. Performance and satisfaction of users who were attempting to detect the difference in two similar images were measured. Decision time was faster for the simultaneous presentation mode than for the two sequential presentations and was also more preferred. Implications of these results for the design of radiologists' workstations and for further research are discussed.

An evaluation facility has been developed, which permits precise physical evaluation of CRT's. It is designed for scanning single element detectors such as photomultipliers or silicon diodes across the active area of the CRT up to dimensions of 15" horizontally by 20" vertically in steps of 2 micrometers. Other detectors include video camera tubes and CCD's. The facility is controlled by an IBM AT compatible PC and includes an analog integrator, an 8000 channel multichannel analyzer and a Biomation 2100 waveform recorder. It is designed for precise measurements of o linearity (signal transfer curve) o contrast resolution o signal-to-noise ratio o spatial resolution (beam spot size) o granularity o phosphor decay o stability o uniformity o jitter of scan lines o line pairing The facility was used for the evaluation of the Toshiba workstation TDF 500, which uses 2 CRT's of 345 mm vertical and 230 mm horizontal dimensions each. The paper describes the facility and discusses some results of the evaluation. It also compares the merits of multi element detectors (i.e. video camera tubes) with the merits of single element detectors (i.e. photomultipliers) in the evaluation of CRT's.

Three-dimensional (3-D) display and volume determination of soft tissue anatomy can be clinically useful in facilitating a physician's diagnosis, treatment, and evaluation. However current commercially available techniques for soft tissue segmentation and display have had limited success in accurately identifying soft tissue structures from two-dimensional (2-D) images. Simple global thresholding fails for a variety of reasons, requiring manual segmentation of soft tissue structures. Our objective is a robust segmentation method that eliminates the need for hand-outlining and yields reproducible results. As a step toward that objective, we have developed a software package that facilitates the use of existing segmentation methods (thresholding and hand-outlining) to segment soft tissue. Tools in this package allow the user to attain a greater success rate using thresholding, thereby decreasing the amount of hand-outlining that is required. The software package attempts to enhance reproducibility by minimizing user interaction.

A technique for reformatting multislice magnetic resonance imaging (MRI) data into arbitrary oblique planes has been developed and implemented on a DEC uVax system and allows flexible user interaction for the definition of the desired plane. The user identifies with a cursor features of interest on 2 or 3 different image levels to be combined in the resulting oblique image. The reformatted image can have arbitrary angulation and is created with a pixel dimension equivalent to the original data set. Resolution ranges from the original in-plane resolution to the slice thickness depending on angulation. This method has been applied to the particular problem of improving the presentation of coronary arteries on a conventional set of multislice spin echo cardiac images. In a study including 16 patients it was found that the visible length of coronary segments was increased on average by a factor of 2.

In previous work3-4 two algorithms were developed called Marching Cubes and Dividing Cubes for the three-dimensional display of objects contained in CT and MRI images. A key to these algorithms is that surface normals are derived from the normalized gradient of the original tomographic images. The resulting images have been subjectively judged superior to images generated with the Cuberille and Ray-casting algorithms. This paper shows how the Cuberille and Ray-casting algorithms can be extended to use the normalized gradient. The image quality attainable using the four existing and the two extended algorithms will be demonstrated and the implementation trade-offs for the algorithms will be discussed. We will also show how the display methods can be extended using recursive applications of custom-designed region-growing algorithms. Results will be shown that demonstrate how the modified algorithms have been used to disarticulate bones and to isolate soft tissue structures in CT and MRI.

The AT&T CommView® image management and communication system (IMACS) at Georgetown University supports multiple work stations for diagnosis, review, and research. Consistancy of the perceived brightness of displayed images throughout the network in the hospital is a critical issue. A simple method has been developed to set and track the brightness of cathod ray tuned used to display radiological images. A computer generated test image (SMPTE) is used. This consists of 11 brightness patches equally spaced over the dynamic range of CRT intensity driving levels. This test image is displayed and viewed in the same manner as patient images. An inexpensive hand-held photometer calibrated in ft-Lamberts is used to measure the brightness patches of the test images.

For certain clinical applications, object manipulation such as cutting or slicing an anatomical structure will add new useful information for physicians (especially surgeons). Performing an object manipulation from a set of axial images is not practical because it is time consuming, difficult to achieve a cut based on the structure of the object, and difficult to keep the slice-to-slice consistency of the cutting. To overcome these difficulties, we present a scheme to cut and slice a three dimensional object from its three dimensional image. The proposed scheme is easy to use and is very flexible so that the surface along which the object is cut can be selected to adapt to the complex shape of the three dimensional object. The cutting surface is not restricted to a planar surface and is defined by drawing curves on a three dimensional view of the object. The user interaction part of the cutting scheme will be presented together with the algorithm to form the cutting surface. A brief discussion of the potential clinical applications is presented to illustrate the usefulness of this nonplanar cutting procedure.

An important component of a radiologist's workload is the "reading" of radiographs and the accompanying reporting process which normally involves dictation and verification of the subsequent transcription. A subset of the patient radiographic file is chosen by the radiologist and arranged on viewboxes (typically a set of 8 viewboxes arranged in two rows of 4). The radiologist reads this subset of films and reports his findings in the context of the accompanying requisition statement. One can distinguish two distinct preliminary phases in the "analog" reading process. The first phase deals with the film selection and placement on the viewboxes (image organization), and the second with the actual film reading (involving image navigation, i.e. the means for accessing the individual images). Our multimedia radiological workstation which will replace the viewboxes by a low-resolution Control Monitor, and a single high-resolution Image Monitor for viewing the radiographic images. The set of radiographic images of the patient are organized by body area and examination view, and are associated with a sketch of a human body on the Control Monitor. This "anthropomorphic" organization of the images helps the radiologist create a mental picture of the entire patient radiographic file. We expect that together with the requisition information, a proper representation of the file will improve the search for relevant radiographic images. The workstation supports image navigation by means of a "soft" representation of the physical viewboxes. In the same manner that radiographic films can be placed onto physical viewboxes, image labels can be dragged with the mouse onto soft viewboxes embedded on the Image Monitor. Evaluation of results from tests conducted with 15 physicians are presented.

The authors report on observations of, and interviews with, physicians using a prototype digital image display and reporting station. While the users generally agree that image quality is clinically satisfactory they are unanimous in their opinion that improvements in the man-machine interface are required before case review by this mechanism is clinically acceptable in a production environment. A model image and information user interface is presented. It was developed in answer to the needs of radiologists and referring physicians operating in the imaging department of a community acute care facility. In such an environment images and related information must be communicated quickly and often simultaneously to different parts of the department and hospital. The user interface to the management system and the management system itself must address the varied functions and the needs of both the medical and clerical staff that perform them. Image enhancement processes, for example, must be restricted to those which quickly provide significantly more perceivable diagnostic information. Little used processes that may occupy significant portions of the display and the console's computing power must be trimmed or eliminated.

Addition of a DMA controller to our compression module has enhanced the system throughput by a factore of two. Advantages of current design strategy which stresses full compliance with industry standards is weighted against its set backs. Potential performance enhancement based on a more customized approach is discussed. It is shown that the compression module naturally expands into a high performance display station with the addition of primitive display controller boards. By sharing resources such as a large frame buffer and DMA controller, an optimal cost/performance scenario can be achieved.

In april 1987 a project started as part of a 2-year joint study agreement between IBM and the Section of Medical Informatics of Leiden University with substantial support of BAZIS and the Leiden University Hospital Department of Diagnostic Radiology. IBM provided us with an RT PC, a 32-bit, RISC-based workstation 1 as host, connected to the powerful 5085 Graphics Processor System. Based on experiments and study, IBM replaced the external 5085 system by a special internal imaging board (Megapel), supporting 1024 x 1024 resolution and 256 shades of grey on its attached display. In this paper the present experimental set-up will be described and analysed and suggestions for the direction of further development will be discussed.

A system for interactive definition, automated extraction, and dynamic interactive display of three-dimensional anatomy has been developed and implemented on a low-cost PC-based image workstation. An iconic display is used for staging predefined image sequences through specified increments of tilt and rotation over a solid viewing angle. Use of a fast processor facilitates rapid extraction and rendering of the anatomy into predefined image views. These views are formatted into a display matrix in a large image memory for rapid interactive selection and display of arbitrary spatially adjacent images within the viewing angle, thereby providing motion parallax depth cueing for efficient and accurate perception of true three-dimensional shape, size, structure, and spatial interrelationships of the imaged anatomy. The visual effect is that of holding and rotating the anatomy in the hand.

We are investigating the technical feasibility of a novel acquisition scheme for peripheral angiography that consists of taking images from a gantry that continuously sweeps back-and-forth while the contrast is flowing. The clinical advantage of such a multiple-pass system is that images at each position are spread over a longer time interval, increasing the chance of imaging contrast filled arteries and thus decreasing the need for retakes. Image acquisition during rapid sweeping is technically feasible. The duration of x-ray pulses is short enough to reduce the extent of motion blurring to less than one pixel, using mA and kV parameters available on our angiography system. Contrast and mask image pair superpositioning is excellent, permitting DSA processing.

Generally, in case of dealing a large-scale and complicated systems which output engineering data, social data, and economical data, the data are often multivariate and change by real-time. Operators are, therefore, required to grasp its status macroscopically at first. And for dealing with such problems of human-interface, we need a method which mediate between function of a computer which have high level data processing capacity and human capacity which excels in intuition and qualitative judgement.

As part of the evaluation of the display requirements needed in PACS, an ROC curve study was performed in order to assess clinical performance and compare two different display workstations using the results from the original films as a reference. It was found that in both cases the ROC curves obtained from the displays were significantly worse than obtained from the original film. It is therefore suggested that considerable care needs to be paid to the performance of such displays for them to fulfil the requirements of a clinically viable PACS system.

In this paper, we propose that the phase modulation of reflective light from the grating encoded plate can be used to achieve pseudocolor images by studying the reflective property of grating encoded plate. This technique offers the advantages which the colors of output image vary with the incident angle of illuminating light real-time and it is convenient to be achieved than other methods.

This paper, by using the hierarchical ideas of Jackson method, presents an aproach to the design of interactive human computer graphic interface. It also points out some of the characteristics of good interactive human computer environments,describes the differences between the programming of interactive human computer interface and the traditional program design, as well as the new requirements of its being viewed as a new kind of software. It summarizes and puts up some guiding principles for the design of user-friendly interactive human computer environments and some considerations of human factors in interactive human computer systems. The aproach presented here has been used for the design of several interactive systems and proven efficient and practical.