Electronic exchange of digital images among health care sites can be accomplished rapidly and economically. Data communications computers attached to existing municipal CATV cable systems can provide transfer of images to remote locations for viewing and archiving. A broadband image communication network has been constructed in the Lansing, Michigan area linking a central image display and interpretation center at Michigan State University to 3 Computed tomography (CT) scanners located in separate community hospitals up to eight miles away. Reconstructed compressed CT images are automatically transmitted from the hospitals to the central viewing site in a few seconds. An electronic mail facility provides for user communications among all network sites. Key to the economic and operational feasibility is the availability of leased channel bandwidth (3MHz total) supplied by two local CATV companies and a university cable service which are all interconnected. Channel characteristics and experience to date with such a complex cable network is reviewed. This demonstration of a metropolitan area image communication network opens new opportunities for expanded relationships and services among hospitals, outpatient clinics, and associated professional groups.

PACS (Picture Archiving and Communications) has become quite a topic lately, with many descriptions of the functionality needed and the cost of the hardware required to implement such a system. The software tasks involved are of enormous importance, and can be designed to provide flexibility and functionality that will reduce overhead costs and provide a variety of interactions for users, suited to their particular task or modality. This paper will attempt to define the software tools and structure needed to give the PACS console a wide variety of applications that can be tailored for a specific site or user. In addition, the modularity of the software goes along with modularity in the hardware to keep the cost down for different types of tasks.

The previously described computerized model (Cywinski & Gray) for calculating the cost of film based PACS in radiology departments has been extended to include projected costs of automated PACS. A survey of 17 diagnostic digital imaging facilities was performed. The survey results for each facility were entered into the model. The output of the model was in the form of a series of spreadsheets which illustrated cost comparisons between film and automated PACS in radiology departments. The analysis of cost comparisons indicated that presently available (M/NET) automated PACS are cost effective and actually save operating costs for the radiology facilities performing at least 10,000 diagnostic imaging procedures per year.

Reliability issues are of prime concern in the design of a digital archive. We investigated the current archiving system to determine how reliable it was and what impact this had clinically. Although the digital archive will significantly improve the reliability of archiving, subsystem failures may negatively impact function. Scenarios are presented to identify potential problems with and solutions to subsystem failures of the digital archive.

As part of a design study for a complete PACS system for a large hospital in England, we were tasked to develop and evaluate a number of potential system architectures. These architectures were to be considered with respect to an operational requirements document. The alternative architectures consisted of systems with distributed and centralized archives, centralized and distributed active storage, and video and digital communications capabilities. Several different architectures are presented along with a brief discussion of their individual merits and brief comparative analysis.

The ultimate purpose of any PACs system is to improve the state of health of individuals and the community. The way one attempts to achieve this goal in our current environment will determine the design and performance specifications of any PACs system. Within the hospital environment, one must determine the functional requirements by understanding that a hospital is intrinsically a place for physicians. It is the referring physician who admits patients, determines their treatment, when the treatment is to be changed, and when the patient is to be discharged. The referring physician is the Chief Executive Officer of patient care. It is a reasonable goal of a diagnostic imaging and PACs systems to furnish timely and accurate information to the Chief Executive Officer of patient care. The role of the referring physician i.e., the C.E.O., is becoming even more critical within the changing economic environment of American medicine. DRG legislation has lead to the greatest change in American medicine since the introduction of Medicare. Previously we have been under a cost reimbursement formula whereby the more one spent the more reimbursement one got. With this cost reimbursement formula, elegant and expensive systems could be justified and actually be profitable. This is now changed. We are now going to be reimbursed by discharge diagnosis at a rate fixed at a national average. The average rate of reimbursement will be that of the 100-bed hospital since this, statistically, is the national average level. In addition, the DRG's impose financial penalties both in cost overruns and in excess length of stay. One of the goals, therefore, of any system would be to help decrease length of patients' stay so that profitability of the hospital can be maintained.

Electronic displays are becoming a common and essential modality for the presentation of medical images. With the implementation of PACS, we anticipate that physicians may rely even more on electronic displays. Without system-by-system adjustments however, a given stored image will not produce the same display on all devices but rather, will be a function of the individual characteristics of the particular display device. We are developing a methodology which will ensure that displayed images are independent of the display device, i.e. are standard. We feel it is important that the selection of a standardization methodology be based upon observer performance. The standardization method we have implemented produces a display scale in which equal changes in driving intensity produce changes in physical luminance that are perceptually equivalent. We refer to this as perceptual linearization. For practical application of this technique, it would be desirable if a single correction function would cause a particular display to be perceptually linear for all observers in all situations. This correction function may depend upon factors such as variation between and within observers, ambient light, etc. We have begun to test the effect of these factors by comparing intra and inter observer variation. Results to date show that the variation between observers does not exceed intra observer variation.

A display system has been developed which is capable of full 2048 X 2048 pixel resolution with an eight-bit gray scale. An essential element of this system is a multibeam cathode ray tube (CRT), which enables the flicker-free display of 4,000,000 picture elements without the penalty of unmanageable video and deflection frequencies. Such a tube has been found to deliver 15-20 ft-Lamberts brightness with actual tube resolution somewhat in excess of the requirement above, over a 10 inch square raster. This paper describes the tube, its electron optics, and performance.

This paper describes the electronic requirements and solutions to those requirements to support a very high resolution interactive medical image display system. The paper is a companion to a paper to be presented at this conference describing the design of the special multibeam CRT which is a key element in the system. The electronic support package for the multibeam tube must be capable of supplying a video bandwidth of more than 400 Mhz. An image memory supplying the 400 MHz. pixel stream must store at least 16 images of 4 million, 8 bit pixels, each. Each image must be refreshed 60 times a second thus requiring effective horizontal sweep rates of 240 KHz. The design which meets these requirements has been completed and will be described in this paper. These requirements were achieved by using an inten sively parallel design in conjuction with the multibeam CRT, described elsewhere. Memory and logical components were combined in such a way as to provide electronic signal storage and transmission speeds far exceeding the current state of the art. In this development, a valuable contribution to medical electronics has been made which makes possible the interactive display, manipulation and analysis, in real time, of medical images with resolution of 2048 x 2048 or 4096 x 4096 pixels. This resolution equals the quality expected from photographic films.

The digital radiology department demands an accurate means to both digitize x-ray film and to represent digital data on film. This paper gives preliminary test results of the performance of a prototype laser scanner and printer system manufactured by the Konishiroku Photo Ind. Co., Ltd., for use in radiological imaging. The laser scanner digitizes x-ray film at a resolution of 2000x2400x10 bits for a 14x17 inch film. A variety of film sizes and optical density ranges can be scanned. The laser printer writes a digital image onto either 8x10 or 14x17 inch film with a pixel size of 80x80 microns. A 14x17 inch film can be written at up to 4288x5275x8 bit pixel resolution. A maximum of 16 images can be placed onto a single film. Software controlled look-up tables are used to correct for nonlinearities in the production of film density from digital data. Image magnification and minification are available under software control. Tests of uniformity, linearity, contrast and spatial resolution, signal to noise ratio, and laser field intensity distribution experiments for the laser scanner and printer have been conducted. This paper presents the results of these tests and some early clinical results.

Traditionally, multiple format recording emulsions for medical video imaging have utilized a film (transparent) base. The major reason for this is probably because the film and camera manufacturers felt the diagnostician is accustomed to viewing x-ray images on a film base and would prefer to view video images that way also. Because of the need to keep radiation exposure to patients at a minimum and the fact that photographic emulsions are generally very inefficient in utilizing x-ray radiation, a film base was the logical requirement for direct x-ray imaging as it enabled the image to be recorded by two emulsions rather than one. The transparent base thus allows viewing a photograph which is the result of the additive effect of the two emulsions. The use of transparent base imposed specific requirements that necessitated the development of a whole complex of equipment designed for the particular use of film such as the processing machines, their chemical solutions, and the famous viewbox and alternators that characterize the radiology departments of today.

There are a number of institutions and equipment manufacturers developing PAC systems or components of such systems. Much effort at our medical center has gone into the development of a workstation for radiologists, and this has been one of the primary devices built by others as well. As a result, diagnostic viewing stations are in operation and their designs are being refined. We have begun to examine the problem posed by the large number of viewing devices which may be required to satisfy the needs of the physicians and clinicians who refer patients to Radiology and want to see the studies which have been performed. Currently, this function is supported by a large number of film viewboxes and alternators both in the Radiology Department and in other hospital locations. It would be prohibitively expensive to replace all such viewing devices with diagnostic workstations, so we have considered the possible alternatives. We call this class of electronic viewing device the image viewstation. The goal of the design of an image viewstation is to satisfy user requirements at reasonable cost. However, we also need to design a device which can be interfaced to a network easily. This has led us to consider systems with both analog and digital image transmission. The former has the advantage of speed, the latter the desired full dynamic range and resolution. To a large extent, the method chosen is network dependent. This paper discusses the designs we have envisioned, the reasons we have chosen them, and how they might be implemented.

In order to define and direct our research efforts toward the development of an effective and efficient PACS display console,we have begun a series of observational studies of radiologists' reading patterns of CT images displayed on film. Preliminary findings indicate that a three-stage process occurs. In each stage specific behaviors have been identified, and some preliminary quantitative aspects have been determined. Depending on the stage when they occur, individual behaviors may serve multiple and/or different functions. We suggest that while certain functions may be necessary, certain specific behaviors might better be replaced by other behaviors. It appears that the manner in which radiologists read hardcopy is, in part, due to habit and, in part, due to limitations imposed by the image display modality. These habits may not necessarily be efficient or necessary with an electronic display. Radiologists have accepted the constraints imposed on them by film, and developed reading habits to operate within these constraints. Guided by our preliminary data, we are developing our concepts of what are the essential requirements of a PACS display.

Computed tomography is widely utilized for the detection and staging of neoplasm. Typical chest, abdomen or pelvis CT scans may produce 10 to 20 transverse slices for each region. The mental reconstruction of the three dimensional anatomy from these transverse sections can be done by a physician who has had training in the analysis and interpretation of cross sectional anatomy and pathology. This mental reconstruction, however, may take years to develop into an efficient tool. With the 3-D reconstructions used in this study, diagnostic information concerning the location, shape and spread of tumor masses can be presented in a simple, intuitive 3-dimensional display. This technique has been found to be useful for improving communication between diagnostic radiologists and consulting physicians.

Presenting three-dimensional information in the form of 3D solid models rather than as a sequence of two-dimensional intensity images provides many benefits in presurgical planning and diagnostic radiography. Although the model generation process does not add information to the sequential slice data, it does present images of organs and bony structures in a form more like the expected view of solid objects in natural scenes. Surface shapes and details of surface variations, which would require practiced observation of two-dimensional intensity data, are readily visible in the solid model displays making this information immediately available to a broad cross section of medical personnel. After a year of experience with a commercially available system, a Contour Medical Systems CEMAX-l000, which accepts input from several CT or MR scanner models and provides basic solid model displays, additional types of solid model viewing have been made available to clinical personnel for preliminary evaluation. The advantages and disadvantages in terms of subjective display quality, information content, and computational cost of several display methods have been investigated. Display of solid models by range encoding, heuristic mappings of intensity levels, and complete reflectance models have been compared for black-and-white and monochromatic color images. The option of displaying multiple objects in contrasting colors both as opaque and transparent objects has also been tested. Methods of surface acquisition from the two-dimensional data have been varied to match the material of interest and the characteristics of the original intensity data allowing improved representation of soft tissue. Finally, the utility of several types of time varying imagery is discussed, including the advantages of viewing rotating solid objects compared to viewing a collage of still pictures in many orientations. Some clinical examples of these experimental image display techniques are presented. The advantages of each are discussed in the context of the computational burden of display generation.

This paper presents a simulation of an image display system or workstation connected to a minicomputer system which is in turn connected to a larger computer system. The use of image display systems in a medical environment is summarized and parameters discussed include image and display resolution, communication bandwidth, and workstation configuration. User activities are broken down into five categories: display, edit, manipulation, generation and browse. Typical user sessions or workloads are then composed of these categories in various proportions and the response time and effectiveness of the simulated activity is measured. Some of the results are summarized in a number of charts and graphs, with conclusions and suggestions.

A cost-effective series of video workstations was developed for the display and manipulation of diagnostic medical images by the radiologist as well as the referring physician. The workstations feature various monitor configurations with 512-visible line display resolution, less than five second image retrieval, and a carefully considered function keyboard. An M/NET workstation is one of a potential twenty such display nodes in a distributed image management system (M/NET). In addition to the necessary display, enhancement, and mensuration features described for the display nodes, the M/NET system also provides for image Acquisition,Management/Archiving, and Communications. The M/NET system featuring the network of 512-visible line, easy-to-operate workstations was designed to provide a cost-effective transition from the current film-based to the future digital imaging department.

Reviewing previously acquired images of a patient when interpreting newly acquired ones is an important aspect of clinical radiology that has received relatively little attention in specifying requirements for automated picture archiving and communications systems (PACS). This paper examines reasons that reviewing comparison images plays an essential role in radiology, and it attempts to describe the range of policies currently employed by radiology departments. Data acquired from the use of the Missouri Automated Radiology System (MARS) at the University of Chicago are presented to quantify the number of comparison examinations that exist in this patient population. The implications of storing, retrieving and displaying these comparison images are discussed under different policies for both current film file systems and PACS.

This presentation is to outline the American College of Radiology (ACR) and National Electrical Manufacturers Associa-tion (NEMA) Digital Imaging and Communica-tion Standard (ACR-NEMA Digital Imaging and Communication), which is a creation of joint effort of many fine knowledgeable people on the ACR-NEMA Digital Imaging and Communication Standards Committee, especially Allen Edwin and Gwilym Lodwick, Co-Chairman and Laura Murphy, Executive of the Committee. We hope that the ACR-NEMA Standard will be useful in the development of digital diagnostic imaging and medical information.

The American College of Radiologists (ACR) and the National Electrical Manufacturers Association (NEMA) have sponsored a joint standards committee mandated to develop a universal interface standard for the transfer of radiology images among a variety of PACS imaging devicesl. The resulting standard interface conforms to the ISO/OSI standard reference model for network protocol layering. The standard interface specifies the lower layers of the reference model (Physical, Data Link, Transport and Session) and implies a requirement of the Network Layer should a requirement for a network exist. The message content has been considered and a flexible message and image format specified. The following Imaging Equipment modalities are supported by the standard interface... CT Computed Tomograpy DS Digital Subtraction NM Nuclear Medicine US Ultrasound MR Magnetic Resonance DR Digital Radiology The following data types are standardized over the transmission interface media.... IMAGE DATA DIGITIZED VOICE HEADER DATA RAW DATA TEXT REPORTS GRAPHICS OTHERS This paper consists of text supporting the illustrated protocol data flow. Each layer will be individually treated. Particular emphasis will be given to the Data Link layer (Frames) and the Transport layer (Packets). The discussion utilizes a finite state sequential machine model for the protocol layers.

Physicians are facing a growing crisis in information management. With their patients living longer, becoming ill with an increasingly complex mix of diseases, health care professionals are being faced with a growing number of diagnoses, treatments, and potential drug-drug interactions to keep track of. The busy clinician needs not only to be a medical specialist but an information technology specialist, to avoid be buried by this avalanche of data.

The number of diagnostic work stations needed in a digital x-ray department varies as a function of department case load, the primary types of diagnostic activity, the number of radiologists on staff, and their relative proficiencies. Other factors which must be considered are the distribution of work load, the amount of time a user is willing to wait for a station, the cost of the stations, and system reliability. There are two extreme solutions to this problem: one work station per department versus one per radiolo-gist. Empirical data available in the literature provides some limiting assumptions. This paper will use that data and analyze the problem using multi-server queuing theory. Estimates using this approach indicate that an "average" department in a medium to large hospital will need five or more work stations to successfully operate an all digital department.

Manipulation of several sets of two-dimensional cross sectional slices of computerized tomography (CT) or magnetic resonance (MR) data and the mental integration of such a large volume of information are major problems encountered by radiologists and surgeons in their attempt to make diagnoses. Contour Medical Systems has developed a physician's imaging console and a user-friendly interface intended to make this information immediately available and to facilitate image analysis. In this paper we will focus on the development of this user interface.

With the recent advances in medical diagnostic imaging technology, the need for an integrated display/processing station is becoming increasingly essential. This paper presents the architecture of a system that could effectively be used in the analysis of imagery derived from varying disciplines. This system has the ability not only to display multiple images at a high-resolution (1280x1024), but also provides the capability of interactive processing any of the displayed images. With extensive built-in graphics features and real-time pixel processing functions, this system would be a very viable component of a PACS environment.

Some error-free and irreversible data compression techniques applied to radiographic images are discussed in this paper. In the case of error-free compression, clipping and bit trunction, run-length coding, run-zero coding and Huffman coding are reviewed. In each case, an example is given to explain the steps involved. In the case of irreversible compression, the full-frame bit allocation in cosine transform domain method is described. Utilizing these compression techniques, we have compressed more than one hundred of radiographic images of different types. Our experience reveals that (a) it is possible to obtain a 3:1 compression ratio for error-free methods, and (b) for irreversible compression, the compression ratio achieved depends on the image type, the image size and the number of bits per pixel. In general, for a 512x512x8 image a 10:1, and for a 1024x1024x8 a 16:1 compression ratio can be achieved. Reconstructed images from these high compression ratio data do not appear to have visual degredation from the original image.

Data Compression will reduce storage device problems and cut down the time required for transmission of high resolution medical images. In order to improve subjective image quality, various mathematical models of the human visual system (HVS) have been proposed for image compression applications. Most of these models are based upon data from psychophysical experiments on human vision. Since some nonlinear characteristics are involved in the human visual system, the psychophysically based model may not reflect the real HVS. In this paper, the physiologically based HVS model is incor-porated with the popular two-dimensional discrete cosine transform (DCT) coding to encode digitized images of diagnostic human tissue sections acquired by a scanning transmission electron microscopy (STEM). Simulation results are compared at 1 and 0.5 bit/pixel between systems with and without the HVS model. Superior subjective image quality was observed from the system with the HVS model. A study of coding scheme mismatch was also conducted to evaluate the robustness of DCT coding for medical image applications. Due to the strong similarity between the statistics of images, simulation results showed little degradation from coding scheme mismatch.

If any data compression scheme for medical images shows some level of success, then this necessarily implies information redundancy in the image data. The information capacity of the system therefore must be larger than the information content. A powerful way to analyze and exploit the possibilities for optimum data-compression, transmission methods, and other useful image-processing options is to seek the eigenfunction representation of the image. By transforming the image-data into the eigen-representation, we obtain a transformed image array (the q-space image) which is the irreducible representation of the data for that imaging system. The coefficients of the irreducible representation are, in the parlance of information theory, "non-interfering symbols". Preliminary studies show that the eigen-representation allows immediate, direct, and predictable compression of the system data capacity to the image data information content. Our analysis of medical images from magnetic resonance imaging (MRI) scans has allowed us to demonstrate some specific data-compression strategies which will be most beneficial, rapid and flexible for PACS applications

The development of image management systems for radiology departments raises many questions regarding the archiving of digitally formatted data. A study was conducted in our department to determine the short-term and long-term archiving requirements. This study was based upon the retrieval rates of patient film jackets archived in the departments film library.

Early in 1983, a three-cable broadband network was installed in The New York Hospital-Cornell Medical Center using well-established cable-TV technology. This network was configured in a vertical tree topology. Currently, it extends over thirteen floors vertically and over two city blocks horizontally. It has now survived several major renovations on the various floors of the hospital. This survivability is a result of the siting of the main tree and of the isolation gained for the branches through the strategic placement of amplifiers. This communications system was designed in a modular fashion for later expansion and so that seven types of functions could be supported on the network without the addition of a new functional level disrupting the functions already existing on the system. Thus far, two functions (real-time image consultation and computer sharing) have been implemented, and two other functions (analog image storage and data base management) are in the prototype stage. Perhaps the most significant feature of our experience thus far has been the ease and utility of analog transmission and storage of images. This experience has lead us to postpone and even de-emphasize digital transmission and storage in our future plans.

Developing several key equipments, we made experimental systems for diagnostic image management, which could make it possible to file and retrieve both digital and analog image data, and to process them for editing, processing and displaying their clinical information on the high resolution monitors. The details of the systems and equipments, and preliminary experiences with them are reported.

A PACS has been designed for the University of Connecticut Health Center to serve all departments acquiring images for diagnosis, surgery and therapy. It incorporates a multiple community communications architecture to provide complete information management for medical images, medical data and departmental administrative matter. The system is modular and expandable. It permits an initial installation for radiology and subsequent expansion to include other departments at the Health Center, beginning with internal medicine, surgery, ophthalmology and dentistry. The design permits sufficient expansion to offer the potential for accepting the additional burden of a hospital information system. Primary parameters that led to this system design were based on the anticipation that departments in time could achieve generating 60 to 90% of their images suited to insertion in a PACS, that a high network throughput for large block image transfers would be essen-tial and that total system reliability was fundamental to success.

One of the most important functions that radiology has to perform is to provide timely and accurate diagnostic reports to referring physicians. Because of the importance of the report process and lack of an ideal reporting system, careful integration of existing reporting systems can not be overlooked.

Computer based information is becoming more common in the practice of modern radiology. Patient information, radiological reports and many radiological images are already in electronic digital format. There is also an increased application of computer technologies throughout the hospital. Use of hospital information system (HIS) and radiological information system (RIS)is receiving wider acceptance in the medical community, in an attempt to provide timely and efficient patient care. As more computer technologies and computer based management tools are utilized in the operation of the Radiology Department, full realization of integrated digital picture archiving and communication system (PACS) encompassing all the subsystems such as digital imaging devices, report generating system, archiving system, picture viewing system, information distribution system and administrative management component is becoming within the reach of currently available technologies (1,2). Only when all the patient data, radiological reports, images and administrative data can be simultaneously managed, will the full benefit of computer technology be realized in the practice of radiology.

The Department of Radiology at the University of North Carolina-Chapel Hill (UNC) has developed the concept for an integrated picture archiving and communications system. A communications network has been installed, and computer facilities are being interfaced within radiology areas. Other projects include the establishment of a computer simulation model of departmental operations and assembly of a limited picture archival and communication system (PACS) beginning with the CT and NMR Modalities. While this approach may provide immediate clinical benefits, the initial thrust has been towards evaluation of prototype systems with flexibility for modifications. PACS operational parameters are being studied for their acceptability in support of radiology clinical services. The goal is to provide objective operational data as a basis for planning system improvements.