Picture Archive and Communication Systems (PACS), which allow the electronic acquisition, storage, transportation, and viewing of medical images, hold the eventual promise of reduced costs, improved image-management logistics, and ultimately, improved patient care. But at what point in the future will PACS really cost less than film-based image management for a given hospital size; and how are these costs affected by the choice of the digital communication network? To address these questions, a static differential cost model has been constructed. Four PAC systems based on different speed networks as well as film, were considered for five different sized hospitals and two time periods. Based on the assumptions outlined, high-speed-network PACS (greater than 150 Mbps) are less costly than those based on low-speed networks for hospitals generating more than 15,000 procedures per year starting in 1995. Further, even though all possible PACS cost savings were not considered, high-speed network PACS appear to be less costly than film for hospitals larger than 30,000 procedures in 1995 and larger than 15,000 in 2000, while low-speed-network PACS should cost less than film for 60,000 and 30,000 procedure hospitals in 1995 and 2000respectively.

We have implemented two high resolution workstations within our pediatric radiology PACS module: a two-monitor 2K x 2K station and a six-monitor 1K x 1K station. The 2K x 2K workstation is under evaluation for primary reading of pediatric radiographs from a computed radiography unit. System implementation and evaluation methods are described. Operational efficiency measures of both film and digital systems are reported. This study is our first attempt to integrate a primary viewing station into a busy clinical environment. The 1K x 1K workstation is available 24-hours a day, 7 days a week for fast reviews by referring physicians. Images from a compated radiography system are available at the workstation in about 8 minutes. A digital voice reporting system is being developed to communicate radiology reports from the 2K x 2K workstation to the 1K x 1K secondary review station.

Seven months experiences of a filmless PACS (named HU-PACS) which covers Radiology, Orthopedic, Internal medicine and General Surgery departments are reported. The PACS has only 20 Image terminals but handles more than 50% of images produced which is about 1000 images per working day. Physicians of the departments have many criticisms and opinions on the PACS but generally speaking it is well accepted and inspiring the physicians to improve the PACS in its image quality and other functions instead of being discarded. Preliminary clinical assessment are performed and reported also.

Since September 1988 a PAC System (CommView by AT&T and Philips) is in operation in the Radiology Department of the University Hospital of Trieste, Italy. A research project is presently in progress aiming at providing factual evidence for the evaluation of this kind of systems as far as operational, technical, clinical and economic aspects are concerned. The general approach to this research consists in implementing and monitoring a PACS in a stepwise way, starting with an "entry-level" system With respect to our actual experience we propose a multi-attribute approach to the problem of justifying PACS, thus classifying attributes into three broad classes: operational, clinical and economic. For each of the relevant problems we distinguish what could be expected in general, what we actually found in practice and what we suppose to do in the future. Justifying PACS turns out to be a difficult task. Within a longer horizon, a horizontalintegration within the Radiology Department and among different Radiology Departments appears to open the more effective perspectives.

The Department of Radiological Sciences at the UCLA School of Medicine is developing an archiving and communication system (PACS) for digitized ultrasound images. In its final stage the system will involve the acquisition and archiving of ultrasound studies from four different locations including the Center for Health Sciences, the Department for Mental Health and the Outpatient Radiology and Endoscopy Departments with a total of 200-250 patient studies per week. The concept comprises two stages of image manipulation for each ultrasound work area. The first station is located close to the examination site and accomodates the acquisition of digital images from up to five ultrasound devices and provides for instantaneous display and primary viewing and image selection. Completed patient studies are transferred to a main workstation for secondary review, further analysis and comparison studies. The review station has an on-line storage capacity of 10,000 images with a resolution of 512x512 8 bit data to allow for immediate retrieval of active patient studies of up to two weeks. The main work stations are connected through the general network and use one central archive for long term storage and a film printer for hardcopy output. First phase development efforts concentrate on the implementation and testing of a system at one location consisting of a number of ultrasound units with video digitizer and network interfaces and a microcomputer workstation as host for the display station with two color monitors, each allowing simultaneous display of four 512x512 images. The discussion emphasizes functionality, performance and acceptance of the system in the clinical environment.

The increasing use of magnetic resonance imaging (MRI) as a clinical modality has put an enormous burden on medical institutions to cost-effectively teach Mill scanning techniques to technologists and physicians. Since MRI scanner time is a scarce resource, it would be ideal if the teaching could be effectively performed off-line. In order to meet this goal, the Radiology Department has designed and developed a Magnetic Resonance Imaging Simulator. The Simulator in its current implementation mimics the General Electric Signa scanner's user-interface for image acquisition. The design is general enough to be applied to other MRI scanners. One unique feature of the simulator is its incorporation of an image-synthesis module which permits the user to derive images for any arbitrary combination of pulsing parameters for spin-echo, gradient-echo, and inversion recovery pulse sequences. These images are computed in five seconds. The development platform chosen is a standard Apple Macintosh-Il computer with no specialized hardware peripherals. The user-interface is implemented in HyperCard. All other software development including synthesis and display functions are implemented under the MPW 'C' environment. The scan parameters, demographics and images are tracked using an Oracle database. Images are currently stored on magnetic disk but could be stored on optical media with minimal effort.

Since October, 1988, a picture archiving and communications system has been our primary means of managing ultrasound images. The ultrasound area was selected because of satisfactory image resolution with PACS, relatively low volume (18-20 cases per day), and minimal clinician request for the images. We have performed over 6,000 examinations, comprising over 50,000 images during this time period. The images and examination reports, stored on optical disks, are available on line for one week before removal to shelf archiving. When shelf archived studies have required de-archiving, most have been retrieved within five minutes. Rapid, reliable image acquisition allowed for decrease in the study performance time. Display monitor illuminance stability eliminates density variation associated with improper exposure of films. Immediate availability of images makes possible on-line monitoring of technologist performed cases. No images have been lost. Technical problems have included image degradation because of pixel shift artifact, hardware and software failure, and duplicate entry of patient demographic data. Despite rapid initial display and review, subsequent image review is 6-10 times slower than with film based studies. If PACS are to succeed, automatic archiving and de-archiving, improved system reliability, and improved speed of display must be developed.

Although trauma has been recognized as the third leading cause of death overall in the United States, it is the leading cause of death in persons less than 40 years of age. There are approximately 150,000 trauma deaths per year. About 50,000 of these occur on the highway, with victims suffering severe multiple system injury. For every death, 2-3 patients suffer major long-term disability. An important estimate states that one in four of these deaths could be prevented1. With this unfortunate fact in mind, trauma systems have been developed across the country over the past decade. A trauma system consists of 1) a sophisticated prehospital system with paramedics initiating resuscitation and rapid transport; 2) on-line communication networks linking prehospital personnel with physicians in trauma centers who provide medical control; and 3) the trauma center, which acts as the hub of the system by coordinating the prehospital efforts and providing definitive trauma care. The trauma center concept has been developed and defmed by the American College of Surgeons, which established three tiers: Levels I, II, and III, with the most severely injured patients being triaged to Level I institutions. Since death and major disability occur within a short time following injury in those patients, diagnosis and therapy must be both simultaneous and expeditious. This is accomplished by having surgeons--trauma surgeons, neurosurgeons, orthopedic surgeons and anesthesiologists--and operating room teams present in the hospital at all times. A level I trauma center recreates, in part or in whole, a hospital environment focused on a limited spectrum of disease. All the problems in medical care related to imaging that are experienced in the trauma center also exist throughout the remainder of the hospital system. However, due to the speed of health care delivery and the concentrated focus of multiple surgical subspecialists and radiologists on the same small number of medical images the problems become severely exacerbated. Virtually all of the most serious problems related to imaging in a level I trauma center can be resolved with digital acquisition and storage of conventional images combined with transmission of already digital images (CT scans) to a central display location. Additional benefits can be expected with the provision of remote access to images from intensive care units and operating rooms and rapid transmission of images to the ShockfTrauma section where the earliest and most intense medical care is administered. Consequently, a PACS system was installed at the Regional Medical Center at Memphis. Installation was completed in the first week of December, 1989 and the system is currently being phased into operation, and evaluated.

A digital image network has been installed in the James Whitcomb Riley Hospital for Children on the Indiana University Medical Center to create a limited all digital imaging system. The total system is composed of commercial components, Philips/AT&T CommView system, and connects an existing Philips Computed Radiology (PCR) system to viewing workstations located in the intensive care unit and the new born nursery. The purpose and design of the system is to input the portable chest images from the PCR system, and to display these images at the remote workstations on high resolution monitors for direct viewing by referring clinicians, thus eliminating some of their visits to the radiology department three floors away. The design criteria includes the ability to centrally control all image management functions on the remote workstations to relieve the clinicians from any image management tasks except for recalling patient images. The principal components of the system are the Philips PCR system, the acquisition module (AM), and the PCR interface to the data management module (DMM). Connected to the DMM are a display workstation (DW), a optical disk drive, and a fiber optic to ethernet gateway. The ethernet link is the network connection to the two results viewing stations (RVS) located approximately 100 meters from the DMM. The DMM acts as an image file server and image archive device. The DMM manages the image database and can load images to both the DW's and the two RVS's. The system has met the initial design specifications and can successfully capture images from the PCR and direct them to the RVS's. Additional studies are beginning to determine the optimal image management procedures such as when to archive and purge images from the DMM.

Experience gained from Siemens PACS installations since 1986 has been of great value and has lead to a continuous upgrading culminating in our today's products. Valuable information not only on technical aspects but also experience in PACS project engineering and operating a PACS has been acquired from a manufacturer's point of view. It might supplement what the PACS users are reporting on their individual results.

As part of a multi-project program, we are conducting a number of studies on the implementation and use of digital radiology in the modern radiology environment. In several of our completed studies, 10 participating radiologists have each been asked to view and interpret 300 or more posteroanterior chest cases, presented on both conventional film and on a high-resolution workstation. Because we believe that radiologists' subjective acceptance of new uses of the digital modalities is important if these techniques are to be more widely utilized, we administered a comprehensive questionnaire to these radiologists to examine their attitudes regarding use of the workstation vs use of film. General adjustment to the digitial modality, personal comfort levels, and perceptions about time required for and resulting accuracy of readings, as well as advantages and disadvantages of the digital system were examined. In general, the radiologists in this study felt slightly less comfortable using the workstation than using film, although they found adjustment to the workstation to be fairly easy. The most important advantage of the workstation was thought to be the ability to control contrast and brightness; the most important disadvantage, that interpreting from the workstation was more tiring than interpreting from conventional film when comparable numbers of images (50) were interpreted in a single reading session.

In a large-scale, multi-reader study to investigate questions surrounding the issue of the implementation of picture archiving and communications systems (PACS) into the modern radiology environment, we examined the effect that the incorporation of a clinical history into the reading process would have on levels of diagnostic accuracy. Because we wanted to test the inclusion of the clinical history in an environment as close to that of the clinical situation as possible, we defined "history" to be a concise, objective, and potentially computer-extractable version of what appears in the patient records, including a statement from the referring physician when this is available. In a series of studies, four radiologists interpreted 247 posteroanterior normal and abnormal chest images on conventional film both with and without accompanying patient histories; five radiologists read the same number of images presented on a high-resolution video workstation with and without clinical histories. There were no significant differences (p = .05) in diagnostic accuracy rates with or without clinical history for either the film or the workstation in cases of interstitial disease, nodules, or pneumothorax. Diagnostic accuracy for the radiologists as a group was not affected by the presence of the clinical history. We concluded that for the interpretation of these abnormalities, the incorporation of clinical history with images in the PACS environment should not be a major goal.

We designed and implemented a high-resolution video workstation as the central hardware component in a comprehensive multi-project program comparing the use of digital and film modalities. The workstation utilizes a 1.8 GByte real-time disk (RCI) capable of storing 400 full-resolution images and two Tektronix (GMA251) display controllers with 19" monitors (GMA2O2). The display is configured in a portrait format with a resolution of 1536 x 2048 x 8 bit, and operates at 75 Hz in a noninterlaced mode. Transmission of data through a 12 to 8 bit lookup table into the display controllers occurs at 20 MBytes/second (.35 seconds per image). The workstation allows easy use of brightness (level) and contrast (window) to be manipulated with a trackball, and various processing options can be selected using push buttons. Display of any of the 400 images is also performed at 20MBytes/sec (.35 sec/image). A separate text display provides for the automatic display of patient history data and for a scoring form through which readers can interact with the system by means of a computer mouse. In addition, the workstation provides for the randomization of cases and for the immediate entry of diagnostic responses into a master database. Over the past year this workstation has been used for over 10,000 readings in diagnostic studies related to 1) image resolution; 2) film vs. soft display; 3) incorporation of patient history data into the reading process; and 4) usefulness of image processing.

Using simulation studies and the data obtained from our clinical investigations, we are comparing some of the techniques commonly employed in ROC analysis. The general areas of our investigation include estimation of the area under the ROC curve and its associated standard error, comparison of the areas under two ROC curves, and simultaneous comparison of the areas under more than two ROC curves. For single ROC curves, maximum likelihood estimation based on the binormal model is being compared to the approximate area obtained using the Wilcoxon Statistic. Standard errors obtained from classical asymptotic theory are being evaluated and compared to standard errors based on jackknifing. With regard to the task of assessing differences in the diagnostic accuracies of two systems, our investigations have compared the maximum likelihood estimation method to a nonparametric method based on the Wilcoxon Statistic, focusing primarily on correlated observations, e.g., from readings of different modalities' images of the same patients. Extensions of existing methodology to the problem of comparing ROC areas from more than two modalities are being addressed as well as problems associated with the incorporation of multiple readers.

In this study, we used a high-resolution workstation (1536x2048) with window width (contrast) and level (brightness) manipulations available to test diagnostic accuracy rates of seven experienced radiologists on 300 posteroanterior chest images in a clinically simulated environment. Observer performance was assessed by receiver operating characteristic analysis, and the results were compared to performance when the same cases were interpreted by the same radiologists using conventional film. Efficiency on the workstation was comparable to, if not better than, that with the conventional film, but accuracy indices (Az) for the detection of interstitial disease and pneumothoraces were lower for the workstation than for conventional film. Since window width and level did not compensate for the lack of resolution, we then examined in a following study the effects of different processing options (reverse mode, unsharp masking) on diagnostic accuracy. These image processing options did not significantly improve readers' performance.

In an ongoing, multi-reader, multi-project program dealing with the interpretation of radiological images, we have examined several issues of methodology which have not as yet been addressed that may impact on the determination of reader performance as measured by receiver operating characteristic (ROC) analysis. Among these are issues associated with the training of observers prior to such studies. We employed three types of observer training that we found to be necessary for the successful performance of such studies: 1) a general instructional session for observers on the study protocol and system operation; 2) practice with an interactive computerized feedback teaching file that demonstrates the imaging systems and familiarized readers with an idea of the types of cases that were used in the study along with their correct diagnoses; and 3) training sessions in which readers were taught the manner in which to distribute answer ratings over an ordinal confidence scale. The possible effect of such types of training on the performance and results of ROC studies should be carefully considered prior to their commencement.

High resolution gray scale displays are being evaluated for the display of 2K x 2K x 12 bit digital images. A comparison has been analyzed for conventional chest radiographs, digitized chest radiographs (2K x 2K x 12 bits) printed on laser film, and a high resolution (2K x 2K x 12 bits) gray scale display. An initial study has been conducted comparing computed radiography laser printed films and an interactive 1K x 1K x 12 bit display.

An Integrated Services Data Network test configuration, developed in collaboration with SBC Technology Resources Inc., has been utilized to gain experience with both hardware and software interface isssues relevant to wide-area extensions of PACS. Example workstations capable of inquiry and display of radiological information have been interfaced to the ISDN trial network and tested. A DEGnet router was used to interface our Ethernet-based PACS to an ISDN network. Average throughput rates of approximately 60Kb/s for the lB (64Kb/s) channel and 120Kb/s for the 2B (l28Kb/s) channel were measured. Results were obtained for both embedded and external ISDN terminal adapters.

In this paper, the PACS network data rate requirement in a radiology department is discussed. The network data transmission rate under real data traffic in a medium-size hospital is evaluated using a simulation study. The network is assumed to be a circuit switching network. A preloading system, in which each workstation has local storage to store the next procedure, is assumed, and the network transmission rate which satisfies the response time of this system is obtained.

There are a number of generic problems that researchers in medical imaging have in common. For example, researchers in MR, CT, PET, SPECT, DSA, and digital radiography all need to display and window digital images. We describe here an integrated network with tools applicable to all of the current areas of medical imaging that enable researchers at Duke to share resources and solve software and hardware problems in a unified effort. We will show examples where efforts in a specific area f imaging research can be readily applied in new ways to different imaging modalities through the facilities provided in this integrated approach. We will point out some of the problems and opportunities in the research environment that are different from those encountered in clinical PACS systems.

The concept of picture archiving and communication systems (PACS) is now widely accepted in the medical community. In order to bring the concept to reality, however, innovative designs and implementations are needed. One such design is a fiber optic star based PACS, conceived by the University of Arizona and Toshiba corporation. This PACS network is based on a multiplexed passive star local area network with wavelength-division multiplexing to provide separate logical channels for transfer of control and image data. The system consists of an Image-Network (INET), for image transfer at a rate of 140 Mbps, and a Control-network (CNET), operating at 10 Mbps, for mediating the flow of image transfers. INET is a circuit switched network where a network supervisor grants users permission to transfer images over it, while CNET employs the CSMA/CD protocol for bus arbitration. Before such a system can be deployed, an accurate evaluation study must be carried out to estimate its performance characteristics. Such evaluations are complicated both by the complexity of the PACS itself and the varied demands that are placed on such a system. An novel approach based on siochastic aciiviiy neiworks, a stochastic extension of Petri nets, is useful in this regard. Stochastic activity networks were used to develop a detailed model of the command and image channels. The performance of the system was then evaluated under realistic workload conditions. In particular, we were able to estimate a number of important performance variables including the image response time, command channel delay, and queue length each type of node and the network supervisor. The results 1) show that stochastic activity networks are an appropriate model type for evaluating picture archiving and communication systems, 2) delineate the workload conditions under which PACS may effectively operate, and 3) show that even when these conditions are exceeded, the command channel load remains extremely light. Results of this type are useful both to designers of other PACS networks and those interested in this particular PACS design.

The PAC system: KIDS (Kyoto University Hospital Image Database and Communication System) has been expanded to include several major digital imaging modalities such as X-ray CT, MRI, DSA and CR. The fiber optic high-speed local area network and the workstation with quick image handling are newly designed. The system (new KIDS) is intended to achieve a film-less environment in the department of radiology and to evaluate the feasibility of a hospital-wide PAC system. The present status of the system at the end of 1989 including a image workstation installed in a lecture hall for clinical education is described.

Victoria General Hospital, which is a part of the Greater Victoria Hospital Society, is a 443 bed community hospital with a full-service medical imaging department that includes MRI. In August 1987, four rooms (chest radiography, GI fluorography, CT, and cardiac angiography) were connected to a Picture Archive and Communications System (PACS). We present the experience gained from two years of study with this prototype PACS and briefly describe its hardware and software configuration. Reported in detail is the measured image transfer performance of the PACS for each of the four image sources. Conventional films require more than 150 seconds from exposure to film availability for reporting. Using PACS, chest and GI images take 77 seconds per image from exposure to viewing, 31 secs for CT, and 40 secs for general angiography. The elapsed times with PACS between the various software processes for each modality, and those needed to set up image archive folders are detailed. The present imaging equipment at VGH, a typical community hospital, is specified and is to be integrated into a department-wide PACS. Required PACS performance levels relative to the clinical demands are described, and compared with the current PACS. The information and experience acquired by testing the VGH PACS has been used in the planning of the full implementation of PACS at VGH (ref. 10. Fisher et al). The next phase (3C) of PACS implementation is described.

In expanding our image management and communications system (IMACS) to include a new machine in the Ultrasound Section, we first modeled the impact of attaching the unit to either of our two acquisition modules (AM). Using the software package available to us, we could predict image queue lengths and average image waiting times (before transmission to the central archive). The AMs have attached a variety of devices with differing image production loads. The modeling allowed us to select the AM which would be least impacted (in terms of resoponse time to the devices sending data to the AM) by the added ultrasound machine. We found that though the input response times would not change much with the ultrasound machine connected to either AM, there was a significant impact on the predicted output queue length, with the more heavily loaded AM suffering larger increases in output queue dwell time if the new machine were connected to it. Based on these results, we elected to redistribute the AM loads, and connect the ultmsound machine so as to maintain a relatively balanced load. We tested certain parameters of the model by measuring input and output response times ofan AM under different, artificially induced, acquisition loading. The changes predicted by the model agreed with those measured in order-of magnitude terms.

A model is presented for evaluating the diagnostic value of an imaging system in clinical use. It is site-specific and includes thedistribution of diagnostic findings, diagnostic accuracy, the importance of the test outcomes to the physician, and the timeliness of information delivery.

Abstract not available.

Picture Archiving and Communication Systems (PACS) represent an enormously expensive technological innovation which has the potential to alter the way in which radiology is practiced. In a 1985 survey of radiologists and radiology administrators, over a quarter of those individu1s expected to have some component of an operational PACS system by 1986. The survey respondants perceived advantages including fast access to files, multiple user simultaneous access to films, and possibly secondary benefits in accuracy in diagnosis and improved communications. Progress in widespread PACS implementation has considerably lagged the rate anticipated in the 1985 study. In order to evaluate the current expectations of PACS and contrast them to the earlier enthusiasm we have undertaken this study of the current perceptions of PACS among both radiologists and clinicians.

Our strategy in studying PACS is to evaluate its clinical implementation working with equipment supplied by an established manufacturer. Fiscal and personnel resources required to design and integrate the hardware components and operational software to develop a functional PACS precluded a bottom up development approach at our institution. Imaging equipment vendors possess more abundant design development resources for this task and therefore can support a more rapid development of the initial components of PACS. For this reason we have chosen to serve as a beta test site to study the viability of the basic PACS components in a clinical setting. Our efforts primarily focus on: (1) image quality; (2) cost effectiveness; (3) PACS/HIS/RIS integration; (4) equipment and software reliability; and (5) overall system performance. The results of our studies are shared with the vendor for future PACS development and refi nement. To attain our investigational goals we have formed an interdisciplinary team of Radiologists, Perceptual Psychologist, Economist, Electrical and Industrial Engineers, Hospital Information System personnel and key departmental administrative staff. For several reasons Pediatric Radiology was targeted as the initial area for our PACS study: a small area representative of the overall operation,tight operational controls and willingness of physicians. We used a step-wise approach, the first step being the installation of PACS exclusively within the physical confines of Pediatric Radiology.

A high-level description of a system to pre-fetch comparison radiographs in an Image Management and Communication System (IMAC) is presented. This rule based system estimates the relevance of previous examinations for comparison to the current examination arid uses this determination to pre-fetch comparison studies. A machine learning module should allow the system to improve its skill in pre-fetching examinations for each individual radiologist. This system could be tailored to fit the desires of individual radiologists.

In this paper we present initial work on the design and implementation of an image database system for medical images. The objective of this project is the development of an environment that will permit the quick and intelligent access of images using their content. An image database is a complex system consisting of many components that must be unified and integrated. These are the image analysis component that derives a representation of the stored images, the database component, and the user interface component. This paper focuses primarily on the justification of the design decisions, an illustration of the major features of such a system, and a requirements definition for its implementation. An object-oriented approach is proposed which meets these requirements. The capabilities of an object-oriented data model are demonstrated and possible extensions, needed to meet the data modeling requirements, are discussed.

We are developing a personal image filing system as an alternative method of archiving radiolOgical images. The system can be configured with three popular computers as host: IBM PC/AT, Macintosh II, and SUN. The images are archived on the 3.5-inch optical disk and can be retrieved by any of the host computers for display.

We conceived a hybrid PACS which combines analogue and digital elements where it is required. We experimented a clinical application of the dynamic image utilizing the analogue element which is a feature of the hybrid PACS. With the analogue optical disk, dynamic image acquisition of 30 frames per second is possible and image recording is also possible with any modality of X-ray fluoroscopy function. In addition, X-ray dose to the patient is one-fifth to one-tenth compared with the conventional cine filming method while its still image is clear and a data acquisition to the PACS is easy. We applied this system to the areas of the otorhinolaryngology, orthopaedics and respiratory disease and could perform a various analysis of the disease status under natural conditions, also we found that this system is useful for judging an effect of a treatment as well as jugding a functional recovery after an operation and an effect of a rehabilitation. We are also planning to extend the application to the areas of the digestive system and the group examination from now on.

An IBM Application System/400 (AS/400) anchors the Mayo Clinic/IBM joint development PACS project. This paper highlights some of the AS/400's features and the resulting benefits which make it a strong foundation for a medical image archival and review system. Among the AS/400's key features are: 1. A high-level machine architecture 2. Object orientation 3. Relational data base and other functions integrated into the system's architecture 4. High-function interfaces to IBM Personal Computers and IBM Personal System/2s' (pS/2TM).

Following our initial investigation of the utility and performance of an AS/400 based token ring network for MRI image archival, we have designed and embarked upon a project to install multiple token ring networks at three physically separated facilities connected with fiber optic communication to archive and display CT and MRI image data. The system design will connect six MRI systems, six CT systems, four optical storage systems, three image display systems, three radiology report scanners, and a foreign archival system. Network communication and archival functions will be managed by an IBM AS/400. Preliminary estimates indicate that savings of media, space, and personnel as well as the additional benefit of online reports with image data will enhance the quality and efficiency of clinical practice. Additionally, we are evaluating a medical image display system (MIDS) to be used for on-demand image review, prefetched image review, and a potential replacement for modality-specific physicians consoles. The MIDS workstation is a promising addition to clinical practice both from the standpoint of an efficient means of image review as well as image interpretation. We consider this joint project to be an important step in the process of understanding the impact and subsequent consequences of electronic image transmission, information communication, and archival within a busy clinical setting. Our performance and design criteria focus upon the impact of this system on the clinical practice within radiology and the subsequent ramifications of PACS throughout the institution.

The first parameter of interest in considering network performance is signaling rate. To understand end-to- end image transfer rate, however, signaling rate is often the parameter of least importance. The basic image transfer in a PACS is disk-to-disk or at best disk-to-memory. The real performance bottlenecks are disk and system bus I/O rates, CPU and operating system efficiency in executing network protocols, and throughput of network interface units. Except for issues of congestion, the end-to-end performance of a 100 Mbps FDDI net may not be significantly different from that of a 10 Mbps Ethernet. The system architecture of workstations and storage nodes is the key to overall PACS network performance. Simulation studies indicate that a high speed disk subsystem combined with an intelligent network interface unit avoids the bottlenecks associated with conventional systems and can make use of high bandwidth media such as broadband ISDN. Experience with a preliminary demonstration of broadband ISDN connectivity as part of the Berlin Communications (BERKOM) project will be presented.

A software package has been developed to aid psychophysical studies on diagnostic image quality. The program, called FEASIBLE, supports all phases of experiments to evaluate the diagnostic quality of PACS components, image acquisition and manipulation methods and display techniques. The software allows the user to design any such an evaluation study, to perform an arbitrary number of psychophysical sessions and to statistically analyze the collected data with the Receiver Operating Characteristics (ROC) technique. Furthermore, the calculated ROC curves can be plotted in graphs to be integrated in publications. Within the scope of the Dutch PACS project, a PACS research project jointly carried out by the Utrecht University Hospital, BAZIS and Philips, FEASIBLE has been used to investigate the diagnostic accuracy of the PACS prototype that was clinically evaluated. The package has been sent - upon request - to more than 20 scientists in the field of radiological imaging. This paper explains the backgrounds and the features of the latest version of the program, and the way to use it in an actual ROC study.

The University of Washington participated, during 1986-89, in a clinical evaluation of commercially available PAC systems under the sponsorship of the MITRE Corporation (Contract N55-200). As part of this work, we developed a general approach to the clinical evaluation of PACS workstations. This paper describes the approach and its application to a specific workstation, the AT&T CommView Diagnostic Workstation.

The functional requirements of a radiological viewing station has been determined by feedbacks from radiologists. In this project, we incorporated all the functional requirements in a custom designed display station with emphasis on affordablility and display speed. Utilization of the open architecture of VME and VSB buses led to expandability of display screens, memory sharing, and availability of off-the- shelf frame buffers.

A major factor in determining the overall utility of a medical Picture Archiving and Communications (PACS) system is the functionality of the diagnostic workstation. Meyer-Ebrecht and Wendler [1] have proposed a modular picture computer architecture with high throughput and Perry et.al [2] have defined performance requirements for radiology workstations. In order to be clinically useful, a primary diagnosis workstation must not only provide functions of current viewing systems (e.g. mechanical alternators [3,4]) such as acceptable image quality, simultaneous viewing of multiple images, and rapid switching of image banks; but must also provide a diagnostic advantage over the current systems. This includes window-level functions on any image, simultaneous display of multi-modality images, rapid image manipulation, image processing, dynamic image display (cine), electronic image archival, hardcopy generation, image acquisition, network support, and an easy user interface. Implementation of such a workstation requires an underlying hardware architecture which provides high speed image transfer channels, local storage facilities, and image processing functions. This paper describes the hardware architecture of the Siemens Diagnostic Reporting Console (DRC) which meets these requirements.

One of the challenges to the PACS workstation designer is to cost effectively fulfill the range of requirements from those of the entry-level system to those of the fully digital hospital. The workstation must therefore support many of the functions of film-based image storage and image information systems. A primary diagnosis workstation must support rapid selection and viewing of patient examination folders. The user must easily and rapidly sequence through a stack of examinations. Activities such as printing films, archiving images, and retrieving older examinations must be initiated easily, without interfering with the diagnostic use of the workstation. Some of these functions are executed by the workstation itself while others are performed by service nodes on the PACS network. The workstation's user interface must make the architecture of the system transparent, providing access to the services while concealing the complexity of their accomplishment. This paper discusses the approach taken to the user interface of a commercial PACS primary diagnosis workstation. This workstation may be configured as an entry-level PACS or as a single workstation in a very complex system. Local image storage, retrieval, archiving and display functions will be described. Networking with modalities, image storage systems and image management systems will also be discussed.

As prototype PACS grow into fully digital departmental and hospital-wide systems, effective information storage and retrieval mechanisms become increasingly important. Thus far, designers of PACS workstations have concentrated on image communication and display functionality. The new challenge is to provide appropriate operator interface environments to facilitate information retrieval. The "Marburg Model" 1 provides a detailed analysis of the functions, control flows and data structures used in Radiology. It identifies a set of "actors" who perform information manipulation functions. Drawing on this model and its associated methodology it is possible to identify four modes of use of information systems in Radiology: Clinical Routine, Research, Consultation, and Administration. Each mode has its own specific access requirements and views of information. An operator interface strategy appropriate for each mode will be proposed. Clinical Routine mode is the principal concern of PACS primary diagnosis workstations. In a full PACS implementation, such workstations must provide a simple and consistent navigational aid for the on-line image database, a local work list of cases to be reviewed, and easy access to information from other hospital information systems. A hierarchical method of information access is preferred because it provides the ability to start at high-level entities and iteratively narrow the scope of information from which to select subsequent operations. An implementation using hierarchical, nested software windows which fulfills such requirements shall be examined.

The clinicians at Shands Hospital at the University of Florida were frustrated by the filmbased system, which served them in taking care of their patients in the Intensive Care Units (ICU). Frequently, they would come to the radiology department three to six floors below their respective ICU only to be stymied by the inability to find a study that was needed for patient care. This led to much consternation and outright animosity towards the personnel in the radiology department. On the other hand, the radiologists were exasperated by the missing films and lack of cooperation of the ICU team of physicians and nurses. The common denominator was an inefficient communications system for sending chest and abdominal films to the ICUs. A solution was proposed, tested, and installed by DuPont. Our initial experience with this system was reported.1 We have now modified and extended this system to all of the ICUs, which represents the basis for this report. The clinical review system (CRS) was initially placed in the medical ICU and very shortly thereafter expanded to serve all six ICUs at Shands Hospital. These include surgical, neurosurgical, cardiothoracic, pediatric and neonatal ICUs. Approximately 30,000 portable examinations are done each year in these areas.

We have evaluated the performance of Ethernet, Fiber Distributed Data Interface (FDDI) and Canstar Super 100 networks in a multimodal Picture Archiving and Communication System (PACS) environmenl image data transfer rates were measured for SUN workstations and PC/ATs running TCP/IP communication protocols. The communication between two computers was on a server-client basis and all transfer rates were measured between computer memories in order to eliminate the bottleneck imposed by storage devices. Four test models were used to evaluate the performance of the three individual networks: bidirectional, centralized parallel and relay models. This paper describes the methods of network throughput evaluation using these four models and presents the resulting performance data for the three networks tested.

The Radiology Department at the Hospital of the University of Pennsylvania is currently expanding its prototype Picture Archiving and Communications System (PACS) into a fully functional clinical system. The first phase of this expansion involves three major efforts: the upgrade of the 10-Mbit token-ring to an 80-Mbit backbone with associated sub-nets, the implementation of a large-scale image archive, and, an interface between the PACS and the Department's Radiology Information System. Upon the completion of this phase, the PACS will serve the storage and display needs of four MRI scanners and four of the Hospital's Intensive Care Units. This paper addresses the implementation of a software suite designed to duplicate and enhance conventional Film Library functions on a PACS. The structure of an electronic 'folder' based upon the ACR/NEMA Digital Imaging and Communication Standard is also introduced.

The effective delivery of health care has become increasingly dependent on a wide range of medical data which includes a variety of images. Manual and computer-based medical records ordinarily do not contain image data, leaving the physician to deal with a fragmented patient record widely scattered throughout the hospital. The Department of Veterans Affairs (VA) is currently installing a prototype hospital information system (HIS) workstation network to demonstrate the feasibility of providing image management and communications (IMAC) functionality as an integral part of an existing hospital information system. The core of this system is a database management system adapted to handle images as a new data type. A general model for this integration is discussed and specifics of the hospital-wide network of image display workstations are given.

The design of workstations for use in picture archival and communications systems (PACS) has received a significant amount of attention; rightfully so, as it is the most noticeable element in such a system. Various aspects of medical imaging workstations have been studied in detail, including feature requirements, user interface, performance requirements, and other human factors considerations. Experience during implementation of the CommView system PACS has shown that it is most important to carefully consider the operational environment in which these workstations will be placed to match the tasks and flow of information to, from, and within the workstation with the tasks and flow of information in the traditional film-based environment. Equally important is the cost of such a workstation. and the well - designed workstation is one that successfully matches the operational environment at a reasonable price. We describe in this paper the process undertaken to optimize a PACS workstation for the operational environment of diagnostic reading and reporting. Operational models have been developed, through interviews with and observations of users at Bowman Gray/Baptist Hospital Medical Center. Duke University Medical Center. San Francisco VA Medical Center, and Georgetown University Hospital. Hardware and software designs of the workstation were optimized to match the workstation to those operational models, including matching the time intervals, and providing easy access to relevant exams and to historical exams to be used for comparison.

The exchange of information between a Radiology Information System (RIS) and a PACS is essential to optimizing the utility of a PACS. Some of the benefits awarded by implementing an interface include a reduction or elimination of repetitious data entry, the availability of more accurate information on the PACS, a reduction in workload for the technologists, registration clerks, transcriptionists, etc, and the availability of more accurate data for automating the PACS. This paper discusses the Georgetown experience of interfacing an HIS/RIS and PACS, by describing the development of the interface and its impact on clinical operations.

Despite the much-discussed advantages of the all-digital radiology department, the speed of electronic display continues to be a major obstacle to its acceptance; physicians generally agree that sophisticated workstation functionality cannot compensate for an interpretation environment that delays diagnosis. Two design schemes have been devised and discussed at length at the Bowman Gray School of Medicine (BGSM) that will improve the efficiency of image transmission significantly. The first of these is image routing and pre-loading. The central archive can use information associated with each exam and a set of rules to predict which workstations will be used to read the exam. The images can thus be sent automatically before the physician arrives at the workstation to interpret a series of exams. The second scheme, which is intimately associated with the first, allows a workstation to manage its own local disk to remove copies of exams so that new ones may be pre-loaded. This disk management algorithm assigns priorities to the exams based on their status in the acquisition/interpretation cycle and performs automatic deletion as the workstation's disk reaches its capacity. The effect is a virtually limitless disk that eliminates the time-consuming task of manual deletion and retrieval of images.

Experience from ten years in Picture Archiving and Communication Systems has led to a clearer understanding of the user, system, and component requirements for PACS, and the evolving market. This paper reports on one manufacturer's considerations and experiences as a PACS supplier.

The Regional Medical Center at Memphis (the MED), including the Presley Memorial Trauma Center (the Trauma Center) is a privately operated, county owned, partially public funded institution that contracts with the University of Tennessee, Memphis (UT,Memphis) for physician clinical services. For approximately the past eight years, the administration of UT,Memphis has encouraged and emphasized the development of "Information Technology". To that end, a number of initiatives were developed on the campus, ranging from the development of a service oriented computer science department and the installation of an extensive broad band cable network to the widespread distribution of personal computers amongst the faculty. More recently, UT,Memphis and Siemens Corporation, the vendor of the UT,Memphis Magnetic Resonance Imaging (MR.I) equipment, collaborated in creating a link between the MRI and a physician workstation located in the Imaging Section of the MED. The MRI is located approximately a half mile from the MED and the link was established over the university broad band cable network. From those beginnings, interest developed within the Department of Raliology in more fully exploring the current PACS technologies. However, a number of resources often associated with the current installation of PACS technologies were lacking at UT, Memphis. No research funds were available from the university or from the Department of Radiology. Until very recently, and well into the final stages of planning for PACS, no significant biomedical engineering resources were available on the campus. The Physics Section of the Department of Radiology was overextended in providing support for a variety of clinical sections of the Department of Radiology. It became apparent that the most appropriate financial support for a PACS system was through the resources of one of the hospitals associated with UT,Memphis and that a major portion of the required technical support would have to come from the vendor and the hospital.

PACS is a topic which has not yet been granted a Seal of Approval by Radiology. One must only count the negative reports published to see that. A change of name from PACS to Image Management, Archival and Communication Systems (IMACS) will also not help the matter at the current stage of development. What is needed are clinical demonstrations of the philosophy behind PACS and of the fact that it works today and is not just an ugly duckling.

In this radiologic communication project two PAC-systems were installed in two locations of one radiologic department of the Rudolf Virchow University Clinic lying about 10 km apart. The realisation of a well-operating PACS-to-PACS communication is related to a fast image- and video- data transfer and a command processing technic. The PACS-workstation has to enable the manipulation and post-processing technics known from CT-, MR- and DSA- imaging.

This session really had its genesis sometime ago when Ed Staab and I had a discussion about the problems of evaluating a PACS. A lot of what we had seen had related to technical evaluation or observer performance studies, and we think those studies are extremely important, but it looked like a lot of the radiology community was interested in going beyond those and looking at PACS as a whole and how it might impact patient care With that thought in mind we began planning the session. We saw a really excellent study by David Gur and his colleagues yesterday on some of the observer performance issues, perhaps the ultimate expression of observer performance studies in terms of evaluating film or digitized film versus CRT displays. As I said, our goal in putting this session together was to try to get some additional insight into clinical evaluation beyond the methodology used for observer performance studies. And perhaps looking at some long range issues that could potentially affect the clinical evaluation of PACS. In that regard I invited, and was happy to receive acceptances from, three speakers. There will be three papers in this session. Unfortunately the last one did not get printed. Harold Kundel and Constantine Gatsonis have their subject titles listed. The third speaker will be David Beard from the University of North Carolina who will be talking about "Is a Work Station as Good as Film?" I think you are in for a very interesting, and I hope thought-provoking session. With that, I would like to introduce Harold Kundel, a radiologist and Professor of Radiology at the University of Pennsylvania, who will speak about the impact of PACS on patient care, efficacy, and outcome studies.

One of the key technologies in developing the picture archiving and communication system (PACS)is the method of presenting pictures. The pictures must be of sufficient quality to enable diagnosis and their presentation must be automatic so as not to interrupt doctors' concentration. The authors are developing a diagnostic work station (WS) which can solve these problems. In this study, we experimented on (1) a picture enhanement method for improving picture quality, and (2) a picture handling environment for simplifying operation. The results suggest that the YS should possess the intelligence to free doctors from complex operations.

The importance of providing voice communication between picture archiving and communication system (PACS) users has been realized by many researchers. In particular, voice communication can aid in the diagnosis of a patient by geographically separated radiologists and physicians. The voice dialogue can also be stored for later review. This work reports the results of a study to investigate the feasibility of introducing packetized voice to a particular PACS network. It is based on a multiplexed passive star local area network with wavelength-division multiplexing to provide separate logical channels for transfer of control and image data. It consists of an image network (INET) for image transfer at a rate of 140 Mbps, and a control network (CNET) operating at 10 Mbps, for controlling the flow of image transfers. The CNET employs the CSMA/CD protocol for bus arbitration. Previous simulation studies have shown that the CNET remains extremely underutilized even under conditions that overload the INET. This suggests that the CNET could be used to carry packetized voice as well as the control data currently carried. Use of the CNET for this purpose must be considered carefully, however, since the CSMA/CD protocol suffers long packet delays at heavy loads. At an early design phase, detailed modeling can be used to determine whether this will occur at various loads for this PACS. Stochastic activity networks provide the modeling framework necessary to carry out a detailed evaluation of this type. In this research, we represent the modified PACS network and the hypothesized voice, control, and image workload as stochastic activity networks. We verify that the network can support packetized voice in CNET while providing a reasonable service to control traffic.

When studying the functions to be supplied by a PACS the need for a link with the HIS/RIS turns out to necessary for two classes of reasons: - the first class deals with the need to supply as much clinical information at the workstations as possible. Whether this information is stored within the HIS/RIS should be transparant to the user. - modelling and simulation of PACS gives evidence that thetraffic load in a PACS can only be handled adequately when the image management system can take into account patient flow and patient medical story. As typical example can be mentioned prefetching algorithms that take care of activation of images from the archive before they are actually needed. The activation process is controlled by events in the hospital that indicate an increased probability that these images will be needed. The first class of reasons lead in general only to use of data that in principle can just be presented on the screen of the workstation. The PACS does not need to be aware of the meaning of the data: no common understanding is necessary. So just data can be transfered that will be presented in a transparant way. The second class of reasons deals with data from the HIS/RIS that should be understood by the PACS to be able to take the required action. So the meaning of the data is to be known. In this second class we have to deal with exchange of information while for the first class exchange just data is sufficient. Since the concepts of PACS and HIS are to a large extent similar it is not evident to which of the two certain functions will be assigned. This assignment of functions is considered and some suggestions for implementation of a link between PACS and HIS/RIS are given.

The Advanced Physicians Workbench (APW) is a radiology image management system display workstation built upon two generally accepted software packages, namely the X Windows System and the Informix database system. Two main purposes for the APW are: (1) to explore the feasibility of building a PACS display station for referring physicians by means of standardized application software development systems and (2) to investigate the use of individual, independent windows for image display and manipulation.

This paper provides a brief technical description of the Mayo/IBM Phase one PACS project. It covers additions to the Phase Zero system which transforms the system from a prototype archival system into an AS/400-driven, on-line, interactive, medical image review and archival system.

Recently we have introduced a method for analyzing Free-response Receiver Operating Characteristic (FROC) experiments. These differ from traditional ROC observer performance experiments by allowing for multiple abnormalities and observer responses per image, and by requiring correct localization for true positive events. We introduced the Alternative FROC (AFROC) plot which is analogous to the ROC plot, and the area under it, which is analogous to the area parameter of ROC methodology. Subsequently we have extended the analysis with a new way of scoring FROC images, termed AFROC analysis. This renders unnecessary the Poisson assumption used in the earlier paper. We also introduced the Free-response Forced Error (FFE) experiment which renders unnecessary the Gaussian distribution assumption of FROC and AFROC analysis. FROC/FFE experiments performed with synthetic images (disk objects at random locations on a noise background) which illustrate the FROC methodology are presented. We wish to bring to the attention of PACS researchers that rather sophisticated efficacy analysis, which has several advantages over the standard ROC approach, is made possible through this methodology.

We are developing a PACS module for thoracic radiology. The Chest PACS is composed of image acquisition subsystem, PACS seiver, display workstation, and communication network The system design is complicated by the clinicaldemandsfor high image quality on theprojectional images which implies a large amount ofdata to be transmitted, stored, and displayed. This paperpresents the details of the system design.

A prototype electronic alternator (EA) has been implemented utilizing an advanced image processing workstation and evaluated for its clinical acceptability as a primary diagnostic workstation. The workstation is based on a PIXAR II image computer and is a node on the University of Washington Digital Imaging Network and Picture Archiving and Communications System (DIN/PACS, hereafter PACS). The goals of this study are to utilize the unique features ofthe workstation (large image memory, a high-speed parallel transfer disk (PTD), and a 2560 x 2048 high-resolution image monitor) to demonstrate the feasibility of a limited model of the EA as a primary diagnostic workstation in the PACS environment, to evaluate its capability in image viewing and diagnosis, and to assess the feasibility of an icon-based user interface to select and rearrange images on the monitor. In this paper, various characteristics of the prototype EA such as image display and processing performance, image storage capacity, and functional specification as well as the results of a clinical evaluation are presented. The main emphasis of the clinical evaluation was on system speed, image quality, and user preference for an icon-based user interface.

A high performance ACR-NEMA computer interface board for the IBM PC/AT bus will be described. The interface board (AT/NSIF) implements the transport/network thru physical layers of the R-NEW standard. The interface board also supplies a session layer mapping function. The session layer mapping function, the transport/network layer protocol and the datalink layer protocol are implented via an on board microprocessor and resident firmware. The interface board makes use of a pair of gate arrays to implement the datalink frame checksum calculation and the physical layer protocol. The two gate array datalink/physical layer implementation will be described. One gate array interfaces the ACR-NEMA EIA-485 transceivers to high speed buffer memory. This gate array, the Datalink/Physical Layer Pipeline Processor (DLPP), is responsible for the ACR-NEMA physical layer word transfer protocol as well as the physical layer parity calculation. The DLPP also handles the datalink layer checksum calculation in real time. The second gate array, the Datalink/Physical Layer Controller (DLC), contains control and status registers for controlling and monitoring the DLPP. The PLC is also responsible for managing the CR-NEM physical layer interface arbitration. Performance measurements for the Datalink/Physical Layer Pipeline Processor will be detailed. AT/PASIF board level point to point communication performance measurements will be described.

A plan for the installation and implementation of a comprehensive computer-based system for the management and communication of digital radiographic images and diagnostic information is described. The paper is based on the authors' experience with, and evaluation of, prototype equipment and systems over the past 5 years. The final configuration will be realized in 1992 at the completion of a 3-phase installation plan. The system will address the clinical, data management, and administrative needs of the different types of users within the department, as well as the requirement to distribute radiographic information and images to locations outside of the department. In order to be considered successful, the system described herein will need to bring about a 90% reduction in both paper- and film-based communication of images and information. The British Columbia Ministry of Health is funding this phase of the project in order to obtain information on which to base decisions regarding installation of similar systems at other sites within the Province and predict with some confidence the cost effectiveness of such decisions.

A demonstration picture archiving and comniunication system (PACS) was developed using standard networking protocols. The systera was developed primarily to gain experience in digital image communications and optical disk usage as well as to raeasure perforittance factors to be used in the design of a substantially larger system. The system consists of a General Electric SIGNA MRI system connected via Ethernet to an IBM P5/2 Model 70, which serves as a gateway to a token-ring network. Resident on the token-ring, along with the P5/2, is an IBM AS/400 computer, an IBM System/36 with an optical storage subsystem and an IBM PS/2-based image review work station. Custom software was developed on each of the computers to control the flow of data and to manage the storage and retrieval of images. Standard TCP/IP software (File Transfer Protocol) is used to transfer the images from the MRI system to the P5/2. The images are converted into ACR/NEMA file format and made available to the communication and archiving system. Under AS/400 control, P5/2 resident programs copy a subset of the ACR/NEMA header information to an AS/400 database, and copy all of the image files to the optical storage subsystem for permanent storage. AS/400 programs were developed to enable database query operations. User-selected images are transferred from the optical storage subsystem either back to the MRI system or to the P5/2-based image review work station.

Recent PACS experiences have highlighted the critical aspects of PACS design. Three main topics are concerned: i) the communication issues, ii) the management of data and iii) the system setting-up and evaluation methodologies; whereas the first and the third aspects are widely presented in the literature, the second one is less discussed in papers although it may have important repercussions on the whole system design. In this paper, we deeply analyse the problem of access to image data from the functional, architectural and implementation viewpoints. The current trends investigated within the SIRENE environment are presented: the proposed solution combines a centralized management of archives with basic indexing and retrieval capabilities and a distributed management of applications on the workstations, offering a good flexibility for the development of specific applications suitable in the various medical departments.

A clinical trial of an Integrated Radiological Information System (IRIS) was conducted at the Ottawa Civic Hospital with the Department of Emergency Medicine and the Department of Radiological Sciences between April 4, and May 12, 1989. During the trial, 319 active Emergency Department cases (905 films) were processed using IRIS. Radiologists examined the digital images on the image screen to formulate a diagnosis, then before dictating a report, they examined the analog films. In 30 cases there was a discrepancy between the information obtained while viewing the digital images on IRIS and the information obtained from the analog films. These anomalous cases were used in an independent study of the discrepancies. In the study, each case was reviewed in both digital and analog form by three physicians who provided a comparative rating of diagnostic quality. Any perceived differences between the digital and analog media were noted. Particular attention was paid to rating the relevance of the IRIS enhancement capabilities. Although ratings for digital images were high, the comparative ratings for the film are in general better. An analysis of the individual cases shows that: (i) most of the discrepancies probably resulted from physician inexperience in reading radiographs in digital form, (ii) the IRIS enhancement facilities significantly increase the ratings of satisfaction or perceived quality of digital images and (iii) an appropriate choice of enhancement may make visible the required diagnostic features for cases where some reviewers did not find the image/digital discrepant.

PACS offers new capabilities, but presents some new limitations as well. We reviewed radiology practice at four medical centers. For interpretation of new images, most radiologists felt that the most recent examination for comparison would suffice, except for chest radiographs, for which three prior exams were frequently needed. All of the centers used alternators to review images, with from 5-14 separate reading areas. PACS capabilities should include organization and delivery of cases for interpretation and consultation to the appropriate area, pnoritization of cases, and rapid availability of inpatient and clinic patient data.

We have implemented a voice recognition interface using a Dragon Systems VoiceScribe-1000 Speech Recognition system installed on an AT&T 6310 personal computer. The Dragon Systems DragonKey software allows the user to emulate keyboard functions using the speech recognition system and replaces the presently used bar code system. The software supports user voice training, grammar design and compilation, as well as speech recognition. We have successfully integrated this voice interface in the clinical report generation system for most standard mammography studies. We have found that the voice system provides a simple, user-friendly interface which is more widely accepted in a medical environment because of its similarities to tradition dictation. Although the system requires some initial time for voice training, it avoids potential delays in transcription and proofreading. This paper describes the design and implementation of this voice recognition interface in our department.

BAZIS has set up an experimental imaging workstation with facilities for image input, storage, presentation and manipulation. A software package called FRACTALS (Filmless Radiology And Computerized Task Analysis Lab Software) has been developed for an IBM RT PC. An important feature of FRACTALS is a logging facility called computerized task analysis. Every time FRACTALS is started, a journal file is created. This file contains information on e.g. the images that has been displayed, options that has been invoked, window & level settings, time stamps. This information can be used to evaluate the functionality of FRACTALS and to perform research on human behaviour at radiology workstations. In this paper we discuss the porting of FRACTALS from the RT PC to other systems, using the X Window System. Furthermore we discuss the evaluation of the suitability of the RT PC input devices for a radiological workstation using the results of event logging.

Major two functions that a PACS workstation is considered to be equipped with are 1) efficient retrieval of image data and 2) supporting or consultation of writing reports, as radiologists have to diagnose increasing number of digital images in routine clinical studies. The authors developed a prototype PACS workstation with high speed image retrieving architecture and computer aided diagnosis and reporting function by using an artificial intelligence technology (AIPACS workstation). When physician selects the patient and his studies, the system performs feature extraction and generates diagnostic report by the inference engine with backward reasoning using the knowledge installed as production rules. Clinical application to the system for thyroid diagnosis showed good correlation with the diagnosis done by the physician.

Rewritable and compact media of magneto-optical disk (MOD) was firstly employed by us to store medical images such as digital X-ray films. The purpose is to test functions of communication media or circulating media as well as of filing media. Another feature of our trial system is to involve voice recording module for diagnostic reports. Method of clinical application and development approach of our PACS with MOD is discribed in this paper. Technology assessment of our system is also briefly descussed in terms of cost-effectiveness and security problem.

One of the reasons for the development of PACS is for use in medical image CAD(Computer Aided Diagnosis). At present, there are few possibilities of introducing image-processed CAD functions to PACS, because we have a number of technical problems with the development of image-processed CAD. However, most of these problems can be solved by improving the development environment of image processing. This paper explains the development environment of image-processed CAD introduced to PACS. The software used to improve the emvironment consists of three units; one is a device server, another is an image processing shell, and still another is an image control window. The major purpose of the improvement of the environment are to conceal device dependent data from users, to reduce programming work by using a hierarchical command system and to realize an easy-to-use environment with user's friendly MIII.

ISDN provides the end user with a large number of services and the flexibility to easily modify the services they receive as their communications needs grow and change. It is our goal to use the advantages of the ISDNtechnology for a medical end user by supporting him with integrated image/voice/data information. We focus our developments to medical scenarios where general PACS concepts are not acceptable, currently or in the near future. An image/voice/data management and communications system using ISDN technology has been designed. First applications are developed - to acquire multiple image modalities such as CT, Angio and NI from several hospital departments, - to communicate with a remote image processing group (Computer Science Department), - to complement visual results wIth voice and data yielding integrated multimedia information, and - to distribute integrated multimedia information to medical end users such as surgeons and therapists. The integrated multimedia information is finally available at a specially configured PC-workstation in complete digital form. An interactive replay of this information complies with a video-standard (either NTSC or PAL) and may discard specific media-modality on demand. At any time a video copy of the complete or constituent integrated information can be saved on VHS-cassettes for further distribution. The image processing group currently creates standardized video clips for frequent surgical scenarios. This paper describes the network model for the in-house and remote ISDN communication of integrated multimedia information, traffic, and performance requirements. The flexibility and acceptance of this approach will be addressed by reviewing several surgical cases.

The Rockwell Digital Imagery Standard (RDIS) format for images was developed in response to the need for an independent image format device. The Rockwell Digital Image Standard format has been designed to be device independent. The RDIS format releases technical staff from hardware limitations and allows the use of whatever image display system is available to them.

New medical image display devices or processes are commonly evaluated by anecdotal reports or subjective evaluations which are informative and relatively easy to acquire but do not provide quantitative nieasures. On the other hand, experinients eniploying ROC analysis, yield quantitative measurements but are very laborious and demand pathological proof of outcome. We have designed and are employing a new approach, which we have termed "agreement experiments," to quantitatively test the equivalence of observer performance on two systems. This was specifically developed to test whether a radiologist using a new display technique, which has some clear advantages over the standard technique, will detect and interpret diagnostic signs as he would with the standard display technique. Agreement experiments use checklists and confidence ratings to measure how well two radiologists agree on the presence of diagnostic signs when both view images on the standard display. This yields a baseline measure of agreement. Agreement measurements are then obtained when the two radiologists view cases using the new display, or display method, compared to the standard technique. If the levels of agreement when one reads from the new and one reads from the standard display are not statistically different from the baseline measures of agreement, we conclude the two systems are equivalent in conveying diagnostic signs. We will report on an experiment using this test. The experiment compares the agreement of radiological findings for chest CT studies viewed on the conventional multiformat film/lightbox to agreement of radiological findings from chest CT images presented on a multiple screen video system. The study consists of 80 chest CT studies. The results were an 86% to 81% agreement between the two viewing modalities which fell within our criteria of showing agreement.

Multimedia database is a fundamental component in the Multimedia Radiology Information System. It provides mean by which different classes of documents (diagnostic reports, images, patient information, etc), involving several media (text, image, graphics, voice), are structured, represented, stored, manipulated and retrieved. In this paper, we first present the requirements for a multimedia database along with the data model used to describe the radiology information and procedures. Then, we describe our approach to the multimedia database design and architecture, and the corresponding prototype instance in terms of implementation and storage strategies. Finally, we give a conclusion with an overview of the research undertaken for enhanced multimedia data management

The paper presents an analysis of the different metal artefact generating effects in X-ray CT and discus3es several algorithmic approaches to reduce their influence. They all share a segmentation step in which the artefact-causing implants are automatically located in the more or less degraded original backprojected image and the subset of affected measurement data is identified. The original image is reconstructed from appropriately scaled ray data that permit a truthful display of implants and the proper assessment of their geometry. Algorithms discussed are simple data substitution by interpolation, data substitution by interpolation enhanced by reprojected high-contrast information and data correction by modelling of artefcat generating physical phenomena.

During the period 1986-1989, the first phase of a research project in the field of computer systems for storage, communication and display of radiological images, has been carried out. This project has been financially supported by the Dutch Ministery of Health Care (WVC). Such systems are internationallyreferred to with the acronym PACS (Picture Archiving and Communication Systems). The project was jointly performed by three partners: - The Utrecht University Hospital (UUH) - BAZIS, Central Development and Support Group HIS, Leiden - Philips Netherlands/Philips InternationalB.V., product division Medical Systems. This paper will give an overview of the experiences and results which evolved from the first phase of this research project. The conclusion is that the PACS concept is attractive, but that further research will be necessary to make it implementable on full-scale. Recommendations for the continuationof PACS research in the Netherlands are made.

Images stored in the central database of a picture archiving and communications system (PACS) must be identified and described with textual information such as patient name, exam procedures, and diagnostic reports. Most radiology departments already use a radiology information system (RIS) for departmental administrative functions, such as exam scheduling, film tracking, billing etc., so this information is readily available. Ideally, the PACS and the RIS should be combined into one complete system, which is not easily achievable due to costly investment of replacing the current RIS and slow clinical acceptance of the PACS. Alternatively, for a successful PACS operation using the existing RIS, the two systems should be linked together so that the PACS can freely retrieve the RIS information whenever the need arises. As a part of the PACS evaluation project, an interface system to link our RIS (DECrad) to our PACS (AT&T CommView) has been implemented using an external microcomputer system running UNIX. The interface system handles proprietary communication protocols on each side, translating the information from one format to the other. Although the interface system translated the RIS transactions and transferred them to the PACS, they were not always associated with the image data at the PACS. This paper presents the design concept and the implementation details of the interface system we have developed. The performance and the problem areas of the interface system are also investigated, as well as suggesting future directions for a better implementation.

This paper discusses the tasks that a Intelligent Diagnostic Radiology Workstation (IDRW) should be able to perform in addition to serving as an image review station. Included in our discussion are the types of data that an IDRW must handle. It is hoped that this discussion will lead to improved, friendlier, and more intelligent workstations which will be accepted by radiologists.

Within the framework of the Dutch PACS project (a cooperation in the PACS field of BAZIS, the Utrecht University Hospital and Philips Medical Systems) a coupling between a HIS and a prototype PACS was realized and evaluated during clinical practice. This one-way coupling is a first step towards a so-called Image Information System, which can be seen as (logically) one system instead of two (a HIS for storage and retrieval of administrative and medical data, and a PACS for the images). The BAZIS/HIS is an integrated HIS, containing about 70 subsystems including a RIS, which is now used in about 40 hospitals in the Netherlands. The main reasons for coupling/integrating a HIS and a PACS can be summerized as: a) the user need for both functional and information integration and b) the need to have access to HIS data for optimizing image database and network management in order to get acceptable response times. As follow-up to the paper presented in the 1989 conference this paper will focus on the experiences gained with this one-way coupling. The event-driven communication took place by sending messages from HIS to PACS in ACRNEMA format. From these experiences recommendations for a next phase: a two-way link between a HIS and a PACS will be given. The necessity for bidirectional datatransfer will be discussed. Directions for future research e.g. on prefetching strategies will be pointed out.

This paper describes the integration of a voice recording module in our existing PACS system at the UCLA Department of Radiology. This module consists of a personnal computer, speech signal digitizer, and a hard disk with appropriate amount of memory storage to handle a given number of reports. The system has the ability to record and store spoken reports in digitized form and play back an existing dictated report. Moreover, the voice module can be activated locally from its own keyboard or by receiving appropriate commands from two remote mini-computers which are used as review and diagnostic display workstations respectively.

The ACR-NEMA standard for communications in digital radiology has been a topic for much discussion lately. The standard was developed to facilitate the development of multivender products which can communicate using a point-to-point standard interface. In this paper we describe, the combined experiences of Toshiba, Radiology, and Electrical and Computer Engineering at the University of Anzona with the ACR-NEMA standard. We also offer a candid evaluation and critique of the standard, for consideration by implementors of the current standard. The paper summarizes a prototype development effort using a high speed fiber optic network and the ACR-NEMA standard interface. We discuss the overall architecture of the prototype and the software design based on the International Standard Organization (ISO) upper layers. The transport layer and below was implemented using a Matrix-developed ACR-NEMA interface board. The paper discusses the problems encountered in the application of the ACR-NEMA standard in a networking environment for Picture Archiving and Communications Systems (PACS). We conclude that the current form of the ACR- NEMA standard does not provide a workable foundation for the development of cost effective and efficient digital radiology interfaces, especially in a networked PACS. We also discuss the PACS environment beyond the current ACR-NEMA standard and make recommendations for the feature direction of the standard.

We previously developed an image reconstruction display for reconstructing images compressed by our hybrid compression algorithm. The hybrid algorithm, which improves image quality, applies Discrete Cosine Transform coding (DCT) and Block Truncation Coding (BTC) adaptively to an image, according to its local properties. This reconstruction display receives the compressed data from the host computer through a BMC channel and quickly reconstructs good quality images using a pipeline-based microprocessor. This paper describes a prototype of a system for compression and reconstruction of medical images. It also describes the architecture of the image compression processor, one of the components of the system. This system consists of the image compression processor, a host main-frame computer and reconstruction displays. Under this system, distributed processing in the image compression processor and the image reconstruction displays reduces the load on the host computer, and supplies an environment where the control routines for PACS and the hospital information system (HIS) can co-operate. The compression processor consists of a maximum of four parallel compression units with communication ports. In this architecture, the hybrid algorithm, which includes serial operations, can be processed at high speed by communicating the internal data. In experiments, the compression system proved effective: the compression processor compressed a 1k x 1k image in about 2 seconds using four compression units. The three reconstruction displays showed the image at almost the same time. Display took less than 7 seconds for the compressed image, compared with 28 seconds for the original image.

At present the management of radiological information is done manually in most hospitals. Images from the digital modalities (e.g. Computer- tomography scanners, X-ray machines) are converted into films and stored in a room. These films are then distributed manually and interpretation is done by reading the films through light boxes. With the advances in semiconductor technology and digital processing techniques it is now possible to automate these operations by using a digital system for archiving and communication of images. A Picture Archiving and Communication System (PACS), is a digital system for acquiring, storing, moving and displaying picture or image information .These systems are highly dependent on the new technology for mass storage of data on optical disks. In the medical environment, the ultimate goal of using digital images and PACS is tohave a radiology department without filmjackets 2,3 A typical PACS consists of a database of medical images, display stations, acquisition systems and the communications medium. There are two possible architectures for PACS, one is the distributed PACS where all the modules are linked together through a high speed communication medium, and the other is a centralized PACS where the acquisition nodes and the display stations are part of the system which comprises the image database 4. In this paper we present analytical queueing models of distributed PACS used in the medical environments. These models are for the image database, the picture viewing stations, and the communication channels.

A PACS research project under the name " FOUNDATIONS FOR A HOSPITAL INThGRATED PICTURE ARCHIVING AND COMMUNICATION SYSTEM" is currently conducted. The overall objective is the modular conception of an experimental knowledge-based medical information management and distribution system. Image/text document access and display will be based on an intelligent user interface for diagnostic image workstations. The concept will comprise the integration of PA.S with the functions of the Hospital Information System. It will be based upon standards and be realized by an interactive collaborationbetween 7 european research groups from universities, hospitals and industries. The multi-disciplinary groups will consist of engineers, computer scientists, medical doctors (both radiologists and clinicians) and physicists. The HIPACS project has a budget of ECU 1.3 million, of which 50% is provided by the Commission of European Communities (CEC) within the scope of the exploratory phase of the Advanced Informatics in Medicine (AIM) research programme. HIPACS is directed by Prof. M. Osteaux, the prime contractor. This paper first explains the backgrounds of the AIM program. Next, the objectives of the HIPACS project are reviewed. The cooperating partners and their (scheduled) research activities are listed. Finally, the continuation of HIPACS within scope of the main phase of AIM is discussed.

This paper discusses the performance aspects ofthe joint Mayo/IBM PACS project. It includes performance measurements made during Phase Zero, and the early stages of Phase One. It also considers the network traffic load distribution and bandwidth utilization for the final Phase One topology. Based on earlier performance measurements and projected load distributions, performance projections are made for the completed Phase One PACS network. Finally, the paper discusses the approach being taken by the department of Computer Science at North Dakota State University to develop a useful model of this PACS network.

The Cost Analysis Model for Digital Imaging Networks (CAMDIN) is a life cycle cost analysis model of the financial impact of PACS technology upon the radiology department at Georgetown University Hospital. The model (which uses the Javelin Plus economic modeling package with data from Georgetown University Hospital and AT&T's CommView system) follows the costs and revenues within radiology both with and without the PACS network. CAMDIN guides the user in building a model tailored to an individual radiology department. It allows for sensitivity studies and produces practical results in several levels of detail clearly and quickly. Sensitivity studies have been performed on this system to determine the relative weights of several factors of PACS implementation.

The aim of the teaching workstation project is to develop a radiology workstation useful to students, practitioners, researchers, and faculty as a radiology learning and research tool. The user will develop skills and experience in interpretation through the teaching file cases as well as an understanding of the usage and capabilities of the host PACS system. This document represents the specifications and design for an effective teaching workstation built upon the Advanced Physicians Workbench, also under development at Georgetown University Hospital.

A new method for the noiseless compression of medical images is described. The method uses the wellknown DPCM technique (i.e. , linear prediction) for decorrelating a given image. However, instead of encoding the pixels of the decorrelated image using a memoryless model as in the conventional method, a source model with several conditioning events (or contexts) is employed. The contexts are based on the horizontal and vertical components of the gradient in the given image as well as the predicted value of a pixel. The statistics under each context of the model are obtained adaptively. In order to encode the decorrelated image as an outcome of such a complex source model, the powerful arithmetic coding technique is employed. Experimental results show that the new method can compress typical medical images 20% to 30% better than the conventional method.

This report describes the refinement of the interaction between a CR system and PACS that has made the automatic distribution of images to specified locations possible. Those interactions between the CR and the PACS that were considered important to the acceptance of the data to be sent to the ICU are described. Display and user interface considerations that were important for physician acceptance of the unit are also discussed. System enhancements and future applications will be developed based on the possible implications of this technology in both Radiology and the ICUs.

We assessed the effect of high background light on observer performance for the detection of a variety of chest radiographic abnormalities. Five observers reviewed 66 digital hard copy chest images formatted to 1 1 x 14 inch size under two display conditions: 1) on a specially prepared 1 1 x 14 inch illuminated panel with no peripheral light and 2) on a standard viewing panel designed for 14 x 17 inch radiographs. The images contained one - or more of the following conditions: pneumothorax, interstitial disease, nodules, alveolar process, or no abnormality. The results of receiver operator characteristic analysis show that extraneous light does reduce observer performance and the detectability of nodules, interstitial disease.

The design and implementation of a standard high speed ACR-NEMA communications interface is described. The upper layers e.g. the Presentation layer, Session layer and part of the Transport/Network layer have been implemented in software. In order to reach the speed requirement of 8M byte/sec. the lower layers e.g. part of the Transport/Network layer and Data Link layer have been implemented in hardware. We have developed and built an interface for an IBM personal computer P5/2 model 50, working under the operating system OS/2. The PS/2, model 50 has been equipped with a fast micro-channel bus, which enables a large throughput. The operating systern OS/2 has a multitasking capability, which enables concurrent programming. In order to minimize the delays, we used this multitasking facility to create a number of parallel operating "threads". The Transport/Network layer functions have been implemented using a receive thread, two send threads and a device driver with three hardware registers. The time to transfer a packet by DMA, to initiate the DMA logic and to execute the required Kernal functions have each been measured and figures are shown. The Data Link layer provides for storage of two packets in two separate random access memories (RAM's). These two RAM's enable a pipelined operation, which minimizes the delay in the Data Link layer.

This paper discusses to which extent manipulation functions which have been applied to images handled in PACS should be documented. After postulating an increasing amount of postprocessing features on PACS-consoles, legal, educational and medical reasons for a documentation of image manipulation processes are presented. Besides legal necessities, aspects of storage capacity, response time, and potential uses determine the extent of this documentation. Is there a specific kind of manipulation functions which has to be documented generally? Should the physician decide which parts of the various pathways he tries are recorded by the system? To distinguish, for example, between reversible and irreversible functions or between interactive and non-interactive functions is one step towards a solution. Another step is to establish definitions for terms like "raw" and "final" image. The paper systematizes these questions and offers strategic help. The answers will have an important impact on PACS design and functionality.

One of the touted potential benefits of Picture Archive and Communications Systems (PACS) is that these systems will save referring physicians significant time by eliminating their trips to andfrom the radiology departmentfile room. To date, this potential savings has not been quantified in any hospital setting. Economic modeling of PACS by the University of Washington demonstrates that the cost savings of PACS is extremely sensitive to small productivity changes in referring physician time savings. To provide analytical strength to this argument and supportfor the overall importance of the referring physician in PACS economics, trip distribution data was collected at the University to determine the average time referring physicians spend traveling to andfrom the radiology departmentfile room. This travel time is significant and it corroborates other work by the authors suggesting the benefits of PACS may be, in fact, larger outside the radiology department than inside it. Information for this study was obtained from two sources. In one case, referring physicians were queried about the time they and their support staffspend requesting and retrievingfilmfilesfrom thefile room by using a comprehensive hospital-wide survey. In the second case, the distance between all major clinics in the University medical center and the number offilm files each clinic checks out weekly were entered into a conventional transportation trip distribution model toforecast the time staffspend on this activity. While some differences in the results occurredfrom the two cases, each case generally showed a significant expenditure of effort approaching approximately two-weeks per year per referring physician with a potential recovery value of between $3 to $8 million dollars annually in additional billable revenue, assuming adequate patient demand.

One of the key determinants influencing how successfully a radiology department can convert from a conventional film-based environment to an exclusively digital imaging environment may be how well referring physician members of the hospital staff who are not radiologists endorse this new system. The benefits of Picture Archive and Communication Systems (PACS) to radiologists are becoming widely accepted and documented; however, physicians who interact with the radiology department represent an important user group whose views on PACS are less well understood. The acceptance of PACS by referring physicians (clinicians) may be critical to the overall utility ofPACS as well as a major drivingforce behind why a hospitalpurchases PACS. The degree to which referring physicians support PACS may be dependent upon many factors. This study identifies several aspects through the administration and analysis ofa survey which improve PACS acceptance by nonradiology physicians. It appears the more patients a referring physician sends to the radiology department, the more time a physician spends traveling to andfrom thefllmflle room retrievingfllms, and, the more interested a referring physician is about computers, the higher his interest is in PACS. If a referring physician believes that PACS will save him or her time, will reduce the incidence oflostfilms, or will cause performance of radiology exams or generation of reports to be more efficient, the referring physician appears more likely to support PACS and to make the initial time investment necessary to learn how PACS equipment operates. The factors which cause referring physicians to support PACS are principally: (1) the elimination oflost, misplaced, and checked outfllms, and (2) the elimination oftrips to and from thefile room. The major distractions ofthe technology are: (1) system reliability, and (2) reduced diagnostic capability. While the high cost ofPACS is also a distraction, it is not the predominant concern.

A purported benefit of digital imaging and archiving of radiographic procedures is the presumption of time savings to radiologists, radiology technologists, and radiology departmentpersonnel involved with processingfilms and managing theflimfile room. As part of the University of Washington's evaluation of Picture Archiving and Communication Systems (PACS)for the U.S. Army Medical Research and Development Command, a study was performed which evaluated the current operationalpractices of the film-based radiology department at the University of Washington Medical Center (UWMC). Industrial engineering time and motion studies were conducted to document the length of time requiredforfilm processing in various modalities, the proportion of the total exam time usedforfilm processing, the amount of time radiologists spent searchingfor and looking at images, and the amount of time file room personnel spent collating reports, making loans, updatingfilm jacket information, and purging files. This evaluation showed that better than one-half of the tasks in the file room may be eliminated with PACS and radiologists may save easily 10 percent of the time they spend reading films by no longer having to searchforfilms. Radiology technologists may also save as much as 10 percent of their time with PACS, although this estimate is subject to significant patient mix aberrations and measurement error. Given that the UWMC radiology department operates efficiently, similar improvements are forecast for other radiology departments and larger improvements areforecastfor less efficient departments.

Image Retrieval Expert System (IRES), a knowledge-based system for automatic image retrieval, is being prototyped at the University of Arizona (U of A). IRES is to couple with the distributed database system designed for Structured PACS (S-PACS)1 to achieve the high system performance required by radiologists. IRES encompasses the "intelligence" of multiple expert radiologists. The system will predict and migrate the "old" images needed for comparison purposes during radiological exam readings from slower or remote storage devices to the local buffers of workstations. The use of IRES with the PACS Distributed Database System (DDBS) is expected to shorten the PACS system response time, save the time of radiologists in selecting films, minimize the turnaround time of the exam interpretation function, and increase diagnostic effectiveness by providing relevant images automatically. This paper presents the implementation details of this IRES prototype.

The delivery of an operable Picture Archiving and Communication System (PACS) that will meet a hospital's high performance requirements is hampered by severe bottlenecks encountered in the database (DB) and communication systems of a centralized architecture. As an alternative to a centralized design, a distributed DB design that physically allocates and potentially replicates data across a hospital has been modeled and evaluated for PACS at the University of Arizona. Because of geographic dispersion of image generation and "local interest" in image retrieval, a distributed DB design approach has a liatural appeal to PACS. This paper examines the issues of a distributed PACS DB design, describes our modeling approach, and presents the results of a preliminary performance evaluation of two distributed DB designs vs. a centralized design for PACS The results of our simulation experiments show that there is a significant improvement of response time in a distributed architecture due to the effects of "local reference" and "load balancing."

The sound design of any system begins with a full understanding of user requirements and the environment in which the system is to operate. The process of obtaining this understanding is called requirement analysis. The design of a Picture Archiving and Communication System (PACS) Database (DB) also starts with a requirement analysis, following a systematic top-down DB design approach. The purpose of a DB requirement analysis is to provide the information necessary for deriving DB design specifications. This paper presents the results and implications of a year long intensive requirement analysis for a PACS DB at the University Medical Center (UMC) of the University of Arizona (U of A).

The design of PAC systems benefits from modelling and simulation. Within the scope of the Dutch PACS project several simulation studies have shown that simulation is a powerful tool to evaluate hardware architectures, image management strategies and clinical working methods before they are actually implemented. Besides using simulation techniques for obtaining insight into a PACS's performance and behaviour, they can also be applied to actively support the design of a PACS. In order to capture the full complexity of PAC systems in a simulation model, and to take full advantage of simulation as a design tool, we are currently developing new techniques for model construction. This paper summarizes the underlying principles of our modelling method, and describes our first experiences with it.

The University Hospital Utrecht (UHU) has been the site for clinical evaluation of a prototype Picture Archiving and Communication System (PACS), as a part of the Dutch PACS project run by BAZIS, Phffips Medical Systems and the UHU in the period 19864989 k The 800 bed UHU moved in July 1989 to a new campus site facifity (Figure 1.). This offered the opportunity to design the new radiology department to be prepared for a future large scale PACS. The set-up includes an ergonomically designed digital reading room, a centrally located computer room, cooling and cabling, a video network covering the total department and facilities for teleradiology. Radiology, Radiotherapy and Nuclear Medicine are organized in one department.

A PACS prototype has been installed and evaluated at the University of Washington. This paper presents the work done in the performance evaluation of the PACS prototype. The work involved network and workstation performance measurements and development of a simulation model based on the performance measurements. The simulation model was then used to do a parametric study of the PACS prototype to pinpoint the bottlenecks and suggest corrective measures. Results show that there are some local bottlenecks in the PACS prototype and an overall global bottleneck in the Data Management System (DM5) which forms the hub of the PACS prototype.

While the cost of Picture Archiving and Communication Systems (PACS) can be substantial, the cost of continuing with present manual methods may become prohibitive in growing departments as the need for additional space and personnel (both technical and professional) to meet the increasing requirements for all image management activities continues to grow. This will occur simultaneously with increasing pressures on problems of the present system, i.e., lost films, lost revenues, delayed reporting and longer diagnostic cycle times. Present methods of image archiving communication and management i.e. the relationship of procedure volume to VFE requirements for professional and technical personnel, costs of film, film storage space, and other performance factors are analyzed based on the database created by the Technology Marketing Group (TMG) computerized cost analysis model applied to over 50 US hospitals. Also, the model is used to provide the projected cost of present methods of film management for an average US 400 +bed hospital based on ten year growth rate assumptions. TMG PACS Tracking data provides confirmation of staffmg pattern correlation to procedure volume. The data presented in the paper provides a basis for comparing the investment in maintaining the status quo to an investment in PACS.

We present a prototype system for computer assisted stereotactic neurosurgery. It is especially suited to integrate vascular data with the stereotactic trajectory and with other information obtained from different imaging modalities. Patient image data from CT, MRI and DSA are acquired using a patient-fixed stereotactic frame with additional external markers for image registration. After preprocessing, all data are available for stereotactic planning and confirmation of electrode positioning. Interactive 3D devices simulate the electrode trajectory, which is projected on a stereoscopic set of angiograms. Alternatively, cross-sections of this trajectory with each of the CT or MR slices can be calculated at the same time. Additional features such as reslicing of the original CT or MR slices along the probe trajectory, are being implemented. All software is written on a 3D graphics workstation. In addition, we use a stereoscopic imaging system employing electro-optical shuttering glasses, and a 3D cursor in stylus form for the simulation of the electrode trajectory.

Once the technical difficulties of PACS have been solved, the main obstacle for its introductionwill be its costs. Even though PACS is still under development, it is clear that the equipment for PACS will be much more expensive than the equipment used in the conventional situation. PACS may, however, pay itself back by allowing savings of film, space and personnel. The conclusions of some recent cost-benefit studies disagree on the costs and benefits of PACS. Some conclude that PACS would pay for itself, whereas according to others PACS would be (much) more expensive. We analyzed the latest cost-benefit studies, to find out why their outcomes diverge. The comparison of the results revealed striking differences. The calculations of the annual costs of a hospital wide PACS varied between 2 and 4 million dollars. These differences could not be explained by differences in the size of the hospitals, as indicated by the number of examinations. They were, in part, caused by the fact that the costs per piece of equipment, material or space varied per study. For instance the costs of 1 square foot of (archive) space varied from $10 to $140 per year. The variation in the given costs of the film based system, depended on the fact whether the time spent by medical personnel on film management was taken into account. The differences among these studies demonstrate clearly the need for uniform, well-defined criteria for the calculation of the costs and savings of PACS.

An essential element in the decision of introducing a PACS will be its financial implication. We all hope PACS will come affordable at a certain point in time because of the decreasing hardware prices. However there is a wide variety of opinions about the moment the savings of PACS will exceed its costs. A software package for cost modelling of PACS, named CAPACITY, was developed by BAZIS in order to to get a better understanding of the cost characteristics of PACS. CAPACITY makes a cost comparison over the years thus indicating the moment of break even. User given input values are checked by a critique module and a file of cases can be maintained. One of the activities within the Dutch PACS project, a cooperative effort of BAZIS, the Utrecht University Hospital (UUH) and Philips Medical Systems is to indicate the costs of a PACS in the Utrecht University Hospital. A number of scenarios for a PACS in the Utrecht University Hospital have been defined and evaluated with the software package CAPACITY. The results will be discussed and it will be shown that the workstations play a major role.

The 422-bed Victoria General Hospital (VGH) and Siemens Electric Limited have since 1983 been piloting the implementation of comprehensive computerized medical imaging, including digital acquisition of diagnostic images, in British Columbia. Although full PACS is not yet in place at VGH, experience to date habeen used to project annual cost figures (including capital replacement) for a fully-computerized department. The resulting economic evaluation has been labelled hypothetical to emphasize that some key cost components were estimated rather than observed; this paper presents updated cost figures based on recent revisions to proposed departmental equipment configuration which raised the cost of conventional imaging equipment by $0.3 million* and lowered the cost of computerized imaging equipment by $0.8 million. Compared with conventional diagnostic imaging, computerized imaging appears to raise overall annual costs at VGH by nearly $0.7 million, or 11.6%; this is more favourable than the previous results, which indicated extra annual costs of $1 million (16.9%). Sensitivity analysis still indicates that all reasonable changes in the underlying assumptions result in higher costs for computerized imaging than for conventional imaging. Computerized imaging offers lower radiation exposure to patients, shorter waiting times, and other potential advantages, but as yet the price of obtaining these benefits remains substantial.

In the past three years, a clinical evaluation of a PACS has been performed in the Utrecht University Hospital as part of the Dutch PACS project. The clinical evaluation focussed on the following aspects: technical evaluation of the prototype PACS equipment coupled to the HIS; diagnostic accuracy studies; studies concerning the impact on the organization of the radiology-department and the referring wards; and cost-savings analysis. Some of the results of these subprojects have already been presented at previous SPIE conferences. In this paper the general condusions are presented about the usefulness of the evaluated PAC-System in the daily routine of radiology department and clinic. By making available the images of radiological examinations fast, complete, reliable and continously on the ward, concrete improvements with regard to the current process could be realized. The possibilities of PACS caused an increasing enthousiasm among the clinicians. By the easier access to all images of their patients during 24 hours/day, they saw more images on the day of the examination and images could be more easily used at consultations of other specialists. The overall conclusion is positive, but a lot of work has to be done to transform PACS from an experimental setup into a routine production system on which a flimless hospital can be based. A complete PACS needs an inteffigent Image Management System, which indudes prefetching algorithms based on data from the Hospital Information System and automated procedures for removing obsolete images from the local buffers in the workstations. As yet PACS is very expensive, and the direct savings in the hospital cannot compensate for the high costs of investment. Possibly PACS can contribute to a shorter stay of patients in the hospital. This will lead to savings for government and health insurance companies and they can be expected to contribute to PAS implementation studies.

The Radiology Department at the Hospital of the University of Pennsylvania is currently expanding its prototype Picture Archiving and Communications System ( PACS ) into a fully functional clinical system. The fi rst phase of this expansion involves three major efforts: the upgrade of the 10-Mbit token-ring to an 80-Mbit backbone with associated subnets, the implementation of a large-scale image archive, and, an interface between the PACS and the Department's Radiology Information System. Upon completion of this phase, the PACS will serve the storage and display needs of four MRI scanners and four of the Hospital' s Intensive Care Units . To achieve these objectives the Department entered into a joint development agreement with Vortech Data Inc. to develop applications-software for an optical jukebox-based image archive. This paper describes the developmental effort with special reference to the Image Archival and Retrieval System ( IRS). The IARS has three major features: a robust ACR/NEMA style protocol for external communication, a hierarchical storage system that incorporates magnetic, optical and shelf storage systems, and, a storage capacity that can be incrementally expanded from 25 GigaBytes to 7 TeraBytes.

For PACS to be clinically desirable, transmiued imagery must be optimized for the various clinical and research tasks. There are several processes that potentially degrade the diagnostic utility of the iransmitted imagery. We examined our CommView PACS system at San Francisco VA Medical Center (SFVAMC) to determine if imagery was altered from the original during passage into and through PACS, and whether the alteration had any discernable effect on current or projected clinical utility of the imagery. Related operational considerations were also examined. This limited review of our CommView system indicates that the attenuation of the image data by frame-grabbing and inability of the workstation to work with processed rather than formatted data from computed radiography limits the potential capabilities of the workstation. The reversible compression algorithm for imagery archival functioned well, but serious high frequency aliasing was introduced in imagery reduced for transmission to remote viewing stations.

The signal to noise ratio and the number of useful bits in data derived from laser film digitization has been assessed for different film densities. The spot size of 2lOjim was not modifiable, and films were digitized to l2bits of grey scale. It was found that the useful dynamic range was typically less then 256, and that the least significant bits were priitiarily noise. This study was part of a series of tests that have been perfonned using resolution patterns, low contrast objects and clinical test series, from which it appears that noise needs to be perceptible in the image for good perfonnance in detecting subtle features. Results from the use of the low contrast phantom, comprising 'lesions' of size well above system resolution but with contrasts ranging from greater than to less than the noise level, indicated that reading from the digital display gave better results than reading from film. It is believed that this gain in contrast perception resulted from the use of an appropriate window setting on the digital display such that noise was clearly visible.

Observers participating in ROC studies are usually required to estimate the confidence with which each observation is made. With a discrete scale, the rating, or score, normally falls into one of 5categories, ranging from 'definitely normal' to 'definitely abnormal'. However, a major problem in data analysis from ROC studies has been found to be caused by observers who have not used the rating scale in a uniform manner, and have made many responses corresponding to the two extreme categories with few responses falling in the middle. The use of a continuous rating scale, with a point selected using a mouse, has assisted in analysis, but only to a limited extent. It has therefore been suggested elsewhere that it is desirable to force observers to select intermediate points. The effect of such an approach on ROC curves was studied by asking a group of observers to re-score a set of difficult clinical images, after training and with continuous feedback on their compliance. Although the resulting fall in the ROC curves was not statistically significant, it is considered unwise to force observers to report in what to them appears to be an unnatural manner.