A complete PACS must store images for decades. For reasons of either media life or drive obsolescence, no storage technology can practically retain data for much more than ten years. Procedures for migration of data to new media must be part of the system to provide the necessary storage life. In preparation for procurement of a PACS archive, clinical procedure volume and use patterns in the Department of Radiology were analyzed and projections of storage, network and retrieval requirements were developed. The analysis showed that an archive based on magnetic disk for short term storage and duplicate copies of high density magnetic tape for long term storage should meet the complete archiving needs of the Department. Advance notice of the need for retrievals from long term storage must be provided by the Radiology Information System for efficient functioning of the archive. The organization and data migration policies of the archive being procured are described.

The ability to display digital radiographs to a radiologist in a reasonable time has long been the goal of many PACS. Intelligent routing, or pre-fetching images, has become a solution whereby a system uses a set of rules to route the images to a pre-determined destination. Images would then be stored locally on a workstation for faster display times. Some PACS use a large, centralized storage approach and workstations retrieve images over high bandwidth connections. Another approach to image management is to provide a high performance, clustered storage system. This has the advantage of eliminating the complexity of pre-fetching and allows for rapid image display from anywhere within the hospital. We discuss the development of such a storage device, which provides extremely fast access to images across a local area network. Among the requirements for development of the image server were high performance, DICOM 3.0 compliance, and the use of industry standard components. The completed image server provides performance more than sufficient for use in clinical practice. Setting up modalities to send images to the image server is simple due to the adherence to the DICOM 3.0 specification. Using only off-the-shelf components allows us to keep the cost of the server relatively inexpensive and allows for easy upgrades as technology becomes more advanced. These factors make the image server ideal for use as a clustered storage system in a radiology department.

In order for PACS (Picture Archiving and Communications System) to better serve our intensive care units (ICUs), we, at University of California, San Francisco, have designed and developed a client/server application that is specifically tailored to provide fast, reliable access to our PACS data from diagnostic viewing stations in the ICUs. One of our utmost design criteria is to ensure consistent delivery of high speed, high performance data throughput, and yet, the system should be cost-effective and render minimal maintenance. As high technology advances, we are able to utilize powerful mass storage device such as raid disk, which serves as a central image repository, to store images and data. We are also able to utilize Asynchronous Transfer Mode (ATM) technology, which is regarded as the prevailing technology for reliable, high speed data communications, to transfer large imagery data sets across systems and networks. This paper describes the design and mechanism of how ICU viewing stations take advantages of sharing a high performance raid disk, and ATM technology in data transfer for timely delivery of images in a clinical setting.

A large distributed image archive has been designed and is currently in the final stages of development as part of the Picture Archiving and Communications Systems at the University of Florida. The problem was to create a large yet efficient image archive to store the images from all the currently supported imaging modalities with expansion capabilities for the future. The design consists of modality acquisition processors (for the receipt, conversion and routing of image data from the diagnostic imaging equipment), medium term organ based magnetic archives (capable of storing patient studies for 4 - 6 months), a 3 Terabyte mixed media long term archive (currently using Digital Linear Tape), and a centralized image management data base. While the archive was primarily designed to support DICOM, certain legacy image formats such as ACR-NEMA 2.0 and native mode nuclear medicine formats are also supported until DICOM compliant acquisition is available from all imaging modalities.

Publications of the International Society of Optical Engineering (SPIE) contain much of the relevant literature on Picture Archiving and Communications Systems (PACS) and related topics. In fact, many PACS-related articles indexed by the National Library of Medicine contain references to articles published by SPIE. Unfortunately, SPIE publications themselves are not indexed by the National Library of Medicine and thus can not be identified through Medline. The lack of a convenient mechanism for searching the SPIE literature is problematic for researchers in medical imaging. With the recent introduction on SPIE's Internet server of their Abstracts Online service and their In-Cite^TM title and author searching software, the SPIE literature has become more accessible. However, the searching process is still a cumbersome and time consuming process, and it is not possible to perform key word searches of manuscript abstracts. In this paper we present results of our work on developing a mechanism to more thoroughly search SPIE publications for PACS-related articles.

The ATM-based Metropolitan Area Network (MAN) of Berlin connects two university hospitals (Benjamin Franklin University Hospital and Charite) with the computer resources of the Technical University of Berlin (TUB). Distributed new medical services have been implemented and will be evaluated within the highspeed MAN of Berlin. The network with its data transmission rates of up to 155 Mbit/s renders these medical services externally available to practicing physicians. Resource and application sharing is demonstrated by the use of two software systems. The first software system is an interactive 3D reconstruction tool (3D- Medbild), based on a client-server mechanism. This structure allows the use of high- performance computers at the TUB from the low-level workstations in the hospitals. A second software system, RAMSES, utilizes a tissue database of Magnetic Resonance Images. For the remote control of the software, the developed applications use standards such as DICOM 3.0 and features of the World Wide Web. Data security concepts are being tested and integrated for the needs of the sensitive medical data. The highspeed network is the necessary prerequisite for the clinical evaluation of data in a joint teleconference. The transmission of digitized real-time sequences such as video and ultrasound and the interactive manipulation of data are made possible by Multi Media tools.

Asynchronous transfer mode (ATM) technology has been implemented within our radiology department's hospital-wide PACS as well as in a wide area network (WAN) connecting affiliated hospitals. This paper describes our implementation strategies and the network performance observed in a clinical setting. The image communication network for our PACS is composed of two network interfaces: ATM (OC-3, 155 Mbps) and Ethernet (10 Mbps). This communication network connects four major campus buildings and two remote hospitals, providing intra- and interbuilding communication for radiologic images including CT, MR, CR, US, and digitized screen-film images. The network links these modalities via their acquisition computers to the PACS controller and to display workstations. The ATM serves as the primary network for transmission of radiologic images and relevant data within the entire PACS. The standard Ethernet is used as a backup network for ATM. It interconnects all PACS components including radiologic imaging systems, acquisition computers, display workstations, the PACS controller, the database servers, and the RIS and HIS. Our communication network operates on a 24 hrs/day, 7 days/week basis. Performance of the ATM network was evaluated in terms of disk-to-disk, disk-to-memory, and memory-to- memory transmission rates. The average memory-to-memory transmission rate over the wide area ATM network was 8.3 MByte/s, which corresponds to transferring a 40-slice (or, 20- MByte) CT examination to a remote site in less than 3 seconds. With the emerging ATM technology, we believe that ATM-based digital communication network is a suitable choice for large-scale PACS involving both LAN and WAN.

As image transmission becomes a more important part of the way radiology departments operate, the need for a high speed network infrastructure has become more important. We have installed a high speed network in the department that uses the latest Asynchronous Transfer Mode (ATM) networking technology combined with Ethernet switching. This network combination is capable of handling a tremendous amount of data traffic while maintaining compatibility with the existing Ethernet environment. These network changes have significantly improved Ethernet throughput on some of the most heavily used segments of the network by effectively isolating common traffic onto different network segments using new network management software and capabilities that are the result of the ATM backbone. Additional capabilities have allowed us to provide a number of serves that would not have been available using older techniques and architecture. Careful planning of the network before any new installations or changes is important for overall network and traffic management. The installation of this high speed network has allowed us to make imaging within the department and throughout the Medical Center and the connected region a reality.

The purpose of this study was to assess the effect of ATM network capabilities on the clinical practice of regional teleradiology, by providing immediate interactive radiology consultations between subspecialists and general radiologists at affiliated academic institutions. PACS installed at three affiliated hospitals (UCLA Medical Center, West LA VAMC and UCLA Olive-View Medical Centers) were connected via an ATM network. Two commercial PACS (Agfa) systems, one at the VAMC and one in an ultrasound outpatient clinic at UCLA were connected via ATM switches (Newbridge, Inc.) and a Santa Monica GTE central office switch. We evaluated this initial system configuration and measured image transfer performance, including memory-to-memory, disk-to-disk, disk-to-archive with and without DICOM protocols. Although the memory-to-memory data rate was 25 Mbps, the average remote disk-to-disk image transfer performance, using DICOM 3.0 communications protocols on SUN SPARCstation 10 servers, was 3 to 5 Mbps. Using these capabilities, timely interactive subspecialty consultations between radiologists was successfully performed while both were at different physical locations. We present the use of ATM technology in a realistic clinical environment and evaluate its impact on patient care and clinical teaching within the radiology departments of 2 institutions. Image communications over a regional PACS using an ATM network can allow interactive consultations between different subspecialist and general radiologists or other specialized radiologist spread over three different medical centers.

Current DICOM establishes connectivity between devices of multiple vendors defined by the storage and query/retrieve service classes. In today's IMAC systems imaging modalities are only sub-optimal integrated with other systems in the network. There is a need for higher integration of modalities with information systems like RIS and HIS. There is also a need for a higher integration of modalities with IMAC (PACS) systems. These requirements can be gained by the evaluation of dozens of Siemens PACS installations worldwide. It is not sufficient only to implement highly advanced functionalities into the PACS components, also imaging modalities and information systems, where most of the operator interactions takes place, must be integrated into an overall system based on common standards. CEN (Comite Europeen de Normalisation) TC 251/WG4 in cooperation with ACR-NEMA WG VI installed two project teams of experts working on the items mentioned above. The results were 2 DICOM/MEDICOM supplements: Supplement 8: `Storage Commitment Service Class' and Supplement 10: `Basic Worklist Management Service Class (Modality Worklist SOP Class)'. Additionally, technical reports were created, e.g. `VIEWS', a paper outlining future concepts for image management. A step towards higher integration and interoperability is achieved with the 2 DICOM supplements. Modalities can efficiently retrieve patient and procedure information from the information system. The operator on the modality can select scheduled procedure steps. The images sent to IMAC systems contain valid identification data. The additional protocol defined by the storage commitment ensures and double-checks, that the images of each procedure are reliably transferred to the IMAC system.

The Bowman Gray School of Medicine of Wake Forest University (BGSM) is installing clinical DICOM workstations and DICOM film digitizers for use in a telemedicine trial over the North Carolina ATM infrastructure. Acquiring equipment thought to be DICOM compliant, but which does not live up to expectations and possibly fails to interoperate with other DICOM equipment, is a concern. In cooperation with RSNA and the Mallinckrodt Institute of Radiology of the Washington University School of Medicine (MIR), BGSM made acceptance of the CEMAX workstations for the telemedicine trial contingent upon a satisfactory test with the MIR central test node (CTN). During the test, both engineers never left their hospitals during the tests drastically cutting travel and per diem costs and valuable time. The successful results exceeded expectations. Images were not only passed back and forth to the CTN by both the CEMAX clinical workstation and the quality assurance workstation, but images arriving from the CTN were entered into the CEMAX clinical database without operator intervention. It was expected that the images would transfer, but the placement into the database, although reasonable to expect, actually occurred. This live test from an independent test source exhibited the ease with which the CEMAX workstations handled the DICOM images in their systems. In the test, the CEMAX workstations were configured as an exporter of DICOM images at BGSM. The CTN was entered in the configuration of the CEMAX workstations as a source of DICOM images. The test was to ensure that plain film, computed tomography, and magnetic resonance images could be easily exchanged between workstations electronically. This paper describes the procedure for using the CTN to provide preliminary testing of DICOM equipment.

The Radiological Society of North America (RSNA) has sponsored demonstrations of the DICOM Standard. The technical details of the demonstrations were initially defined by the MedPACS section of NEMA beginning in 1992 and in subsequent years by RSNA. The Mallinckrodt Institute of Radiology was awarded a series of contracts by the RSNA to provide Central Test Node (CTN) software at the RSNA annual meetings to help vendors demonstrate interoperability. A goal of the early demonstrations was to have vendors communicate with one implementation of the Standard to minimize the amount of testing that needed to be performed prior to the RSNA Annual Meetings. Now that the Standard is better understood and implementations are more mature, vendors are offering more commercial products that utilize the Standard and which communicate with equipment manufactured by other vendors. The DICOM Standard is a technical specification which defines interfaces between equipment. These interfaces are currently defined for network, point to point and disk media. As a technical specification, the Standard does not tell the reader how to design an implementation nor does it provide insight into the thought process of the writers. This paper describes some of the implementation questions and issues that we have seen while working with the CTN software and vendors during the last four years. Examples of problem areas include application entity titles, generation of DICOM unique identifiers and queries for the query/retrieve service classes. The DICOM Standard defines the message formats for association negotiation and exchange of data across the network but provides flexibility in how these messages are used. A specific example is that an application is allowed to propose every service class defined by the Standard when requesting an association but may choose to exchange messages using only one service class once the association is established (for example, storage of CR data). We captured messages sent by vendor applications to the CTN during the 1995 RSNA DICOM demonstration. This paper will discuss some of the ways that vendors have chosen to propose associations and how they utilized the Standard to perform query/retrieve operations.

The DICOM services for patient, study and results management are recent additions to the ACR-NEMA standard. Commercially available RIS systems and the HL7 standard do not match the DICOM state transition models for these management tasks well. The purpose of this effect is to model the interaction between a DICOM-compliant PACS and an RIS, based on the DICOM management models, to provide a basis for standards-based PACS-RIS integration. We report on a state transition model for a RIS that correlates both to the HL7 trigger events usually supported and to the major features of the various state transition models incorporated in DICOM.

In this paper we discuss the implementation and use of a WWW interface to a DICOM PACS that allows users to select, move, and display images that are currently available in the PACS and to view their corresponding radiology reports. This system allows our users to query the archive from any workstation (such as Unix, DOS, and Mac) that supports a WWW browser. To use this system, the user first runs a WWW browser such as Mosaic, Netscape, or Lynx and specifies a URL on one of our Unix workstations. This URL refers to an HTML file that contains a query form. This query form contains a number of fields such as patient name and medical record number. The user may specify any or all fields as well as wildcards in fields such as the name field. Once the form is completed, the user presses a button to submit the request. The HTML form submits the query to a C program that executes on the Unix server. This program accepts as input the form field values that the user specified. This program then communicates with the archive via DICOM requests to determine those patients that match the search criteria. The user may then choose a patient which in turn causes the studies for this patient to be displayed. Finally, the user may select a study which causes those images to be retrieved from the archive and displayed via the Web browser. The result of this system is an easy to use interface to a DICOM PACS with the option to query and move images from the PACS. In summary, a system that integrates the ease of use of WWW browsers with a DICOM PACS is discussed. We are currently incorporating information from our RIS as well. This allows us to obtain extensive patient demographics, exam information, and textual radiological reports and associate this information with information from the PACS.

The purpose of this research and development effort was to solve several image management problems in Picture Archiving and Communications Systems (PACS). First, the patient and study information associated with images was not always correct and only rarely complete. This was due to human error in entering information on a console, and from incomplete data entry forms on image producing equipment. Second, in at least one area, Computed Radiography, the task of data entry was so time consuming that throughput suffered. The third problem was image routing, both within the Department of Radiology and throughout the hospital. The last identified problem was the lack of a common key between the PACS database the Radiology Information System (RIS) database so reports and images could be associated with each other. An RIS/PACS interface was developed in which RIS packets were sent to a PACS process at the time request forms were printed. These packets were parsed to various acquisition computers based on the modality type where they were stored in a MSQL Database for use in validating studies as they were completed prior to being transferred to PACS. DICOM header information from incoming studies were `matched' to a database entry based on the medical record number and modality. Whenever possible, an additional match was made on an accession number stored in the header. A match could result in the DICOM header being completed with detailed information about the patient, patient location, requesting service, and the procedure or study. In the case of the Kodak CR unit, patient and study information were sent directly to the CR workstation where they were accessible using a bar code interface at the time plates were ready to be processed. Routing within the radiology department was determined by comparing the RIS procedure code with an MSQL table to locate the workstation(s) used for viewing this type of study. The data of birth was used to determine whether the study should also be routed to a pediatric workstation. Finally, the accession number as assigned by the RIS was placed in the image header to allow matching of images and reports. The RIS/PACS system now matches patient, study, and other RIS information to PACS images to have improve routing, display, information accuracy, and efficiency. This system was built using a legacy RIS system without an HL7 interface, but the processes were created in a modular fashion that will make them easy to convert to HL7 when an expected new RIS is put in place. In order for a PACS to operate efficiently in an information intensive environment, the data associated with images must be correct, complete, and must contain `hooks' to other information systems in a medical center. The RIS/PACS interface is crucial to a successful PACS implementation.

We argue that efficient access to digital images for the Intensive Care Unit (ICU) requires the PACS to be strongly integrated within the overall working environment. This includes tailoring the imaging environment towards the specific working organization as well as integrating it with the HIS. The purpose of this project was to implement an integrated environment for chest images generated using a recently introduced phosphorplate scanner. PACS-HIS-Modality integration is reflected in our environment in different ways. The user interface of the viewing station is centered around often used patterns in ICU viewing. Information about bed occupancy is retrieved from the HIS and exploited in the viewing station. A digital connection between the phosphorplate scanner and the HIS ensures that images are correctly related to other patient information and to previous images of the patient. Using minor adaptations to the existing HIS, PACS and HIS have been made to cooperate in integrated presentation of images and reports. As work in progress, this integration has been expanded towards hypermedia, linking annotations in images to sections in a textual or spoken report. We discuss the relation between PACS and the global information environment, emphasizing organizational issues rather than technological aspects.

A PACS system can only contribute to the efficiency and quality of the operation of a radiology department unless it is an integral part of the workflow. This requires a thorough workflow analysis before deployment. Integration of the Radiology Information System (RIS) with the PACS is essential. DICOM is a necessary, but not a sufficient condition for the interaction between the various parts of the RIS and the PACS.

The paper describes various initial modeling steps for radiological departments, which take the data-flow, function-flow, and work-flow into consideration in different ways. Important contents of DICOM and HL7 are described and their possibilities and limitations to support overall managed workflow are analyzed. The architecture of SIENET, the Siemens PACS, its openness and integration capabilities are presented and illustrated in practical examples of typical data and workflow scenarios. In conclusion, managed workflow is possible today and specific parts are already successfully in operation. Here DICOM is the basis for far-reaching standardized communication in a heterogeneous landscape. For overall managed workflow, however, further efforts must be undertaken in modeling radiological activities as well as in the integration capabilities of various software applications and functions.

There exists tremendous opportunity in hospital-wide resource optimization based on system integration. This paper defines the resource planning and scheduling requirements integral to PACS, RIS and HIS integration. An multi-site case study is conducted to define the requirements. A well-tested planning and scheduling methodology, called Constrained Resource Planning model, has been applied to the chosen problem of radiological service optimization. This investigation focuses on resource optimization issues for minimizing the turnaround time to increase clinical efficiency and customer satisfaction, particularly in cases where the scheduling of multiple exams are required for a patient. How best to combine the information system efficiency and human intelligence in improving radiological services is described. Finally, an architecture for interfacing a computer-aided resource planning and scheduling tool with the existing PACS, HIS and RIS implementation is presented.

Sonography can be highly dependent on real-time imaging and as such is highly physician intensive. Such situations arise mostly during complicated ultrasound radiology studies or echocardiology examinations. Under those circumstances it would be of benefit to transmit real-time images beyond the immediate area of the ultrasound laboratory when a physician is not on location. We undertook this study to determine if both static and dynamic image transfer to remote locations might be accomplished using an ultrafast ATM network and PACS. Image management of the local image files was performed by a commercial PACS from AGFA corporation. The local network was Ethernet based, and the global network was based on Asynchronous Transfer Mode (ATM, rates up to 100 Mbits/sec). Real-time image transfer involved two teaching hospitals, one of which had 2 separate ultrasound facilities. Radiologists consulted with technologists via telephone while the examinations were being performed. The applications of ATM network providing real time video for ultrasound imaging in a clinical environment and its potential impact on health delivery and clinical teaching. This technology increased technologist and physician productivity due to the elimination of commute time for physicians and waiting time for technologists and patients. Physician confidence in diagnosis increased compared to reviewing static images alone. This system provided instant access for radiologists to real-time scans from remote sites. Image quality and frame rate were equivalent to the original. The system increased productivity by allowing physicians to monitor studies at multiple sites simultaneously.

The AT&T extended multimedia interface (EMMI) is being tested for applicability to wide area network access. Two EMMI units are used. One in the ultrasound examination room is connected to the National Television Standards Committee (NTSC) output of the ultrasound modality and audio communication equipment for the technologist. The other in the ultrasound reading room is connected to a NTSC monitor and audio communication equipment for the radiologist. The EMMIs are interconnected via an ATM permanent virtual circuit that establishes a connection between an ultrasound technologist and radiologist during a procedure. The test monitors 3 items: (1) How often can ultrasound studies be completed by a radiologist without on-site intervention, i.e., by viewing remotely only? (2) How often and on which exam does a technologist require direction remotely from a radiologist to complete a study? (3) How often and which exams require the radiologist to go to the ultrasound examination room to complete the study?

Transmission of high quality images between hospitals would be of value by exposing residents at individual institutions to a greater mix of disease processes. This problem is particularly serious in ultrasound where individual hospitals may not perform the entire range of examinations. We undertook this study to assess the effectiveness of image transmission via a PACS/ATM global network in improving ultrasound education among residents at affiliated hospitals. Image management was performed by AGFA PACS; global network was Asynchronous Transfer Mode. Selected cases from the two hospitals (OB/GYN cases at one, vascular at the other) were transmitted. Readout/teaching sessions included cases performed at base hospital and those received via network. Evaluation forms were collected from participants at both institutions. No image degradation occurred with transmission. Residents' exposure to ultrasound cases increased at the two hospitals. The system was considered an excellent teaching tool by all faculty and residents surveyed.

The authors have an in-kind grant from NASA to investigate the application of the Advanced Communications Technology Satellite (ACTS) to teleradiology and telemedicine using the Jet Propulsion Laboratory developed ACTS Mobile Terminal (AMT) uplink. We have recently completed three series of experiments with the ACTS/AMT. Although these experiments were multifaceted, the primary objective was the determination and evaluation of transmitting real- time compressed ultrasound video imagery over the ACTS/AMT satellite link, a primary focus of the author's current ARPA Advanced Biomedical Technology contract. These experiments have demonstrated that real-time compressed ultrasound video imagery can be transmitted over multiple ISDN line bandwidth links with sufficient temporal, contrast and spatial resolution for clinical diagnosis of multiple disease and pathology states to provide subspecialty consultation and education at a distance.

In a cooperation between the University Hospital Rotterdam (AZR) and HISCOM, HIS supplier, a multimedia information system for endoscopy (ENSIS) has been developed in a project funded by the Dutch Ministry of Health Care. An integral part of this project was an evaluation of costs and effects of this system. The system has been implemented on the gastroenterology department and the internal medicine ward in the AZR. The results indicate that the anatomical knowledge of requesting physicians improved with the system. Both the response time and availability of endoscopy images improved greatly. Because of the use of off-the-shelve technology (possible because of the relatively small resolution requirements of endoscopy images) ENSIS can be implemented at relatively low costs.

Current digital video archiving and networking systems are not well suited for high bandwidth echocardiography applications. We discuss an application of a structured hypermedia network system to archiving and management of digital echocardiography records. The server technology used in this project was based on Hyper-G information system. IT was demonstrated that a distributed echocardiography hypermedia record can be navigated hierarchically, by following hyperlinks (browsing) and by a search utility. A record integrity was preserved by strict enforcement of bi-directional linking between the objects within a domain. Distinct features of the discussed hypermedia network are patient centered hypermedia data model, link database and record input, access and retrieval facilities.

The purpose of this development effort was to evaluate the possibility of using digital technologies to solve image management problems in the Department of Radiology at the University of Florida. The three problem areas investigated were local interpretation of images produced in remote locations, distribution of images to areas outside of radiology, and film handling. In all cases the use of a laser film digitizer interfaced to an existing Picture Archiving and Communication System (PACS) was investigated as a solution to the problem. In each case the volume of studies involved were evaluated to estimate the impact of the solution on the network, archive, and workstations. Communications were stressed in the analysis of the needs for all image transmission. The operational aspects of the solution were examined to determine the needs for training, service, and maintenance. The remote sites requiring local interpretation included were a rural hospital needing coverage for after hours studies, the University of Florida student infirmary, and the emergency room. Distribution of images to the intensive care units was studied to improve image access and patient care. Handling of films originating from remote sites and those requiring urgent reporting were evaluated to improve management functions. The results of our analysis and the decisions that were made based on the analysis are described below. In the cases where systems were installed, a description of the system and its integration into the PACS system is included. For all three problem areas, although we could move images via a digitizer to the archive and a workstation, there was no way to inform the radiologist that a study needed attention. In the case of outside films, the patient did not always have a medical record number that matched one in our Radiology Information Systems (RIS). In order to incorporate all studies for a patient, we needed common locations for orders, reports, and images. RIS orders were generated for each outside study to be interpreted and a medical record number assigned if none existed. All digitized outside films were archived in the PACS archive for later review or comparison use. The request generated by the RIS requesting a diagnostic interpretation was placed at the PACS workstation to alert the radiologists that unread images had arrived and a box was added to the workstation user interface that could be checked by the radiologist to indicate that a report had been dictated. The digitizer system solved several problems, unavailable films in the emergency room, teleradiology, and archiving of outside studies that had been read by University of Florida radiologists. In addition to saving time for outside film management, we now store the studies for comparison purposes, no longer lose emergency room films, generate diagnostic reports on emergency room films in a timely manner (important for billing and reimbursement), and can handle the distributed nature of our business. As changes in health care drive management changes, existing tools can be used in new ways to help make the transition easier. In this case, adding digitizers to an existing PACS network helped solve several image management problems.

Two hundred examinations were tracked through the neuroradiology reading room. The median time from completing the exam to starting the interpretation was 1.6 hours for 76% and 18 hours for 24% of the cases. The delay occurred because exams done after 4:00 PM were held for the next day along with exams for which relevant prior studies were unavailable. The median times required for previewing, interpreting and consulting on a case were 7, 8 and 5 minutes respectively.

In this paper we present a specific example of the current state-of-the-art in desktop image access in the GI section of the Department of Radiology at the Hospital of the University of Pennsylvania. We describe a system which allows physicians to view and manipulate images from a Philips digital fluoroscopy system at the workstations in their offices. Typically they manipulate and view these images on their desktop Macs and then submit the results for slide making or save the images in digital teaching files. In addition to a discussion of the current state-of-the-art here at HUP, we also discuss some future directions that we are pursuing.

The productivity-improving features of a direct digital radiography system for projection radiography are introduced and the integration of the system with PACS is discussed. A flat panel digital array, an array controller, and a system controller with video monitor and standard interface to a local or wide area communications network are the main components of the direct digital radiography system of which prototypes have been built and tested in laboratory settings. When used in radiology room for projection radiography, the digital array converts x-ray photons into digital image data and makes the data available immediately for display on a video monitor for the technologist's review. Upon the technologist's acceptance of the image, an industry-standard network allows the transmission of the image to a workstation where additional image processing can be performed and where the image can be viewed on a high resolution display by the radiologist. If so desired, the image may be routed to a laser printer, digital mass storage medium, or to a PACS communication interface. By connecting the direct digital radiography system with a PACS network, radiographs may be shared electronically like images of other digital modalities. In this scenario, productivity improvements would come mainly from shorter patient exams as a result of the immediate availability of the review images, wider exposure latitude and ease of handling of the digital array, and from the electronic transport, storage and retrieval of image files and patient data in a PACS environment.

The first installation of the `STATVIEW Classic' image network system from E-Systems Medical was at William Beaumont Hospital in December 1993. The installation was intended to support a new critical care tower. Three hundred critical care beds are served by this system, with ten display stations. The system uses non-proprietary hardware for image display and the display software runs in Microsoft Windows. The system consists of a Lumisys model 150 laser digitizer with necessary control equipment, a file server consisting of a IBM compatible 486SX computer with 500 Megabytes of storage, and display stations based on IBM compatible 486DX (66 MHz) computers with NEC 6FGp monitors. We decided to expand the system by purchasing the new Statview DX system from EMED. In August 1995, installation of a DICOM 3.0 image network, the `STATVIEW DX' from E-Systems Medical, was completed. Considerations which were important during the installation of the system and the response of the medical staff to the image network will be discussed.

In previous studies of workstation usage by physicians in the Medical Intensive Care Unit (MICU) of the University of Pennsylvania Medical Center, we have compared which image manipulation functions were used when viewing digitized film and phosphor plate (CR) images. We had found some differences in the use of such features, particularly a decrease in the use of brightness and contrast adjustment, for CR images. We had predicted this because of the improved uniformity of CR images compared to conventional film [Horii 95]. We were interested in how radiologists viewing similar CR images would differ (if at all) from their MICU physician colleagues. We found that the radiologists used the brightness and contrast adjustments to a greater degree than the MICU physicians when viewing CR images from MICU patients. Use of other image manipulation functions by the radiologists was considerably lower; in this first series of readings, zoom, invert grayscale, and high-resolution display controls were not used at all.

Our purpose was to determine the economic effects associated with the introduction of PACS and computed radiology (CR) in a medical intensive care unit (MICU). Clinical and financial data were collected over a period of 6 months, both before and after the introduction of PACS/CR in our medical intensive care unit. Administrative claims data resulting from the MICU stay of each patient enrolled in our study were transferred online to our research database from the administrative databases of our hospital and its affiliated clinical practices. These data included all charge entries, sociodemographic data, admissions/discharge/transfer chronologies, ICD9 diagnostic and procedure codes, and diagnostic related groups. APACHE III scores and other case mix adjusters were computed from the diagnostic codes, and from the contemporaneous medical record. Departmental charge to cost ratios and the Medicare Resource-Based Relative Value Scale fee schedule were used to estimate costs from hospital and professional charges. Data were analyzed using both the patient and the exam as the unit of analysis. Univariate analyses by patient show that patients enrolled during the PACS periods were similar to those enrolled during the Film periods in age, sex, APACHE III score, and other measures of case mix. No significant differences in unadjusted median length of stay between the two Film and two PACS periods were detected. Likewise, no significant differences in unadjusted total hospital and professional costs were found between the Film and PACS periods. In our univariate analyses by exam, we focused on the subgroup of exams that had triggered primary clinical actions in any period. Those action-triggering exams were divided into two groups according to whether the referring clinician elected to obtain imaging results from the workstation or from the usual channels. Patients whose imaging results were obtain from the workstation had significantly lower professional costs in the 7 days following the action-triggering exam compared with patients whose exam information was obtained through the usual channels. No significant differences in patient age or APACHE score were noted between these two groups. Our preliminary results show possible evidence of a reduction in unadjusted cost of care for the subgroup of patients whose action-triggering exams were first encountered on the PACS workstation. Whether these results represent clinicians' differential workstation utilization patterns or true effects of PACS/CR will be determined by further data analysis, including case mix adjustment, subgroup analysis, and multivariate modeling.

A clinical comparison of computed radiography (CR) versus screen-film for imaging the critically-ill neonate is performed, utilizing a modified (hybrid) film cassette containing a CR (standard ST-V) imaging plate, a conventional screen and film, allowing simultaneous acquisition of perfectly matched CR and plain film images. For 100 portable neonatal chest and abdominal projection radiographs, plain film was subjectively compared to CR hardcopy. Three pediatric radiologists graded overall image quality on a scale of one (poor) to five (excellent), as well as visualization of various anatomic structures (i.e., lung parenchyma, pulmonary vasculature, tubes/lines) and pathological findings (i.e., pulmonary interstitial emphysema, pleural effusion, pneumothorax). Results analyzed using a combined kappa statistic of the differences between scores from each matched set, combined over the three readers showed no statistically significant difference in overall image quality between screen- film and CR (p equals 0.19). Similarly, no statistically significant difference was seen between screen-film and CR for anatomic structure visualization and for visualization of pathological findings. These results indicate that the image quality of CR is comparable to plain film, and that CR may be a suitable alternative to screen-film imaging for portable neonatal chest and abdominal examinations.

An assessment of changes in health-care professional behavior as a result of the introduction of a PACS (picture archiving and communication system) display station to an adult medical- surgical intensive care unit (ICU) is investigated via pre- and post-PACs evaluations. ICU display station utilization and the impact on clinical operations are also examined. Parameters measured both pre- and post-PACS ICU display station placement include the number of films per patient day, the number of clinician reviews of a patient's images per day and the percentage of images on which the unit interacts with a radiologist. The elapsed times from the time of exposure to the time of: review by the referring physician, radiologist-unit interaction and clinical action based on image information are also measured. The results of this investigation suggest that the introduction of a PaCS display station in the ICU may reduce the number of exams per patient day, decrease the elapsed time from the time of exposure to the time of review by the unit clinician, and improve the time to clinical action. Note, however, that it does not appear to change the percentage of total images on which the unit interacts with a radiologist.

The purpose of this paper is to present a control theory and a fault tolerance algorithm developed for real time monitoring and control of acquisition and preprocessing of computed radiographs for PACS and Intensive Care Unit operations. This monitoring and control system uses the event-driven, multilevel processing approach to remove computational bottleneck and to improve system reliability. Its computational performance and processing reliability are evaluated and compared with those of the traditional, single level processing approach.

In order to determine the effect of an image workstation, viewing patterns and related clinical actions were evaluated in a randomized prospective study. During 16 weeks of Computed Radiography data collection, an image workstation was conveniently available to the Medical Intensive Care Unit clinicians. The workstation was not available for clinical use during 16 weeks of Analog Film data collection. Viewing patterns were evaluated by comparing viewing times. Patient care was evaluated by comparing the time of performing image based clinical actions. The percentage of routine exams viewed before AM Radiology Conference increased from 0% during the Analog Periods to 27% during the CR PACS Periods. Clinicians selected images taken during the first few days of the patient's admission for viewing before conference. Images taken later in admission were viewed during or after conference. On days when radiology conference was not held, images were viewed significantly earlier when the workstation was available. Clinical actions based on images viewed on the workstation were performed significantly earlier. When an image workstation was available routine images were viewed sooner and image based actions occurred earlier.

Teleradiology offers significant improvement in efficiency and patient compliance over current practices in traditional film/screen-based diagnosis. The increasing number of women who need to be screened for breast cancer, including those in remote rural regions, make the advantages of teleradiology especially attractive for digital mammography. At the same time, the size and resolution of digital mammograms are among the most challenging to support in a cost effective teleradiology system. This paper will describe a teleradiology architecture developed for use with digital mammography by GE Corporate Research and Development in collaboration with Massachusetts General Hospital under National Cancer Institute (NCI/NIH) grant number R01 CA60246-01. The testbed architecture is based on the Digital Imaging and Communications in Medicine (DICOM) standard, created by the American College of Radiology and National Electrical Manufacturers Association. The testbed uses several Sun workstations running SunOS, which emulate a rural examination facility connected to a central diagnostic facility, and uses a TCP-based DICOM application to transfer images over a satellite link. Network performance depends on the product of the bandwidth times the round- trip time. A satellite link has a round trip of 513 milliseconds, making the bandwidth-delay a significant problem. This type of high bandwidth, high delay network is called a Long Fat Network, or LFN. The goal of this project was to quantify the performance of the satellite link, and evaluate the effectiveness of TCP over an LFN. Four workstations have Sun's HSI/S (High Speed Interface) option. Two are connected by a cable, and two are connected through a satellite link. Both interfaces have the same T1 bandwidth (1.544 Megabits per second). The only difference was the round trip time. Even with large window buffers, the time to transfer a file over the satellite link was significantly longer, due to the bandwidth-delay. To compensate for this, TCP extensions for LFNs such as the Window Scaling Option (described in RFC1323) were necessary to optimize the use of the link. A high level analysis of throughput, with and without these TCP extensions, will be discussed. Recommendations will be made as to the critical areas for future work.

We are developing a wide area test bed network using the Advanced Communication Technology Satellite (ACTS) from NASA for high speed medical image transmission. The two test sites are the University of California, San Francisco, and the National Library of Medicine. The first phase of the test bed runs over a T1 link (1.544 Mbits/sec) using a Very Small Aperture Terminal. The second phase involves the High Data Rate Terminal via an ATM OC 3C (155 Mbits/sec) connection. This paper describes the experimental set up and some preliminary results from phase 1.

Recent advances in telecommunications via the Internet have significantly improved international communication. We report a case of teleconsultation using the World Wide Web (WWW) and related technologies on the internet to facilitate the successful diagnosis and treatment of a rare illness that occurred in Beijing, China. This case report demonstrates some of the advantages and disadvantages of using such a system to gather, to retrieve medical information, and to facilitate communication among physicians in different countries. It also demonstrates international teleconsultation using the internet and WWW is possible and effective in this case which saved this young student's life.

This paper investigates the design and technical efficacy of an integrated PC based platform for three different medical applications. The technical efficacy of such a telemedicine platform has not been evaluated in the literature and optimal technical requirements have not been developed. The first application, with the Department of Surgery, Division of Urology, tests the utility of a telemedicine platform including radiology images for a surgical stone disease consultation service from an off site location in West Virginia. The second application, with the Department of Internal Medicine, Division of Clinical Pharmacology, investigates the usefulness of telemedicine when used for a clinical pharmacology consultation service from an off-site location. The third application, with the Department of Pediatrics, will test telemedicine for trauma care triage service first within an off-site location in Virginia and then from there to Georgetown University Medical Center.

A pilot project was developed to explore the role of subspecialty radiology support to rural medicine sites over a long-distance network. A collaborative relationship between 2 rural radiology practices and an academic health was established. Project objectives included: (1) Does the subspecialty consultation significantly change diagnosis patterns at the rural site? (2) Is there value added as measured by improved clinical care or an overall decreased cost of care? (3) Can a collaborative model be economically self-supportive? (4) Does the collaborative model encourage and support education and collegial relationships? Two rural hospitals were selected based on the level of imaging technology and willingness to cooperate. Image capture and network technology was chosen to make the network process transparent to the users. DICOM standard interfaces were incorporated into existing CT and MRI scanners and a film digitizer. Nuclear medicine images were transferred and viewed using a proprietary vendor protocol. Relevant clinical data was managed by a custom designed PC based Lotus Notes application (Patient Study Tracking System: PaSTS) (Pennsylvania Blue Shield Institute). All data was transferred over a Frame Relay network and managed by the Pennsylvania Commonwealth sponsored PA Health Net. Images, other than nuclear medicine, were viewed on a GE Advantage viewing station using a pair of 2 X 2.5 K gray scale monitors. Patient text data was managed by the PaSTS PC and displayed on a separate 15' color monitor. A total of 476 radiology studies were networked into the AHC. Randomly chosen research studies comprised 82% of the case work. Consultative and primary read cases comprised 17% and 1% respectively. The exercise was judged effective by both rural sites. Significant findings and diagnoses were confirmed in 73% of cases with discrepant findings in only 4%. One site benefited by adopting more advanced imaging techniques increasing the sophistication of radiology services. The primary value for the referring sites was the added confidence provided by the subspecialty overreads. An educational value was recognized by all. In conclusion, the networking of rural health care sites to an AHC subspecialty radiology practice was successful primarily in increasing the diagnostic confidence at the rural site. Other benefits included: education; increased rural imaging and an opportunity to provide primary interpretation when the rural radiologist is not available. However, the rate of rural generated consultation was low (17%) and is unlikely to support the costs of a high speed network. To support, rather than replace, rural radiology requires a lower cost network and a mechanism for payment for these services.

Telemedicina is a Spanish project that covers teleradiology and radiosurgery areas. This project is under the frame of the Spanish Broadband National Plan (PLANBA). The final technical tests are being ended over Ethernet and ATM and it is planned to get their first clinical results on February-96. Two pilots will be installed: one in Madrid linking two sites through a ATM network (provided by Telefonica, Spanish PTT) and a second one in Asturias (north of Spain) using ISDN primary access (2 Mbps). The system handles still images, voice and video records, scanned documents, text and slides allowing doctors to interchange this data using cooperative tools. The system is based on a multimedia Unix platform with voice, video and videoconference devices and boards. The platform will be used in several ways: as desktop videoconferencing, primary diagnosis and review. Communications are based on ATM (over AAL5) at 155 Mbps and ISDN (primary access). The protocol used in both networks is TCP/IP. The application is written in C++ (object oriented design and programming) and C. GUI is built under X-Windows and Motif. The codification of video is MJPEG done through dedicated hardware. The system is integrated in a small PACS (previously installed); the images are captured from the modalities such as CT using the DICOM standard and it is connected with the Radiological Information System. The application allows collaborative work: telepointer, shared windows, editors and actions. Main news of the Telemedicina project will be the incorporation of broadband networks (ATM at 155 Mbps) and the integration of collaborative work. This two aspects allow the doctors to improve their work speeding up the transmission and retrieval of medical records. Also this platform can be used to achieve several goals: such as primary diagnosis, videoconference, review.

This paper describes the current status of the second generation PACS at UCSF commenced in October 1992. The UCSF PACS is designed in-house as a hospital-integrated PACS based on an open architecture concept using industrial standards including UNIX operating system, C programming language, X-Window user interface, TCP/IP communication protocol, DICOM 3.0 image standard and HL7 health data format. Other manufacturer's PACS components which conform with these standards can be easily integrated into the system. Relevant data from HIS and RIS is automatically incorporated into the PACS using HL7 data format and TCP/IP communication protocol. The UCSF system also takes advantage of state-of-the-art communication, storage, and software technologies in ATM, multiple storage media, automatic programming, multilevel processes for a better cost-performance system. The primary PACS network is the 155 Mbits/sec OC3 ATM with the Ethernet as the back-up. The UCSF PACS also connects Mt. Zion Hospital and San Francisco VA Medical Center in the San Francisco Bay area via an ATM wide area network with a T1 line as the back-up. Currently, five MR and five CT scanners from multiple sites, two computed radiography systems, two film digitizers, one US PACS module, the hospital HIS and the department RIS have been connected to the PACS network. The image data is managed by a mirrored database (Sybase). The PACS controller, with its 1.3 terabyte optical disk library, acquires 2.5 gigabytes digital data daily. Four 2K, five, 1,600-line multiple monitor display workstations are on line in neuroradiology, pediatric radiology and intensive care units for clinical use. In addition, the PACS supports over 100 Macintosh users in the department and selected hospital sites for both images and textual retrieval through a client/server mechanism. We are also developing a computation and visualization node in the PACS network for advancing radiology research.

Three dimensional visualization of tumor and normal tissues are often valuable in precision treatment planning for radiation therapy of cancer. This is often not possible in many remote treatment facilities because of the high cost of imaging and computer equipment. At the University of Southern California, School of Medicine we are developing and testing a high speed, wide area, computer communications network to provide access to such resources from remote locations. In this paper, we present our concept of a Virtual Academic Medical Center, our network design and report on initial evaluations of effectiveness and clinical acceptability of 3D, CT simulation and treatment planning from a distance.

We have begun a project to implement an Electronic (Filmless) Radiology Practice (ERP) at Mayo Clinic Jacksonville. This project is integrated with the implementation of a project (Automated Clinical Practice--ACP) to eliminate circulation and archival of the current paper Medical Record. The ERP will result in elimination of screen/film radiography and the transmittal of film throughout the institution by the end of 1996. In conjunction with the ACP, paper and film will not circulate within the clinic by the end of this year.

Many studies analyzing the costs of film-based and filmless radiology have focused on multi- modality, hospital-wide solutions. Yet due to the enormous cost of converting an entire large radiology department or hospital to a filmless environment all at once, institutions often choose to eliminate film one area at a time. Narrowing the focus of cost-analysis may be useful in making such decisions. This presentation will outline a methodology for analyzing the cost per exam of film-based and filmless solutions for providing portable chest exams to Intensive Care Units (ICUs). The methodology, unlike most in the literature, is based on parallel data collection from existing filmless and film-based ICUs, and is currently being utilized at our institution. Direct costs, taken from the perspective of the hospital, for portable computed radiography chest exams in one filmless and two film-based ICUs are identified. The major cost components are labor, equipment, materials, and storage. Methods for gathering and analyzing each of the cost components are discussed, including FTE-based and time-based labor analysis, incorporation of equipment depreciation, lease, and maintenance costs, and estimation of materials costs. Extrapolation of data from three ICUs to model hypothetical, hospital-wide film-based and filmless ICU imaging systems is described. Performance of sensitivity analysis on the filmless model to assess the impact of anticipated reductions in specific labor, equipment, and archiving costs is detailed. A number of indirect costs, which are not explicitly included in the analysis, are identified and discussed.

A remote consultation system is available at the University of Pittsburgh Medical Center (UPMC) which links four outlying hospitals in Western Pennsylvania and Eastern Ohio. This system has the potential to improve short and long term clinical outcomes and to reduce overall medical care cost by establishing improved emergency triage capability. An EMED, Inc. teleradiology system permits rapid, high-quality transfer of digitized film and CT images from the remote sites to the tertiary care center (UPMC). The images are sent over dial-on- demand ISDN and SW56 lines from the remote hospitals to a central server where they are transmitted to a dual 2K monitor workstation in the Emergency Department, thirteen Eastman Kodak PDS workstations within UPMC, and to three physician homes. Transmission to a workstation at each of the physician homes over ISDN lines enables `after hours' consultation. The radiographic images along with voice and fax communications provide a technique where physicians in outlying hospitals will be able to consult with specialists at any time. A study is in progress to evaluate the effectiveness of this system in terms of perception of utility and its potential to improve emergency triage capability, as well as selection of the appropriate transportation mode (helicopter versus ambulance).

It is a well known fact that managed care and new treatment technologies are revolutionizing the health care provider world. Community Health Information Network and Computer-based Patient Record projects are underway throughout the United States. More and more hospitals are installing digital, `filmless' radiology (and other imagery) systems. They generate a staggering amount of information around the clock. For example, a typical 500-bed hospital might accumulate more than 5 terabytes of image data in a period of 30 years for conventional x-ray images and digital images such as Magnetic Resonance Imaging and Computer Tomography images. With several hospitals contributing to the archive, the storage required will be in the hundreds of terabytes. Systems for reliable, secure, and inexpensive storage and retrieval of digital medical information do not exist today. In this paper, we present a Medical Image Archive and Distribution Service (MIADS) concept. MIADS is a system shared by individual and community hospitals, laboratories, and doctors' offices that need to store and retrieve medical images. Due to the large volume and complexity of the data, as well as the diversified user access requirement, implementation of the MIADS will be a complex procedure. One of the key challenges to implementing a MIADS is to select a cost-effective, scalable system architecture to meet the ingest/retrieval performance requirements. We have performed an in-depth system engineering study, and developed a sophisticated simulation model to address this key challenge. This paper describes the overall system architecture based on our system engineering study and simulation results. In particular, we will emphasize system scalability and upgradability issues. Furthermore, we will discuss our simulation results in detail. The simulations study the ingest/retrieval performance requirements based on different system configurations and architectures for variables such as workload, tape access time, number of drives, number of exams per patient, number of Central Processing Units, patient grouping, and priority impacts. The MIADS, which could be a key component of a broader data repository system, will be able to communicate with and obtain data from existing hospital information systems. We will discuss the external interfaces enabling MIADS to communicate with and obtain data from existing Radiology Information Systems such as the Picture Archiving and Communication System (PACS). Our system design encompasses the broader aspects of the archive node, which could include multimedia data such as image, audio, video, and free text data. This system is designed to be integrated with current hospital PACS through a Digital Imaging and Communications in Medicine interface. However, the system can also be accessed through the Internet using Hypertext Transport Protocol or Simple File Transport Protocol. Our design and simulation work will be key to implementing a successful, scalable medical image archive and distribution system.

We have designed a teleradiology and telemedicine architecture over the World-Wide Web using current HIS, RIS and PACS. Our implementation allows remote access to hypermedia medical record and automatic management of interactive communications between referring physician and consultants. Security and privacy issues are also discussed. Its successful use in a telemedicine trial to China involving hundreds of doctors has shown the potential trend of telemedicine over the World-Wide Web.

The object of this study is using Internet resources to provide a cost-effective, user-friendly method to access the medical image archive system and to provide an easy method for the user to identify the images required. This paper describes the prototype system architecture, the implementation, and results. In the study, we prototype the Intelligent Medical Image Retrieval (IMIR) system as a Hypertext Transport Prototype server and provide Hypertext Markup Language forms for user, as an Internet client, using browser to enter image retrieval criteria for review. We are developing the intelligent retrieval engine, with the capability to map the free text search criteria to the standard terminology used for medical image identification. We evaluate retrieved records based on the number of the free text entries matched and their relevance level to the standard terminology. We are in the integration and testing phase. We have collected only a few different types of images for testing and have trained a few phrases to map the free text to the standard medical terminology. Nevertheless, we are able to demonstrate the IMIR's ability to search, retrieve, and review medical images from the archives using general Internet browser. The prototype also uncovered potential problems in performance, security, and accuracy. Additional studies and enhancements will make the system clinically operational.

Medicine employs a wide variety of digital data in various formats for clinical and research activities. To make medical data available for all users and equipment, it must be stored in a general format that is understandable by all end users. Also data must be structured so that it can be retrieved, manipulated, and stored efficiently. In this article, we propose a paradigm for representation of digital data such that it encompasses various modes of medical data without imposing demanding compliance requirements. Also the distributed nature of the proposed data format allows efficient data storage/transmission through use of a hierarchical data servers. Such a distributed data format provides not only a universal and uniform means of data storage/communication but also a valuable resource for medical research and public health planning.

Clinical prototypes of digital tele-ultrasound systems at the Bowman Gray School of Medicine have provided insight into various design architectures. Until network equipment costs decrease, hybrid systems often provide good cost/feature mixes by using high-cost networking equipment only when digital networking is required. Within the hospital using remote ultrasound system, a video and audio router interconnects the video output of ultrasound modalities and technologist communications subsystems. This is done either manually or by remote signaling, depending on the size of the ultrasound infrastructure and the cost of a remote signaling subsystem. For extramural sites and in hospital areas too distant for cost- effective analog switching techniques, an appropriate coder/decoder (CODEC), with echo cancellation, is used to transfer the audio and visual information to a CODEC in the viewing station location. The CODECs can be T1 (1.544 Mbps) CODECs for areas that cannot be reached economically at asynchronous transfer mode (ATM) data rates. This is contingent upon the diagnostic quality of the output of the T1 CODECs. Otherwise, high-speed CODECs are used with 45 Mbps DS-3 or ATM transmission facilities. This system allows full use of existing hospital infrastructures while adapting to emerging data communications infrastructures being implemented.

The Toshiba General Hospital introduced a departmental RIS/PACS (Radiology Information System/Picture Archiving and Communication System) in the radiology department in May, 1993. It has been used routinely since that time. In order to provide efficient means for clinicians to find and read many images, the system has been expanded to the neurosurgery and urology clinics and wards since May, 1995, and five image referring workstations now provide digital images to clinicians. In this paper we discuss an algorithm for image migration, one of the key issues to accomplish the expansion to outpatient clinics successfully, and propose the WYWIWYG (what you want is what you get) image transfer logic. This is the logic used to transfer images that physicians require refer without increasing the traffic between the image server and referring workstations. We accomplish the WYWIWYG logic by prioritizing exams the physicians have not yet viewed and by finding historical exams according to the modality, anatomy, and marking. Clinicians gave us comments from their first use of the system and suggested that the PACS enables clinicians to review images more efficiently compared to a film-based system. Our experience suggests that it is a key to the effective application of PACS in outpatient clinics to incorporate consideration patterns of clinicians on the migration algorithm.

The Geneva PACS is based on a distributed architecture, with different archive servers used to store all the image files produced by digital imaging modalities. Images can then be visualized on different display stations with the Osiris software. Image visualization require to have the image file physically present on the local station. Thus, images must be transferred from archive servers to local display stations in an acceptable way, which means fast and user friendly where the notion of file must be hidden to users. The transfer of image files is done according to different schemes including prefetching and direct image selection. Prefetching allows the retrieval of previous studies of a patient in advance. A direct image selection is also provided in order to retrieve images on request. When images are transferred locally on the display station, they are stored in Papyrus files, each file containing a set of images. File names are used by the Osiris viewing software to open image sequences. But file names alone are not explicit enough to properly describe the content of the file. A specific utility has been developed to present a list of patients, and for each patient a list of exams which can be selected and automatically displayed. The system has been successfully tested in different clinical environments. It will be soon extended on a hospital wide basis.

For the major imaging disciplines as radiology, PACSystems are still very costly because of the large image volumes involved. However, for low resolution images PACS appears to be feasible and cost effective, using mass produced general purpose hardware. Because endoscopy images fall within this category the Rotterdam University Hospital and BAZIS (lateron HISCOM) started in 1993 the development of an experimental ENdoScopy Information System (ENSIS) consisting of functions for the image acquisition and multimedia reporting. A multimedia report consists of a selection of the images acquired, annotations, spoken comments and a textual report. In addition the system offers retrieval of completed multimedia endoscopy reports by authorized users throughout the hospital as part of the existing electronic medical record. The multimedia functions of ENSIS, are developed for personal computers running MSDOS using a Novell fileserver with optical disks. From September 1994 the system is in clinical operation and from that moment the majority of the examinations (6000 per year) have resulted in a multimedia report. In this paper we will describe the functionality and design of the system.

We report on our experience with a recently introduced phosphorplate system (AGFA Diagnostic Center) from the viewpoint of overall operation efficiency. A first factor that determines efficiency is the time it takes to enter patient and examination information. A second factor is robustness of the automated image processing algorithms provided in the CR, as this determines the need for interactive image reprocessing on the workstation or for film retake. Both factors are strongly influenced by the integration of the modality within the HIS, whereby information about the patient and the examination request is automatically transferred to the phosphorplate system. Problems related to wrongly entered patient information are virtually eliminated. In comparison with manual entry of patient demographic data, efficiency has increased significantly. The examination information provided by the HIS helps the CR system to select optimal processing parameters automatically in the majority of situations. Furthermore, the image processing algorithms turn out to be rather robust and independent of pathology. We believe that both the HIS connection and the robustness of internal image processing contribute to making the current percentage of retakes and reprocessing in the order of 1.2% and 0.9% respectively, compared to more than 8% of retakes in the previous analogue systems.

A Picture Archiving and Communication System centered on a shared image file server can support a filmless hospital. Systems based on this architecture have proven themselves in over four years of clinical operation. Changes in healthcare delivery are causing radiology groups to support multiple facilities for remote clinic support and consolidation of services. There will be a corresponding need for communicating over a standardized wide area network (WAN). Interactive workflow, a natural extension to the single facility case, requires a means to work effectively and seamlessly across moderate to low speed communication networks. Several schemes for supporting a consortium of medical treatment facilities over a WAN are explored. Both centralized and distributed database approaches are evaluated against several WAN scenarios. Likewise, several architectures for distributing image file servers or buffers over a WAN are explored, along with the caching and distribution strategies that support them. An open system implementation is critical to the success of a wide area system. The role of the Digital Imaging and Communications in Medicine (DICOM) standard in supporting multi- facility and multi-vendor open systems is also addressed. An open system can be achieved by using a DICOM server to provide a view of the system-wide distributed database. The DICOM server interface to a local version of the global database lets a local workstation treat the multiple, distributed data servers as though they were one local server for purposes of examination queries. The query will recover information about the examination that will permit retrieval over the network from the server on which the examination resides. For efficiency reasons, the ability to build cross-facility radiologist worklists and clinician-oriented patient folders is essential. The technologies of the World-Wide-Web can be used to generate worklists and patient folders across facilities. A reliable broadcast protocol may be a convenient way to notify many different users and many image servers about new activities in the network of image servers. In addition to ensuring reliability of message delivery and global serialization of each broadcast message in the network, the broadcast protocol should not introduce significant communication overhead.

The PACS developed previously at the Delft University did not allow for real time interaction between specialists at different viewing stations. Recently, we enhanced our PACS with a Consultation Capability to provide a user with the ability to consult colleagues at other viewing stations. This consultation capability enables viewing the same image(s) simultaneously at different workstations and provides overlay information of pointers and rectangles, generated at one site, to become visible at the other site(s). In our earlier PACS, viewing of images is initiated by starting the Xclient application at one of the viewing stations. This procedure has been enhanced to enable a user to select two modes of operation namely either single- or multi-user session mode. In the new multi-user mode, all viewing stations involved in the same session, communicate with each other via a central control application (called consultation server), which is part of the enhanced supervisor. In the multi-user mode, only one user is allowed to select images, which will be displayed at all viewing stations involved in the session. The new viewing program sends it mouse information and framing information to the consultation server, which distributes this information to the other connected viewers. It is desired to remain DICOM compatible, however, ACR/NEMA has not (yet) handled the standardization of commands concerned with real-time consultation. Two different DICOM SOP (Service Object Pair) classes have been created. The first class is used to setup consultation, whereas the second class is used to exchange consultation data.

The current developments in the health care sector facilitate the deployment of telemedical applications which promise to decrease costs and improve treatment quality. The regular use of telemedicine as a means of exchanging knowledge and patient information over long distance requires a natural integration of human communication with all available information sources like imaging modalities and patient data. ATM and ISDN are network technologies which allow the transmission of both digital data and real-time information like video and speech and therefore recommend themselves for telemedical applications. In the future a heterogeneous infrastructure with ATM, ISDN and other network technologies can be expected, increasing the importance of `connectivity' among the network services. This contribution describes an application scenario for teleradiology using a combination of ATM and ISDN wide area networks and experiences with the experimental use of such a system.

The development of a hospital wide PACS is in progress at the University Hospital of Geneva and several archive modules are operational since 1992. This PACS is intended for wide distribution of images to clinical wards. As the PACS project and the number of archived images grow rapidly in the hospital, it was necessary to provide an easy, more widely accessible and convenient access to the PACS database for the clinicians in the different wards and clinical units of the hospital. An innovative solution has been developed using tools such as Netscape navigator and NCSA World Wide Web server as an alternative to conventional database query and retrieval software. These tools present the advantages of providing an user interface which is the same independently of the platform being used (Mac, Windows, UNIX, ...), and an easy integration of different types of documents (text, images, ...). A strict access control has been added to this interface. It allows user identification and access rights checking, as defined by the in-house hospital information system, before allowing the navigation through patient data records.

The quantity of image data created in a large radiology practice has long been a challenge for available archiving technology. Traditional methods ofarchiving the large quantity of films generated in radiology have relied on warehousing in remote sites, with courier delivery of film files for historical comparisons. A digital community archive, accessible via a wide area network, represents a feasible solution to the problem of archiving digital images from a busy practice. In addition, it affords a physician caring for a patient access to imaging studies performed at a variety ofhealthcare institutions without the need to repeat studies. Security problems include both network security issues in the WAN environment and access control for patient, physician and imaging center. The key obstacle to developing a community archive is currently political. Reluctance to participate in a community archive can be reduced by appropriate design of the access mechanisms.

Facing a world undergoing fundamental and rapid change, healthcare organizations are seeking ways to increase innovation, quality, productivity, and patient value, keys to more effective care. Individual clinics acting alone can respond in only a limited way, so re- engineering the process key which services are delivered demands real-time collaborative technology that provides immediate information sharing, improving the management and coordination of information in cross-functional teams. StreamWorks is a development stage architecture that uses a distribution technique to deliver an advanced information management system for telemedicine. The challenge of StreamWorks in telemedicine is to enable equity of the quality of Health Care of Telecommunications and Information Technology also to patients in less favored regions, like India or China, where the quality of medical care varies greatly by region, but where there are some very current communications facilities.

This paper describes the global system design of a PACS for nuclear medicine images. This NM PACS provides facilities for image capture, storage, display, manipulation and analysis. The NM PACS workstation displays besides images also the patient data from the HIS database. The NM PACS is equipped with well-defined HIS interface, which provides interoperability with HIS systems. The system design of the NM PACS is based on: a twin client-server concept, i.e. each workstation can run a HIS client and a PACS client, each interfaced with their own server. The HIS and the PACS servers are in turn inter-connected. The PACS images can be retrieved and displayed by evoking a command to a HIS menu. The X-protocol, together with GUI tools, such as Builder Xcsessory and the Motif tools in the Xmt library, are used to create the software modules that displays, manipulates and analyzes the images. The image file storage architecture consists of a single layer, namely an array of magnetical disks.

High speed networking is a crucial ingredient in medical information systems. ATM (Asynchronous Transfer Mode) and FDDI (Fiber Distributed Data Interface) networking have overcome their high costs and it is now possible to deploy these technologies widely. However, the impact of these networks on the performance of medical information systems is not well documented. Structured experiments are required to evaluate networking in complex systems of hardware and software. A test system designed to simulate an image delivery and display system using Ethernet, FDDI and ATM networking was established. Bottlenecks in this system related to networking protocols and hardware, as well as operating system and disk operations were identified and examined. Special attention was given to the DICOM (Digital Imaging and Communications in Medicine) protocol layered on TCP/IP (Transmission Control Protocol/Internet Protocol). Although the test ATM network was rated at 155 Mbps at the physical layer, observed maximum throughputs using TCP with optimized parameters were 79 Mbps for memory to memory transfers and 50 Mbps for disk to memory transfers. The default parameters yielded much lower throughputs with rates of 28 Mbps for memory to memory transfers. When the DICOM protocol was layered on top of TCP, a performance degradation of 10% to 78% of the optimal TCP rate was observed, depending on the type of image study being transferred (CR, CT, or MR). These initial data indicate the performance of a medical information system can be limited by a series of factors. Image data types have performance characteristics based on their image and study size. Appropriate selection and tuning of higher level protocols also makes a substantial contribution to system performance. Once network bandwidth exceeds ethernet speeds, disk operations are rate limiting factors in image retrieval.

Current generation of picture archiving and communication systems (PACS) lacks the capabilities to permit content-based searches to be made on image data and to visualize and render 3D image data in a cost effective manner. The purpose of this research project is to investigate a framework that will combine the storage and communication components of PACS with the power of content-based image indexing and 3D visualization. This presentation will describe the integrated architecture and tools of our experimental system with examples taken from applications of neurological surgical planning and assessment of pediatric bone age.

The integration of PACS and teleradiology capabilities into a healthcare enterprise in a clinically acceptable manner is both difficult and expensive. In order to justify the purchase of such systems, an understandings of the associated benefits is needed. By creating and executing process models that simulate the activities, infrastructure and communications within the radiology department both before and after the introduction of PACS and analyzing the resulting metrics, estimates of potential improvements in cost, efficiency and quality of delivered care can be quantified. A project to model and analyze processes within the MRI center at Southwest Texas Methodist Hospital in San Antonio, Texas, is described. The resulting process models, which utilized real-world metrics and process data collected at Methodist Hospital, was used to predict the effects of PACS on parameters such as operational costs, operational efficiency, resource consumption, length of patient stay and personnel requirements. Graphical illustrations of the models are presented, as are reports indicating predicted savings and efficiency increases that would result from the introduction of PACS into the MRI center.