The Wake Forest University School of Medicine and the Wake Forest University/Baptist Medical Center (WFUBMC) are implementing a second generation PACS. The first generation PACS provided helpful information about the functional and temporal requirements of the system. It highlighted the importance of image retrieval speed, system availability, RIS/HIS integration, the ability to rapidly view images on any PACS workstation, network bandwidth, equipment redundancy, and the ability for the system to evolve using standards-based components. This paper deals with the network design and implementation of the PACS. The physical layout of the hospital areas served by the PACS, the choice of network equipment and installation issues encountered are addressed. Efforts to optimize fault tolerance are discussed. The PACS network is a gigabit, mixed-media network based on LAN emulation over ATM (LANE) with a rapid migration from LANE to Multiple Protocols Over ATM (MPOA) planned. Two fault-tolerant backbone ATM switches serve to distribute network accesses with two load-balancing 622 megabit per second (Mbps) OC-12 interconnections. The switch was sized to be upgradable to provide a 2.54 Gbps OC-48 interconnection with an OC-12 interconnection as a load-balancing backup. Modalities connect with legacy network interface cards to a switched-ethernet device. This device has two 155 Mbps OC-3 load-balancing uplinks to each of the backbone ATM switches of the PACS. This provides a fault-tolerant logical connection to the modality servers which pass verified DICOM images to the PACS servers and proper PACS diagnostic workstations. Where fiber pulls were prohibitively expensive, edge ATM switches were installed with an OC-12 uplink to a backbone ATM switches. The PACS and data base servers are fault-tolerant, hot-swappable Sun Enterprise Servers with an OC-12 connection to a backbone ATM switch and a fast-ethernet connection to a back-up network. The workstations come with 10/100 BASET autosense cards. A redundant switched-ethernet network will be installed to provide yet another degree of network fault-tolerance. The switched-ethernet devices are connected to each of the backbone ATM switches with two-load-balancing OC-3 connections to provide fault-tolerant connectivity in the event of a primary network failure.

Local area networks in hospitals with connection to the Internet enable remote access to medical data and the deployment of distributed medical services. The use of standardized protocols like DICOM as required by the heterogeneous hard- and software infrastructure aggravates the problem that intruders can potentially gain access to sensitive data. Different levels of data protection are therefore required depending on the utilization of secured or publicly accessible networks, the use of standardized communication, and the differing national data security regulations. To investigate different speed-optimized data security concepts, we constructed exemplary scenarios with distributed telemedical services utilizing DICOM-conform software systems. The hospital networks are separated from the Internet by firewalls. Communication between the DICOM applications was made possible by integrating a security level between the DICOM upper layer protocol and the TCP/IP interface, while encrypting the whole datastream using the Secure Socket Layer Protocol (SSL). A DICOM-conform encryption of selected parts of the DICOM messages and files was developed, that encodes only patient-relevant data. Additionally a security proposal of the DICOM working group on security was implemented and analyzed. Data were encrypted by using either symmetric (public and private key) or symmetric (secret key) methods. This sped up the overall data transfer rate and allowed the DICOM-conform, off-line data storage.

A new network security protocol that uses smart IC cards has been designed to assure the integrity and privacy of medical information in communication over a non-secure network. Secure communication software has been implemented as a library based on this protocol, which is called the Integrated Secure Communication Layer (ISCL), and has been incorporated into information systems of the National Cancer Center Hospitals and the Health Service Center of the Tokyo Institute of Technology. Both systems have succeeded in communicating digital medical information securely.

Traditional PACS systems are based on two-tier client server architectures, and require the use of costly, high-end client workstations for image viewing. Consequently, PACS systems using the two-tier architecture do not scale well as data increases in size and complexity. Furthermore, use of dedicated viewing workstations incurs costs in deployment and maintenance. To address these issues, the use of digital library technologies, such as the World Wide Web, Java, and CORBA, is being explored to distribute PACS data to serve a broader range of healthcare providers in an economic and efficient manner. Integration of PACS systems with digital library technologies allows access to medical information through open networks such as the Internet. However, use of open networks to transmit medical data introduces problems with maintaining privacy and integrity of patient information. Cryptography and digital timestamping is used to protect sensitive information from unauthorized access or tampering. A major concern when using cryptography and digital timestamping is the performance degradation associated with the mathematical calculations needed to encrypt/decrypt an image dataset, or to calculate the hash value of an image. The performance issue is compounded by the extra layer associated with the CORBA middleware, and the use of programming languages interpreted at the client side, such as Java. This paper study the extent to which Java-based cryptography and digital timestamping affects performance in a PACS system integrated with digital library technologies.

Previously, we developed and implemented a lossless compression technique that provided a very high compression ratio for a variety of medical imaging modalities. We have extended our approach to satisfy additional requirements for the clinically acceptable implementation of lossless compression of digital medical images. Our new algorithm, called APC Codec (Rice) consists of a novel combination of techniques including adaptive prediction, Rice entropy coding, and multithreading. In order to demonstrate the clinical performance of our technique, we processed a large number of medical images (n greater than 10,000) obtained during the routine operation of the UCLA Clinical PACS. We report the resulting compression ratio and time statistics for different modalities and anatomies. The modalities tested were computed radiography (CR), magnetic resonance (MR), computed tomography (CT), and the anatomical regions included the brain, chest, abdomen and extremities. A comparison to the UNIX compress utility is provided as a performance benchmark.

The biomedical digital library of the future is expected to provide access to stores of biomedical database information containing text and images. Developing efficient methods for accessing such databases is a research effort at the Lister Hill National Center for Biomedical Communications of the National Library of Medicine. In this paper we examine issues in providing access to databases across the Web and describe a tool we have developed: the Web-based Medical Information Retrieval System (WebMIRS). We address a number of critical issues, including preservation of data integrity, efficient database design, access to documentation, quality of query and results interfaces, capability to export results to other software, and exploitation of multimedia data. WebMIRS is implemented as a Java applet that allows database access to text and to associated image data, without requiring any user software beyond a standard Web browser. The applet implementation allows WebMIRS to run on any hardware platform (such as PCs, the Macintosh, or Unix machines) which supports a Java-enabled Web browser, such as Netscape or Internet Explorer. WebMIRS is being tested on text/x-ray image databases created from the National Health and Nutrition Examination Surveys (NHANES) data collected by the National Center for Health Statistics.

Hospital enterprises are being created through mergers and acquisitions of existing hospitals. One area of interest in the PACS literature has been the integration of information systems and imaging systems. Hospital enterprises with multiple information and imaging systems provide new challenges to the integration task. This paper describes the requirements at the BJC Health System and a testbed system that is designed to acquire images from a number of different modalities and hospitals. This testbed system is integrated with Project Spectrum at BJC which is designed to provide a centralized clinical repository and a single desktop application for physician review of the patient chart (text, lab values, images).

This paper describes the acquisition of usage statistics and analysis of data-flow patterns for our DICOM-conformant PACS during its first year of operation. The system currently supports one MR, nine ultrasound, and three CT scanners, and has allowed our department to fully eliminate the production and use of film in these modalities. The aim was to not only aid in trouble-shooting, but to provide a means of examining usage patterns and quantifying data-flow and storage requirements to help identify where refinements should be focused. Of particular interest were statistics quantifying turn-around times for user-initiated transfers and retrieval from long-term storage; the quantities of data acquired, moved, retrieved and prefetched, (sorted by modality and anatomy); usage patterns of clinicians within the hospital, and the quantity of data accessed via a WWW interface to our PACS. The results have been instrumental in refining our physical network plan, modifying retrieval and transmission algorithms, and providing objective measures of the performance of our PACS. During the evolution of the system, the same data have allowed us to retrospectively examine whether certain modifications yielded improvements that were significant and whether expectations were met. The logging process continues since it is now relied upon as a tool for monitoring nearly all useful system parameters.

The purpose of this work was to develop a set of controls for image navigation and manipulation for use at a workstation. The intended focus was to provide a set of controls that would be most useful for the task of reading plain radiographs, such as those from chest, abdominal, and musculoskeletal imaging. We thought that most workstation controls were better suited to the interpretation of cross sectional imaging rather than plain radiography, and from current reports on user ergonomics and our own experience, developed a prototype control set. A goal of this design was to create controls that could be operated without the need to display menus on the workstation or select items from them. The control set was also designed to be operable without having to do visual searches for the controls or observe the controls while they were being used.

The main purpose of this project is to implement a Digital Image and Communications (DICOM) compliant online tape library system over the Internet. Once finished, the system will be used to store medical exams generated from U.S. ARMY Mobile ARMY Surgical Hospital (MASH) in Tuzla, Bosnia. A modified UC Davis implementation of DICOM storage class is used for this project. DICOM storage class user and provider are implemented as the system's interface to the Internet. The DICOM software provides flexible configuration options such as types of modalities and trusted remote DICOM hosts. Metadata is extracted from each exam and indexed in a relational database for query and retrieve purposes. The medical images are stored inside the Wolfcreek-9360 tape library system from StorageTek Corporation. The tape library system has nearline access to more than 1000 tapes. Each tape has a capacity of 800 megabytes making the total nearline tape access of around 1 terabyte. The tape library uses the Application Storage Manager (ASM) which provides cost-effective file management, storage, archival, and retrieval services. ASM automatically and transparently copies files from expensive magnetic disk to less expensive nearline tape library, and restores the files back when they are needed. The ASM also provides a crash recovery tool, which enable an entire file system restore in a short time. A graphical user interface (GUI) function is used to view the contents of the storage systems. This GUI also allows user to retrieve the stored exams and send the exams to anywhere on the Internet using DICOM protocols. With the integration of different components of the system, we have implemented a high capacity online tape library storage system that is flexible and easy to use. Using tape as an alternative storage media as opposed to the magnetic disk has the great potential of cost savings in terms of dollars per megabyte of storage. As this system matures, the Hospital Information Systems/Radiology Information Systems (HIS/RIS) or other components can be developed potentially as interfaces to the outside world thus widen the usage of the tape library system.

Three data-base architectures may be distinguished among Picture Archiving and Communication Systems (PACSs): (1) Configurations with logically and physically centralized data- base and file server, (2) systems with physically distributed file servers and a logically centralized data-base, and (3) installations with logically and physically distributed data- bases and file servers. A brief overview of these architectures and their scaleability, performance, and fault- tolerance is given. A PACS for an existing large university hospital is designed for the first as well as the second architecture using given image production data and workflow. We evaluate the fault-tolerance of the two architectures. By modeling the work-flow and employing queuing theory, solutions with practically realizable data transfer requirements are found for both architectures. With today's performance and cost of computers, storage, and information management technologies, the second and third architectures are preferably implemented, depending on the size of the installation. The architectures offer almost unlimited scaleability, very high fault-tolerance, and optimized workflow. We describe a modern commercial PACS that adheres to the open-systems concept and consists of software application programs that run, independent of specific computer and network components, on off-the-shelf hardware and under standard multi-platform operating systems and utilize commercial data-base management systems and network managers. The system is based on the second architecture with multiple islands of functionality, each with servers and archive modules and a physically distributed data-base. Our PACS architecture supports browser technology: Workstations use the data-base to determine the location of needed information and then, through the image browser, mount the appropriate file server for access. The architecture supports a concept similar to domain name server (DNS) directory services on the Internet. The system can be expanded to enterprise-wide installations with a logically distributed data-base. Openness, scaleability, and longevity of a PACS also strongly depend on the architecture of software applications in the operating and tool-set environment as well as on the distribution of image processing tasks across a PACS. These issues are discussed in the last section of our paper. We are presenting an image processing strategy that provides a consistent rendering of image gray-scale and spatial resolution throughout the entire PACS.

Building cost effective PACS solutions is a main concern in developing countries. Hardware and software components are generally much more expensive than in developed countries and also more tightened financial constraints are the main reasons contributing to a slow rate of implementation of PACS. The extensive use of Internet for sharing resources and information has brought a broad number of freely available software packages to an ever-increasing number of users. In the field of medical imaging is possible to find image format conversion packages, DICOM compliant servers for all kinds of service classes, databases, web servers, image visualization, manipulation and analysis tools, etc. This paper describes a PACS implementation for review and storage built on freely available software. It currently integrates four diagnostic modalities (PET, CT, MR and NM), a Radiotherapy Treatment Planning workstation and several computers in a local area network, for image storage, database management and image review, processing and analysis. It also includes a web-based application that allows remote users to query the archive for studies from any workstation and to view the corresponding images and reports. We conclude that the advantage of using this approach is twofold. It allows a full understanding of all the issues involved in the implementation of a PACS and also contributes to keep costs down while enabling the development of a functional system for storage, distribution and review that can prove to be helpful for radiologists and referring physicians.

This paper describes the design and implementation of a low- cost PACS conforming to the DICOM 3.0 standard. The goal was to provide an efficient image archival and management solution on a heterogeneous hospital network as a basis for filmless radiology. The system follows a distributed, client/server model and was implemented at a fraction of the cost of a commercial PACS. It provides reliable archiving on recordable CD and allows access to digital images throughout the hospital and on the Internet. Dedicated servers have been designed for short-term storage, CD-based archival, data retrieval and remote data access or teleradiology. The short-term storage devices provide DICOM storage and query/retrieve services to scanners and workstations and approximately twelve weeks of 'on-line' image data. The CD-based archival and data retrieval processes are fully automated with the exception of CD loading and unloading. The system employs lossless compression on both short- and long-term storage devices. All servers communicate via the DICOM protocol in conjunction with both local and 'master' SQL-patient databases. Records are transferred from the local to the master database independently, ensuring that storage devices will still function if the master database server cannot be reached. The system features rules-based work-flow management and WWW servers to provide multi-platform remote data access. The WWW server system is distributed on the storage, retrieval and teleradiology servers allowing viewing of locally stored image data directly in a WWW browser without the need for data transfer to a central WWW server. An independent system monitors disk usage, processes, network and CPU load on each server and reports errors to the image management team via email. The PACS was implemented using a combination of off-the-shelf hardware, freely available software and applications developed in-house. The system has enabled filmless operation in CT, MR and ultrasound within the radiology department and throughout the hospital. The use of WWW technology has enabled the development of an intuitive we- based teleradiology and image management solution that provides complete access to image data.

We have developed a portable DICOM-compliant low- cost PC NT- based display workstation with Asynchronous Transfer Mode (ATM) and Ethernet connectivity to Picture Archiving and Communication System (PACS). This paper describes the hardware components and software system of the workstation, and compares the performance of ATM and Ethernet connectivity in the workstation. The display workstation consists of a DOME Md2/PCI board, two 1,600 X 1,280 resolution monitors, and an asynchronous transfer mode (ATM) OC-3 communication board. Implemented software includes: (1) a DICOM server, which automatically receives images from the DICOM-compliant PACS Controller: (2) a patient folder manager, which incorporates the local database to group multiple studies from individual patients; (3) a sophisticated display program with automatic image pre-processing functions, such as pre-set window and level; and (4) a DICOM query and retrieve services to retrieve images from DICOM-compliant archive. The display workstation demonstrates that: (1) the PC/NT-based display workstation is 100% DICOM-compliant. (2) DICOM query and retrieve services support convenient multi-level retrieval of historical images from PACS archive. (3) Integration of DICOM and ATM into the display workstation facilities and speeds up image communication. (4) Patient folder management provides hierarchical image storage for individual patients, which consequently facilitates review of image. This paper demonstrates that the PC/NT based full-functioned display workstation can be implemented cost-effectively for PACS applications.

We are in the process of conducting a research of full-field direct digital telemammography using three protocols: telediagnosis, teleconsultation, and telemanagement. To conduct this research project, an asynchronous transfer mode network based telemammography system was developed across two remote campuses in our facility. The hardware and software components of this system are detailed. The system was embedded in a clinical environment for a four-month test. Some preliminary study results from the current phase of this study are reported.

The current version of the Global PACS software system uses a Java-based implementation of the Remote Consultation and Diagnosis (RCD) system. The Java RCD includes a multimedia consultation session between physicians that includes text, static image, image annotation, and audio data. The JAVA RCD allows 2-4 physicians to collaborate on a patient case. It allows physicians to join the session via WWW Java-enabled browsers or stand alone RCD application. The RCD system includes a distributed database archive system for archiving and retrieving patient and session data. The RCD system can be used for store and forward scenarios, case reviews, and interactive RCD multimedia sessions. The RCD system operates over the Internet, telephone lines, or in a private Intranet. A multimedia consultation session can be recorded, and then played back at a later time for review, comments, and education. A session can be played back using Java-enabled WWW browsers on any operating system platform. The JAVA RCD system shows that a case diagnosis can be captured digitally and played back with the original real-time temporal relationships between data streams. In this paper, we describe design and implementation of the RCD session playback.

The functionality of many teleradiology systems is limited to image transfer to a remote workstation. It is not possible to access reports or images from older studies, or to create new reports of current cases. Additionally, teleradiology systems from different vendors are usually not interoperable. Future teleradiology systems should provide bi-directional image transfer as well as image annotation, the possibility to create and send back text reports, and access to previous reports. Furthermore, it becomes increasingly important that teleradiology systems are conformant to the DICOM standard. The advanced teleradiology service (ATS) is compatible to the DICOM standard and is based on the functional model of the standard. In addition to receiving current images which is straightforward, the ATS supports retrieving of correlative images, studies and reports. One further feature of ATS is the review of images which is supported with flexible image compression to keep the demand for bandwidth minimal. Additionally, creating and modifying multimedia results is supported. The multimedia results are based on 'Structured Reporting' (SR) which is a supplement of the DICOM standard. SR offers a mechanism to represent results for many fields of application in medicine. The types of documents supported by SR range from simple text descriptions to multimedia interpretation reports. For this reason it is important to evaluate scenarios and to define the requirements of SR in the radiological environment. The definition of ATS is intended to provide a tested reference architecture for teleradiology to which vendors can adhere in order to develop teleradiology systems with DICOM 'plug and play' capability.

Real-time consultation between referring physicians or radiologists with an expert is critical for timely and adequate management of problem cases. During consultation, both sides need to (1) synchronously manipulate high resolution digital radiographic images or large volume MR/CT images, (2) perform interpretation interactively, and (3) converse with audio. We present a specific designed teleconsultation system with bi-directional remote control technology to meet critical teleconsultation application with high resolution and large volume medical images in a limited bandwidth network environment. We give the system design and implementation methods, and also describe the teleconsultation procedure and protocol used in this system. Finally, preliminary results are discussed.

During the past 5 years (1992 - 1997) the Department of Radiology of the University of New Mexico Health Sciences Center has developed an active teleradiology program. Contracts are in place to provide both routine and emergency image interpretations 24 hours per day, every day of the year. Several rural hospitals are served as well as the Navajo Indian Health Service. Areas of success: include significantly improved radiologic service to the rural sites, specialty consultations to general radiologists, successful teaching of teleradiology practice to radiology residents and staff, good diagnostic quality images, a small but real profit, improved quality assurance for the rural sites, and no significant medical-legal problems. Failures include: significant telecommunications problems, lack of acceptance and utilization by some of the rural sites, poor QA compliance by some sites, a long period of disappointing technical support by equipment vendors, and slow acceptance of DICOM by equipment manufacturers. The successes outweigh the failures. We would do it again -- but somewhat differently. We offer advice to institutions developing a new rural teleradiology operation.

The changes that have overtaken the U.S. healthcare industry in the last five years could be best characterized as tectonic shifts. Every aspect of the healthcare market has been affected by the changes in Government policy and the attitude of society to issues in Healthcare. Most of these changes have been viewed as adversarial both to the health care provider and to the consumer. Healthcare reform was to have made healthcare more affordable and more available. Although healthcare reform was not passed, attempts have been made nationwide to address the ills of the system. These attempts have been largely half-hearted and weak-kneed. In most instances, only half a solution has been provided. There has been no improvement in the quality of care. In fact, in many instances, there has been degradation in quality and it has not become more available. We are faced with seemingly conflicting mandates -- providing quality care making it more available working under severe capitation constraints and attracting and retaining a quality workforce. How do we address these problems? We have to change. We have to adopt the military paradigm of agility, adaptability and flexibility applicable to military science to our field of endeavor. We have to consider achieving all our goals without sacrificing any aspect. The most obvious step is to improve efficiency. This can be done best by incorporating the advantages that information technology has bestowed on other fields of endeavor. Properly applied information technology will provide the answer to improving efficiency in the Healthcare field. In the Department of Defense (DoD), we are now embarking on an extremely exciting new idea -- rendering the entire Virtual Radiology Environment (VRE). The business of radiology in the military therefore, is being re-engineered on several fronts. This is achieved in several sequential steps: (1) Equipping every radiology department to become digital and PACS-network capable. (2) Information Retrieval, Routing and Archiving -- developing a network. (3) Relook at staffing patterns and intelligent management of those patterns. (4) Arming ourselves for the future -- learning to enjoy chaos. In this paper, we then formulate a master plan for developing large hospital PAC Systems that will service the entire Regional Medical Command (RMC) and Medical Treatment Facilities (MTFs). The systems engineering methodology is presented, as well as some important characteristics of the VRE. The resulting Virtual Radiology Environment will enable to improve access and quality and reduce cost in the Army.

The U.S. Army Medical Command, lead by the Brooke Army Medical Center, has embarked on a futuristic project which will revolutionize the practice of radiology in the DoD. The U.S. Army Virtual Radiology Environment (USAVRE) is a CONUS-based network that connects all the Army's major medical centers and Regional Medical Commands (RMC). The purpose of the USAVRE is to improve the quality, access, and cost of radiology services in the Army via the use of state-of-the-art medical imaging, computer, and networking technologies. The USAVRE contains multimedia-viewing workstations for static and dynamic modality cases. The storage and archiving systems are based on a distributed computing environment using Common Object Request Broker Architecture (CORBA) middleware protocols. Collaboration between archive centers and viewing workstations are managed by CORBA functions and multimedia object streams. The underlying Telecommunications network is an ATM based backbone network that connects to the RMC regional networks and PACS local networks at medical centers and RMC clinics. The U.S. Army Information Systems Engineering Command (USAISEC) at Ft. Huachuca, AZ is responsible for the ATM backbone network to the RMC sites. The virtual Radiology services in a USAVRE must be applied to several radiology modalities in a virtual network environment. In this discussion, we assume the existence of several PACS networks within a USAVRE environment that have a need to exchange multimedia images and patient information. We define a multimedia collaborative distributed computing environment (DCE) in medical imaging and radiology as a collection of collaborating PACS networks with workstations and image archive systems for the purposes of acquiring and exchanging patient static and video sequence images; storage, retrieval, and archival of those images; performing image analysis and multimedia consultation on patient cases; operation and management of the network to optimize its resources; and to improve the quality and access of the radiology services to the patients. This paper describes the open systems architecture for the USAVRE, including the PACS and Global PACS user equipment. This project is a collaborative effort between military, university, and industry centers with expertise in Teleradiology and Global PACS applications.

In a situation of shrinking health care budgets, increasing cost pressure and growing demands to increase the efficiency and the quality of medical services, health care enterprises are forced to optimize or complete re-design their processes. Although information technology is agreed to potentially contribute to cost reduction and efficiency improvement, the real success factors are the re-definition and automation of processes: Business Process Re-engineering and Workflow Management. In this paper we discuss architectures for the use of workflow management systems in radiology. We propose to move forward from information systems in radiology (RIS, PACS) to Radiology Management Systems, in which workflow functionality (process definitions and process automation) is implemented through autonomous workflow management systems (WfMS). In a workflow oriented architecture, an autonomous workflow enactment service communicates with workflow client applications via standardized interfaces. In this paper, we discuss the need for and the benefits of such an approach. The separation of workflow management system and application systems is emphasized, and the consequences that arise for the architecture of workflow oriented information systems. This includes an appropriate workflow terminology, and the definition of standard interfaces for workflow aware application systems. Workflow studies in various institutions have shown that most of the processes in radiology are well structured and suited for a workflow management approach. Numerous commercially available Workflow Management Systems (WfMS) were investigated, and some of them, which are process- oriented and application independent, appear suitable for use in radiology.

Because of its exciting potential to improve clinical service, as well as reduce costs, a voice recognition system for radiological dictation was recently installed at our institution. This system will be clinically successful if it dramatically reduces radiology report turnaround time without substantially affecting radiologist dictation and editing time. This report summarizes an observer study currently under way in which radiologist reporting times using the traditional transcription system and the voice recognition system are compared. Four radiologists are observed interpreting portable intensive care unit (ICU) chest examinations at a workstation in the chest reading area. Data are recorded with the radiologists using the transcription system and using the voice recognition system. The measurements distinguish between time spent performing clerical tasks and time spent actually dictating the report. Editing time and the number of corrections made are recorded. Additionally, statistics are gathered to assess the voice recognition system's impact on the report cycle time -- the time from report dictation to availability of an edited and finalized report -- and the length of reports.

The impact of direct digital capture on projection radiography has been explored with Sterling Diagnostic Imaging's DirectRay amorphous selenium detector in several clinical trials. Progress, key lessons learned, and immediate plans are discussed below. Plans for a clinical field trial involving a dedicated chest system are described. The existing screen-film workflow is compared with the workflow of an integrated digital chest system. The flexibility of the digital system is described as it relates to a phased approach to PACS implementation. The relationship of DirectRay images with DIOCM is described as it relates to interoperability. This includes the image processing architecture and the development of the DICOM Digital X-Ray (DX) Information Object Definition (IOD).

This paper describes an integrated system designed to provide efficient means for DICOM compliant cardiac imaging archival, transmission and visualization based on a communications backbone matching recent enabling telematic technologies like Asynchronous Transfer Mode (ATM) and switched Local Area Networks (LANs). Within a distributed client-server framework, the system was conceived on a modality based bottom-up approach, aiming ultrafast access to short term archives and seamless retrieval of cardiac video sequences throughout review stations located at the outpatient referral rooms, intensive and intermediate care units and operating theaters.

The developments in information technologies -- computer hardware, networking and storage media -- has led to expectations that these advances make it possible to replace 35 mm film completely by digital techniques in the catheter laboratory. Besides the role of an archival medium, cine film is used as the major image review and exchange medium in cardiology. None of the today technologies can fulfill completely the requirements to replace cine film. One of the major drawbacks of cine film is the single access in time and location. For the four catheter laboratories in our institutions we have designed a complementary concept combining the CD-R, also called CD-medical, as a single patient storage and exchange medium, and a digital archive for on-line access and image review of selected frames or short sequences on adequate medical workstations. The image data from various modalities as well as all digital documents regarding to a patient are part of an electronic patient record. The access, the processing and the display of documents is supported by an integrated medical application.

An independent evaluation of PACS has been carried out at Hammersmith Hospital. This paper describes one element: the use of a questionnaire instrument to assess radiology service user's views on the quality of the service being provided; major causes of dissatisfaction with the service; the extent to which images are unavailable; and the consequences of images unavailability. The objective was to investigate some of the key claims made for the PACS technology. The principal research design was a 'before and after' comparison at Hammersmith Hospital. A number of other, comparator, hospitals were included in this survey in order to allow inferences to be made about the any observed changes at Hammersmith. The questionnaire was distributed on three occasions before PACS was operational at Hammersmith and on one occasion afterwards. Across all data collection rounds and all sites, very high levels of satisfaction with image quality were reported. When asked about satisfaction with the written reporting service, a larger proportion of respondents across all sites and rounds indicated their discontent Following the introduction of PACS, the proportion of respondents indicating that lost ward or outpatient images was a problem was significantly lower and the rate of re-examination was lower.

The aim of the study described in this paper was to identify and measure any changes in radiation dose attributable to the implementation of the hospital wide Hammersmith PACS system. The authors believe this study to be a comprehensive study of changes in patient dose for the examination of the lateral lumbar spine associated with the introduction of PACS: the study has fully monitored the many factors which affect these patient doses. Some very optimistic claims of large dose reductions with the introduction PACS at other hospitals have been based upon minimal reported evidence, or comparisons have been made with very old high dose systems with film/screen sensitivities as low as 100. In this study the patient doses with the PACS system have been compared with those measured when a conventional film/screen system with a sensitivity of 300 was used.

The objective of this research was to assess the costs and benefits associated with the introduction of a small PACS system into an intensive care unit (ICU) at a district general hospital in north Wales. The research design adopted for this study was a single center randomized controlled trial (RCT). Patients were randomly allocated either to a trial arm where their x-ray imaging was solely film-based or to a trial arm where their x-ray imaging was solely PACS based. Benefit measures included examination-based process measures, such as image turn-round time, radiation dose and image unavailability; and patient-related process measures, which included adverse events and length of stay. The measurement of costs focused on additional 'radiological' costs and the costs of patient management. The study recruited 600 patients. The key findings from this study were that the installation of PACS was associated with important benefits in terms of image availability, and important costs in both monetary and radiation dose terms. PACS-related improvements in terms of more timely 'clinical actions' were not found. However, the qualitative aspect of the research found that clinicians were advocates of the technology and believed that an important benefit of PACS related to improved image availability.

DICOM provides integration solutions for PACS to interconnect multi-vendor imaging equipment, regardless of computer hardware and operating system platforms, for communication of images. This paper describes our three-year experience integrating heterogeneous imaging systems into a large-scale PACS based on DICOM. Vendor imaging equipment connected to the PACS include (1) six CT and MR imaging devices; (2) two CR systems; (3) two laser film digitizers; (4) one commercial ultrasound PACS module; and (5) eight DICOM-based display workstations. The image communication software implemented in the PACS was modified from Mallinckrodt's Central Test Node software and University of California at Davis' DICOM software to support the Unix-based and Windows NT-based computer systems, respectively. Images acquired from the individual CT/MR devices, CR systems, film digitizers, and the commercial PACS module were transmitted from the acquisition computers to the PACS Controller, where these images were routed to the display workstations and the archive server. Problems during integration included: (1) missing slice image(s) from an image sequence during the acquisition process; (2) incorrect data encoding in image header; (3) DICOM Service Object Pair Class not supported by the Service Class Provider vendor; and (4) DICOM conformance mismatching and shadow group conflict between the individual vendors. These problems were resolved by the mechanisms we implemented in our Image Acquisition Gateways and PACS Controller. The DICOM communications standard provides a simple means of interconnecting multi- vendor equipment for communication of images. However, operating a large-scale PACS configured with heterogeneous imaging systems requires significant effort in software integration to ensure data integrity and operation efficiency.

Interoperability of imaging devices (Modality), PACS and RIS is a crucial point determining the effectiveness and the performance of a digital diagnostic radiology department. This paper represents an integrated environment of Modality, PACS and RIS based on a communication interface using the DICOM standard. In the environment, RIS acts as the master system and provides overall image management and move control functions. This allows a considerably simplified interface design and an employment of standard DICOM services for the implementation. All important RIS - PACS inter-operation functions like Image Pre-fetching, Auto-Routing, Report Delivery, etc., are supported. Using the DICOM Standard, the system integration has been achieved with an open communication protocol and therefore is expected to be supported by various vendors.

Recently, a DICOM-compliant image server application (Query/Retrieve Service Class Provider; Q/R SCP) is available commercially or freely in small or large scale. It is better to use these products as an image management sub system of sophisticated image networking applications such as a teaching file system and electric clinical patient record exchange. We describe a new facility to construct medical image database systems using Query/Retrieve Service Class applications. We propose a new Internet Standard; DICOM-QR URL scheme. It is an expansion of the URL (Uniform Resource Locator) standard to comply the DICOM standard. The DICOM-QR URL conveys the complete information to access to the specified images on the Q/R SCP. It is encoded into the compact ASCII string that is easily stored in fields of relational databases or text documents. Finally, we propose that the DICOM-QR URL scheme should be standardized following the Internet Standards Process and we propose many colleagues to join the process.

Purpose -- To integrate radiological DICOM images into our currently existing web-browsable Electronic Medical Record (MINDscape). Over the last five years the University of Washington has created a clinical data repository combining in a distributed relational database information from multiple departmental databases (MIND). A text-based view of this data called the Mini Medical Record (MMR) has been available for three years. MINDscape, unlike the text based MMR, provides a platform independent, web browser view of the MIND dataset that can easily be linked to other information resources on the network. We have now added the integration of radiological images into MINDscape through a DICOM webserver. Methods/New Work -- we have integrated a commercial webserver that acts as a DICOM Storage Class Provider to our, computed radiography (CR), computed tomography (CT), digital fluoroscopy (DF), magnetic resonance (MR) and ultrasound (US) scanning devices. These images can be accessed through CGI queries or by linking the image server database using ODBC or SQL gateways. This allows the use of dynamic HTML links to the images on the DICOM webserver from MINDscape, so that the radiology reports already resident in the MIND repository can be married with the associated images through the unique examination accession number generated by our Radiology Information System (RIS). The web browser plug-in used provides a wavelet decompression engine (up to 16-bits per pixel) and performs the following image manipulation functions: window/level, flip, invert, sort, rotate, zoom, cine-loop and save as JPEG. Results -- Radiological DICOM image sets (CR, CT, MR and US) are displayed with associated exam reports for referring physician and clinicians anywhere within the widespread academic medical center on PCs, Macs, X-terminals and Unix computers. This system is also being used for home teleradiology application. Conclusion -- Radiological DICOM images can be made available medical center wide to physicians quickly using low-cost and ubiquitous, thin client browsing technology and wavelet compression.

When a set of medical images are recorded on film, some contrast windowing of the raw data must be performed. Whoever subsequently sees the film sees exactly what has been recorded. If the raw data is sent to another system, maybe at a remote site, there is no standard way to specify the layout and image presentation parameters. The meta language, Interscreen, described here, overcomes this problem since it allows the recording of all information needed to reproduce a display on any system, while retaining freedom to interact with the data. By separating presentation data from raw image data, a number of different displays may be defined for a given set of images using different display thresholding, layout and lookup tables (different views of the data). It operates at a level above that of page description languages such as HTML. The facilities support a number of sub-object specification languages and can be expressed in DICOM or text tag form for example. Interscreen provides a hierarchy of entities for describing a display window and the layout and appearance of the various objects it contains.

The Government released a Request for Proposal (RFP) in January 1997 for a Digital Imaging and Communications in Medicine (DICOM) Picture Archiving and Communication System (PACS) known as the Digital Imaging Network-Picture Archival and Communication System (DIN-PACS). The RFP included the requirement for each of the submitting vendors to support benchmark testing of their proposed architectures. The benchmark test equipment, protocols, and procedures, were developed through a joint effort between the Government, its contracting office, and consulting engineers. The intent of the benchmark test was to evaluate each proposed architecture in nine specific areas. This paper presents the final benchmark test methods used to evaluate a DICOM PACS architecture.

The purpose of this research was to develop an objective methodology for evaluating the clinical utility of imaging workstation functions. To test this methodology, we compared a general purpose workstation features against a task-based workstation features. Prior to conducting the study, radiologists input was used to develop these workstation features in order to answer a formulated research question. Next, a comprehensive list of features was compiled for each workstation. From this list, qualitative predictions were made as to which workstation would produce greater clinical utility. The four parameters used to measure clinical utility included: efficiency (time based measures), user satisfaction, confidence in diagnosis and accuracy in diagnosis. Radiologists read a total of two hundred cases on both workstations and these cases where categorized by complexity. Analysis found that the task-based workstation has greater clinical utility overall. The task-based workstation was found to be significantly more efficient, brought greater confidence in diagnosis, and was found to be more satisfying to use overall. There was no significant difference in accuracy of diagnosis between workstations. It was found that our quantitative results matched with the qualitative prediction made about clinically utility prior to the onset of the study; therefore, validating the methodology.

With the increasing number of PACS image display workstations deployed in clinical operations, a PACS image server needs to be tuned more effectively to achieve optimal system throughput. There are many factors that affect the performance of image transfer, including the application software program, the storage subsystem, and the network system. This paper primarily addresses the issues involving the image storage subsystem. Related research has been done to address the issues in network system and application software. One primary factor that affects the overall performance of a PACS image server is the throughput and reliability characteristics of the magnetic storage devices for image caching. A redundant array of inexpensive disks (RAID) is a mean to provide reliable and high performance storage. The UCLA PACS image server relies on a software-based RAID system to fulfill the bandwidth requirement and reliability demand. An extensive evaluation on various hardware and software parameters in the I/O subsystem has been conducted to derive an optimal configuration, which specifically attempts to address the following questions: How many disks should be in a stripe group? What is the optimal stripe chunk size? How big should the maxcontig be? What should be the IO request size? And how many concurrent I/O accesses should be allowed?

The experimental site for the evaluation reported was the Hammersmith Hospital, London. The study adopted an economic perspective in that the focus was on the change in various elements of cost and the change in various parameters of benefit following the implementation of a hospital-wide PACS. Comparison was made of hospital operation in a film-based situation with operation in a PACS-based situation. Some of the research activities focused on the radiology service itself at Hammersmith, others focused on other areas of the hospital where radiological information was seen as an important component in clinical decision-making, and others looked outside the hospital. In terms of operational, clinical and patient benefits, the evaluation found no significant indicators of disadvantages of PACS and many examples of significant actual measurable benefits or perceived advantages by users. However, as one might expect, these advantages come at a significant net cost.

The purpose of this work was to develop a method for systematically testing the reliability of a CR system under realistic daily loads in a non-clinical environment prior to its clinical adoption. Once digital imaging replaces film, it will be very difficult to revert back should the digital system become unreliable. Prior to the beginning of the test, a formal evaluation was performed to set the benchmarks for performance and functionality. A formal protocol was established that included all the 62 imaging plates in the inventory for each 24-hour period in the study. Imaging plates were exposed using different combinations of collimation, orientation, and SID. Anthropomorphic phantoms were used to acquire images of different sizes. Each combination was chosen randomly to simulate the differences that could occur in clinical practice. The tests were performed over a wide range of times with batches of plates processed to simulate the temporal constraints required by the nature of portable radiographs taken in the Intensive Care Unit (ICU). Current patient demographics were used for the test studies so automatic routing algorithms could be tested. During the test, only three minor reliability problems occurred, two of which were not directly related to the CR unit. One plate was discovered to cause a segmentation error that essentially reduced the image to only black and white with no gray levels. This plate was removed from the inventory to be replaced. Another problem was a PACS routing problem that occurred when the DICOM server with which the CR was communicating had a problem with disk space. The final problem was a network printing failure to the laser cameras. Although the units passed the reliability test, problems with interfacing to workstations were discovered. The two issues that were identified were the interpretation of what constitutes a study for CR and the construction of the look-up table for a proper gray scale display.

Purpose/Background: Asynchronous transfer mode (ATM) network technology has recently been used for high speed transmission of radiological images between hospitals and inside hospitals. However, the number of clinical sites which routinely use this technology is limited. The purpose of this study was to analyze the very early impact of an ATM link between a large tertiary referral center and small peripheral clinic on cost and clinical practice. Methodology: An ATM link using 155 bps (OC3) technology was installed between the West Los Angeles VA Medical Center and the Sepulveda VA, a large outpatient facility which provides full service radiological services. The West Los Angeles VA Medical Center is a large tertiary referral center with sub-specialist radiologist. The clinical impact of this ATM link between a large full-scale DICOM-3 compliant PACS system at the West LA VA on a smaller PACS system at the Sepulveda VA was evaluated. Results: The ability to freely exchange complicated MRI and CT studies between a tertiary referral center and a clinic could have a direct impact on patient care. Over the last six months, all and CT studies from Sepulveda VA were readily available via the ATM connection to all radiologists at the West LA VA. On average the workload at the Sepulveda VA in CT and MRI was about one tenth of the same workload at West LA VA, thus creating interesting possibilities for sharing or radiologist resources. Conclusions: Although our preliminary data and work loads have been too limited to draw any final conclusions yet, we feel that future results will show that the ability to provide immediate and fast interactive consultation between general radiologists in a large outpatient facility and sub- specialists at a tertiary referral center can have an impact upon the quality of patient care.

Although a number of authors have described the challenges and benefits of filmless operation using a hospital-wide Picture Archival and Communication System (PACS), there have been few descriptions of a multi-hospital wide area PACS. The purpose of this paper is to describe our two and a half year experience with PACS in an integrated multi-facility health care environment, the Veterans Affairs Maryland Health Care System (VAMHCS). On June 17, 1995 the Radiology and Nuclear Medicine services became integrated for four medical centers forming the VA Maryland Health Care System creating a single multi-facility imaging department. The facilities consisted of the Baltimore VA (acute and outpatient care, tertiary referral center), Ft. Howard (primarily long term care), Perry Point (primarily psychiatric care), and the Baltimore Rehabilitation and extended care facility (nursing home). The combined number of studies at all four sites is slightly more than 80,000 examinations per year. In addition to residents and fellows, the number of radiologists at Baltimore was approximately seven, with two at Perry Point, one at Ft. Howard, and no radiologists at the Rehabilitation and Extended Care facility. A single HIS/RIS, which is located physically at the Baltimore VAMC is utilized for all four medical centers. The multi- facility image management and communication system utilizes two separate PAC Systems that are physically located at the Baltimore VA Medical Center (BVAMC). The commercial system (GE Medical Systems) has been in place in Baltimore for more than 41/2 years and is utilized primarily in the acquisition, storage, distribution and display of radiology and nuclear medicine studies. The second PACS is the VISTA Imaging System, which has been developed as a module of the VA's HIS/RIS by and for the Department of Veterans Affairs. All of the radiology images obtained on the commercial PACS are requested by the VISTA Imaging System using DICOM query/retrieve commands and are stored on a separate server and optical jukebox. Additionally, the VISTA system is used to store all images obtained by all specialties in the medical center including pathology, dermatology, GI medicine, surgery, podiatry, ophthalmology, etc. Using this two PAC system approach, the hospital is able to achieve redundancy with regard to image storage, retrieval, and display of radiology images. The transition to a 'virtual' multi-facility imaging department was accomplished over a period of two years. Initially, Perry Point and Ft. Howard replaced their general radiographic film processors with Computed Radiography (CR) units. The CR units and subsequently, the CT and Ultrasound systems at Perry Point were interfaced (DeJarnette Research Systems) with the commercial PACS located in Baltimore. A HIS/RIS to modality interface was developed (DeJarnette and Fuji Medical Systems) between the computed radiography and CT units and VISTA Information System at Baltimore. A digital dictation system was recently implemented across the multi- facility network. The integration of the three radiology departments into a single virtual imaging department serving four medical centers has resulted in a number of benefits. Economically, there has been the elimination via attrition of one and a half radiologist FTE's (full time equivalents) and an administrative position resulting in an annual savings of more than $375,000 per year. Additionally, the expenditures for moonlighter coverage for vacation, meeting, and sick leave have been eliminated. There is now subspecialty coverage for primary or secondary interpretation and for peer review.

We have previously described a system for delivering radiology information to the desktop computers used for the electronic medical record (EMR). The system was built with the ability to record physician usage to a database. This usage information was then studied to help understand the value and requirements of an application that could display radiology information on the EMR workstations. This system was used by both primary care physicians and specialists primarily in the out-patient setting. We found that while there was substantial variation in usage both within and between the two physician groups, there was a high degree of support for maintaining image display capabilities on the workstations.

To evaluate a conventional radiology image management system, by investigating information accuracy, and information delivery. To discuss the customization of a picture archival and communication system (PACS), integrated radiology information system (RIS) and hospital information system (HIS) to a high volume emergency department (ED). Materials and Methods: Two data collection periods were completed. After the first data collection period, a change in work rules was implemented to improve the quality of data in the image headers. Data from the RIS, the ED information system, and the HIS as well as observed time motion data were collected for patients admitted to the ED. Data accuracy, patient waiting times, and radiology exam information delivery were compared. Results: The percentage of examinations scheduled in the RIS by the technologists increased from 0% (0 of 213) during the first period to 14% (44 of 317) during the second (p less than 0.001). The percentage of images missing identification numbers decreased from 36% (98 of 272) during the first data collection period to 10% (56 of 562) during the second period (p less than 0.001). Conclusions: Radiologic services in a high-volume ED, requiring rapid service, present important challenges to a PACS system. Strategies can be implemented to improve accuracy and completeness of the data in PACS image headers in such an environment.

Recent economic analyses of picture archiving and communication system (PACS) and computed radiography (CR) suggest that these systems are not cost saving from the perspective of a large, subspecialized department. Nevertheless, adoption of these systems is increasing. We examine the reasons for this apparent contradiction by analyzing a recently published economic model of PACS. Through sensitivity analysis and re-examination of model assumptions, we suggest several scenarios under which PACS adoption may be more economically favorable. We also discuss the non-economic factors that may influence adoption decisions.

The BJC Health System (BJC) and the Washington University School of Medicine (WUSM) formed a technology alliance with industry collaborators to develop and implement an integrated, advanced clinical information system. The industry collaborators include IBM, Kodak, SBC and Motorola. The activity, called Project Spectrum, provides an integrated clinical repository for the multiple hospital facilities of the BJC. The BJC System consists of 12 acute care hospitals serving over one million patients in Missouri and Illinois. An interface engine manages transactions from each of the hospital information systems, lab systems and radiology information systems. Data is normalized to provide a consistent view for the primary care physician. Access to the clinical repository is supported by web-based server/browser technology which delivers patient data to the physician's desktop. An HL7 based messaging system coordinates the acquisition and management of radiological image data and sends image keys to the clinical data repository. Access to the clinical chart browser currently provides radiology reports, laboratory data, vital signs and transcribed medical reports. A chart metaphor provides tabs for the selection of the clinical record for review. Activation of the radiology tab facilitates a standardized view of radiology reports and provides an icon used to initiate retrieval of available radiology images. The selection of the image icon spawns an image browser plug-in and utilizes the image key from the clinical repository to access the image server for the requested image data. The Spectrum system is collecting clinical data from five hospital systems and imaging data from two hospitals. Domain specific radiology imaging systems support the acquisition and primary interpretation of radiology exams. The spectrum clinical workstations are deployed to over 200 sites utilizing local area networks and ISDN connectivity.

The U.S. Department of Veterans Affairs is integrating imaging into the healthcare enterprise using the Digital Imaging and Communication in Medicine (DICOM) standard protocols. Image management is directly integrated into the VistA Hospital Information System (HIS) software and clinical database. Radiology images are acquired via DICOM, and are stored directly in the HIS database. Images can be displayed on low- cost clinician's workstations throughout the medical center. High-resolution diagnostic quality multi-monitor VistA workstations with specialized viewing software can be used for reading radiology images. DICOM has played critical roles in the ability to integrate imaging functionality into the Healthcare Enterprise. Because of its openness, it allows the integration of system components from commercial and non- commercial sources to work together to provide functional cost-effective solutions (see Figure 1). Two approaches are used to acquire and handle images within the radiology department. At some VA Medical Centers, DICOM is used to interface a commercial Picture Archiving and Communications System (PACS) to the VistA HIS. At other medical centers, DICOM is used to interface the image producing modalities directly to the image acquisition and display capabilities of VistA itself. Both of these approaches use a small set of DICOM services that has been implemented by VistA to allow patient and study text data to be transmitted to image producing modalities and the commercial PACS, and to enable images and study data to be transferred back.

An Image Management And Communication (IMAC) system adapted to the X-ray department at Sahlgrenska University Hospital has been developed using standard components. Two user demands have been considered primary: Rapid access to (display of) images and an efficient worklist management. To fulfil these demands a connection between the IMAC system and the existing Radiological Information System (RIS) has been implemented. The functional modules are: check of information consistency in data exported from image sources, a (logically) central storage of image data, viewing facility for high speed-, large volume-, clinical work, and an efficient interface to the RIS. Also, an image related database extension has been made to the RIS. The IMAC system has a strictly modular design with a simple structure. The image archive and short term storage are logically the same and acts as a huge disk. Through NFS all image data is available to all the connected workstations. All patient selection for viewing is through worklists, which are created by selection criteria in the RIS, by the use of barcodes, or, in singular cases, by entering the patient ID by hand.

The Medical Information, Communication and Archive System (MICAS) is a multi-vendor incremental approach to PACS. MICAS is a multi-modality integrated image management system that incorporates the radiology information system (RIS) and radiology image database (RID) with future 'hooks' to other hospital databases. Even though this approach to PACS is more risky than a single-vendor turn-key approach, it offers significant advantages. The vendors involved in the initial phase of MICAS are IDX Corp., ImageLabs, Inc. and Digital Equipment Corp (DEC). The network architecture operates at 100 MBits per sec except between the modalities and the stackable intelligent switch which is used to segment MICAS by modality. Each modality segment contains the acquisition engine for the modality, a temporary archive and one or more diagnostic workstations. All archived studies are available at all workstations, but there is no permanent archive at this time. At present, the RIS vendor is responsible for study acquisition and workflow as well as maintenance of the temporary archive. Management of study acquisition, workflow and the permanent archive will become the responsibility of the archive vendor when the archive is installed in the second quarter of 1998. The modalities currently interfaced to MICAS are MRI, CT and a Howtek film digitizer with Nuclear Medicine and computed radiography (CR) to be added when the permanent archive is installed. There are six dual-monitor diagnostic workstations which use ImageLabs Shared Vision viewer software located in MRI, CT, Nuclear Medicine, musculoskeletal reading areas and two in Radiology's main reading area. One of the major lessons learned to date is that the permanent archive should have been part of the initial MICAS installation and the archive vendor should have been responsible for image acquisition rather than the RIS vendor. Currently an archive vendor is being selected who will be responsible for the management of the archive plus the HIS/RIS interface, image acquisition, modality work list manager and interfacing to the current DICOM viewer software. The next phase of MICAS will include interfacing ultrasound, locating servers outside of the Radiology LAN to support the distribution of images and reports to the clinical floors and physician offices both within and outside of the University of Rochester Medical Center (URMC) campus and the teaching archive.

Purpose: To evaluate the clinical acceptability of personal computer (PC)-based PACS in daily practice of HIS-RIS- Modality-PACS coupling. Materials And Methods: We are developing a hospital-wide PC-based PACS which uses HIS/RIS network. PACS servers are connecting to HIS-RIS network and store images on hard disks and magneto-optical disk (MOD) juke-boxes from 8 FCRs and 3 CT scanners. We developed the remote control software of PACS terminal from HIS terminal for physicians' easy operation in the out-patent clinics. We investigated the network workloads and display time in the daily work. Clinicians' opinion was recorded on a 5-point scale for image quality, response, and function. Result: The network workload is under the limitation and display time is within twenty-six seconds. The quality was acceptable in 54%. The response was acceptable in 25%. The function of PACS was acceptable in 40%. Conclusion: PC-based HIS/RIS/PACS coupling has possibility of acceptance.

The role of the Project Leader in implementing a hospital-wide PACS is to lead the PACS Project Team. This involves ensuring that the structure of this team fairly represents the entire hospital, convening regular Project Team meetings, chairing these meetings and drawing up their agenda and documenting their minutes. Between meetings the Project Leader must follow up action items, verifying that they are being implemented in a timely fashion. The Project Leader is ultimately responsible for the decisions of the Project Team, in particular the assessment of the progress of the whole project, step by step. Experience of the implementation of the Hammersmith Hospital PACS is described in relation to the role of the Project Leader and Project Team.

The University of Pennsylvania Radiology Department is moving towards having a 'Filmless' Emergency Department (ED). Different modalities like computed radiography (CR), computed tomography (CT) and magnetic resonance imaging (MR) together with the IDXRad Radiology Information System (RIS) have been integrated into a centralized PACS. ED radiologists can view studies simultaneously from one of many four-monitor diagnostic workstations using various unread study worklists with patient demographics and examination details, along with prior images and reports. ED clinicians can immediately view the diagnostic quick readings, preliminary and final reports.

The 'Infrastructure Approach' is a PACS implementation methodology wherein the archive, network and information systems interfaces are acquired first, and workstations are installed later. The approach allows building a history of archived image data, so that most prior examinations are available in digital form when workstations are deployed. A limitation of the Infrastructure Approach is that the deferred use of digital image data defeats many data quality management functions that are provided automatically by human mechanisms when data is immediately used for the completion of clinical tasks. If the digital data is used solely for archiving while reports are interpreted from film, the radiologist serves only as a check against lost films, and another person must be designated as responsible for the quality of the digital data. Data from the Radiology Information System and the PACS were analyzed to assess the nature and frequency of system and data quality errors. The error level was found to be acceptable if supported by auditing and error resolution procedures requiring additional staff time, and in any case was better than the loss rate of a hardcopy film archive. It is concluded that the problem of data quality compromises but does not negate the value of the Infrastructure Approach. The Infrastructure Approach should best be employed only to a limited extent, and that any phased PACS implementation should have a substantial complement of workstations dedicated to softcopy interpretation for at least some applications, and with full deployment following not long thereafter.

The aim of this project was to implement the installation of a major hardware and software upgrade into the Hammersmith hospital-wide PACS without any loss of data and with minimum disruption to the continuing clinical service throughout the hospital. The extant 40 Gbyte RAID was replaced by a new 256 Gbyte centralized RAID short term storage (STS) device at the hub of the Hammersmith Hospital PACS. This change entailed a complete database transfer, as well as hardware and software conversions to each of the 168 PACS workstations in the hospital.

Recent wide adoption of the DICOM 3.0 standard by ultrasound equipment vendors created a need for practical clinical implementations of cardiac imaging study visualization, management and archiving, DICOM 3.0 defines only a logical and physical format for exchanging image data (still images, video, patient and study demographics). All DICOM compliant imaging studies must presently be archived on a 650 Mb recordable compact disk. This is a severe limitation for ultrasound applications where studies of 3 to 10 minutes long are a common practice. In addition, DICOM digital echocardiography objects require physiological signal indexing, content segmentation and characterization. Since DICOM 3.0 is an interchange standard only, it does not define how to database composite video objects. The goal of this research was therefore to address the issues of efficient storage, retrieval and management of DICOM compliant cardiac video studies in a distributed PACS environment. Our Web based implementation has the advantage of accommodating both DICOM defined entity-relation modules (equipment data, patient data, video format, etc.) in standard relational database tables and digital indexed video with its attributes in an object relational database. Object relational data model facilitates content indexing of full motion cardiac imaging studies through bi-directional hyperlink generation that tie searchable video attributes and related objects to individual video frames in the temporal domain. Benefits realized from use of bi-directionally hyperlinked data models in an object relational database include: (1) real time video indexing during image acquisition, (2) random access and frame accurate instant playback of previously recorded full motion imaging data, and (3) time savings from faster and more accurate access to data through multiple navigation mechanisms such as multidimensional queries on an index, queries on a hyperlink attribute, free search and browsing.

A cost effective, high capacity digital archive system is one of the remaining key factors that will enable a radiology department to eliminate film as an archive medium. The ever increasing amount of digital image data is creating the need for huge archive systems that can reliably store and retrieve millions of images and hold from a few terabytes of data to possibly hundreds of terabytes. Selecting the right archive solution depends on a number of factors: capacity requirements, write and retrieval performance requirements, scaleability in capacity and performance, conformance to open standards, archive availability and reliability, security, cost, achievable benefits and cost savings, investment protection, and more. This paper addresses many of these issues. It compares and positions optical disk and magnetic tape technologies, which are the predominant archive mediums today. New technologies will be discussed, such as DVD and high performance tape. Price and performance comparisons will be made at different archive capacities, plus the effect of file size on random and pre-fetch retrieval time will be analyzed. The concept of automated migration of images from high performance, RAID disk storage devices to high capacity, Nearline^R storage devices will be introduced as a viable way to minimize overall storage costs for an archive.

We have replaced our shared-media Ethernet and FDDI network with a multi-tiered, switched network using OC-12 (622 Mbps) ATM for the network backbone, OC3 (155 Mbps) connections to high-end servers and display workstations, and switched 100/10 Mbps Ethernet for workstations and desktop computers. The purpose of this research was to help PACS designers and implementers understand key performance factors in a high- speed switched network by characterizing and evaluating its image delivery performance, specifically, the performance of socket-based TCP (Transmission Control Protocol) and DICOM 3.0 communications. A test network within the UCLA Clinical RIS/PACS was constructed using Sun UltraSPARC-II machines with ATM, Fast Ethernet, and Ethernet network interfaces. To identify performance bottlenecks, we evaluated network throughput for memory to memory, memory to disk, disk to memory, and disk to disk transfers. To evaluate the effect of file size, tests involving disks were further divided using sizes of small (514 KB), medium (8 MB), and large (16 MB) files. The observed maximum throughput for various network configurations using the TCP protocol was 117 Mbps for memory to memory and 88 MBPS for memory to disk. For disk to memory, the peak throughput was 98 Mbps using small files, 114 Mbps using medium files, and 116 Mbps using large files. The peak throughput for disk to disk became 64 Mbps using small files and 96 Mbps using medium and large files. The peak throughput using the DICOM 3.0 protocol was substantially lower in all categories. The measured throughput varied significantly among the tests when TCP socket buffer was raised above the default value. The optimal buffer size was approximately 16 KB or the TCP protocol and around 256 KB for the DICOM protocol. The application message size also displayed distinctive effects on network throughput when the TCP socket buffer size was varied. The throughput results for Fast Ethernet and Ethernet were expectedly lower but the patterns were interestingly different from those for ATM. To achieve the optimum throughput in a TCP-based high-speed switched medial imaging network, the size of the TCP socket buffer is the most important parameter to optimize. If the DICOM 3.0 protocol is used, however, the performance gain by tuning system parameters is minimal, particularly if small files are used. Compared to socket-based TCP, the decrease in throughput caused by DICOM 3.0 protocol overhead is significantly larger in a high-speed switched network. This suggests that the protocol itself is the bottleneck in high-speed networks and that the protocol should be fine-tuned to take advantage of the services provided by such networks and not to duplicate them. To design a successful high-speed PACS network, it is important that bandwidth-demanding workstations and servers be on the same subnet and use the same technology so that no routing and data conversions are required.

This paper describes the incorporation of radiology teaching files within our existing filmless radiology Picture Archiving and Communications System (PACS). The creation of teaching files employs an intuitive World Wide Web (WWW) application that relieves the creator of the technical details involving the underlying PACS and obviates the need for knowledge of Internet publishing. Currently, our PACS supports filmless operation of CT, MRI, and ultrasound modalities, conforming to the Digital Imaging and Communications in Medicine (DICOM) and Health Level 7 (HL7) standards. Web-based teaching files are one module in a suite of WWW tools, developed in-house, for platform independent management of radiology data. The WWW browser tools act as liaison between inexpensive desktop PCs and the DICOM PACS. The creation of a teaching file is made as efficient as possible by allowing the creator to select the images and prepare the text within a single application, while finding and reviewing existing teaching files is simplified with a flexible, multi-criteria searching tool. This efficient and easy-to-use interface is largely responsible for the development of a database, currently containing over 400 teaching files, that has been generated in a short period of time.

Storage and archive systems are essential components of a Picture Archiving and Communication System (PACS) and become increasingly important in medical enterprise-wide information systems. Building a large-scale, long-term medical data storage system is complex, involving accurate identification of system requirements, judicious tradeoff between flexibility and complexity and effective execution of implementation plans. The proposed object-oriented software architecture fulfills the key system requirements, while maintains a high degree of flexibility for extending and customizing the framework in various medical data archive application areas. The complexity in developing such a framework has been proven manageable with various prototypes.

Purpose/Background: As medical centers purchase commercial PACS to integrate with their imaging equipment, the fact that each component is often purchased at different times and with differing levels of DICOM conformance may become a problem. When the PACS is installed, major incompatibilities are often discovered between DICOM conformant pieces of equipment. Commercial PACS vendors are sometimes reluctant to take responsibility for the final integration of these components. The purpose of this paper is to document the nature and extent of these problem at one medical center, to report on how we are trying to solve them, and recommendations for the future. Methods: A large commercial PACS (IMPAX^TM, Bayer Corporation, Agfa division) installed at the West Los Angeles VA medical Center was recently upgraded to be DICOM 3.0 conformant. Recently purchased DICOM conformant CT and MRI units and an older CT were already present. The problems in interfacing and integrating these imaging modalities with the upgraded DICOM-compliant PACS were investigated. Both software and hardware incompatibilities and human factors in bringing the different commercial vendors together were analyzed. Results: Major hardware and software incompatibilities were discovered. The use of DICOM-gateways is not always able to solve problems of incompatibility. The typical lag time between discovery of a problem and provision of a solution was 4 to 5 months. For more severe problems and also for a few of the lesser problems, resolution took over a year and is still ongoing. Repeated requests had to be made to the vendors to get together with us and solve these problems.

As radiology departments move closer to enterprise-wide electronic imaging systems, the importance of timely and accurate patient and exam information cannot be overstated. In addition, automated entry of patient and exam information into image acquisition modalities (e.g., computed radiography, CR) enhances technologist productivity, which is critical in high volume practices (e.g. a major trauma center). A difficulty arises, however, since the authoritative source of patient and exam information may be a Radiology Information System (RIS) or a Hospital Information System (HIS) packages its data differently than most imaging modalities. To bridge this gap, we have implemented an interface engine to affect the translation of Health Level Seven (HL7) messages from the RIS (or HIS) into Digital Imaging Communications In Medicine (DICOM) messages for seven CR systems (Fuji Medical Systems, Burbank, CA). Recent measurements suggest that manual technologist transcription of information into the CR unit takes 45.4 plus or minus 3.0 seconds. Since the introduction of the HL7/DICOM broker, the time has dropped to 5.4 plus or minus 0.1 seconds to produce the same information.

We present an efficient and daily used solution to archive the whole research image data set produced in our Multi-Modality Medical Imaging Research Laboratory. It is divided into two parts, (1) the IMASERV project whose purpose is to backup and distribute the data generated by the acquisition devices (3 PET scanners, 2 MRI, 3 SPECT systems and 1 phosphor imager system) and (2) the SILO project whose purpose is to archive and retrieve data issued from processing softwares available on the 80 Unix Workstations of the local area network. The two main technical difficulties were the computer architecture differences and to permit a uniform access to the data in the image formats available on our site. A database system was used to store information related to the description of the archived data. Softwares developments were necessary to push and pull data towards and from the archiving system. Automatic data transfer procedures between the imaging devices and the archiving system were implemented with a high quality control level. Data consulting and retrieval are available through suitable web interfaces.

Two sources of information play key roles in a collection of medical images such as computer tomographs, X-rays and histological slides, they are (1) textual descriptions relating to the image content and (2) visual features that can be seen on the image itself. The former are traditionally made by human specialists (e.g. histopathologists, radiographers, etc.) who interpret the image, and the latter are the inherent characteristics of images. This research program aims to study the architectural issues of a system which combines and interprets the information inherent in these two media to achieve automatic intelligent browsing of medical images. To give the research some practical significance, we applied the architecture to the design of the I-BROWSE system which is being developed jointly by the City University of Hong Kong and the Clinical School of the University of Cambridge. I- BROWSE is aimed to support intelligent retrieval and browsing of histological images obtained along the gastrointestinal tract (GI tract). Within such an architecture, given a query image or a populated image, a set of low level image feature measurements are obtained from a Visual Feature Detector, and with the help of knowledge bases and reasoning engines, the Semantic Analyzer derives, using an semantic feature generation and verification paradigm, the high level attributes for the image and furthermore automatically generates textual annotations for it. If the input image is accompanied with annotations made by a human specialist, the system will also analyze, combine and verify these two level of information, i.e., iconic and semantic contents. In the paper, we present the architectural issues and the strategies needed to support such information fusion process as well as the potentials of intelligent browsing using this dual- content-based approach.

The U.S. Army Medical Command, lead by the Brooke Army Medical Center, has embarked on a visionary project. The U.S. Army Virtual Radiology Environment (USAVRE) is a CONUS-based network that connects all the Army's major medical centers and Regional Medical Commands (RMC). The purpose of the USAVRE is to improve the quality, access, and cost of radiology services in the Army via the use of state-of-the-art medical imaging, computer, and networking technologies. The USAVRE contains multimedia viewing workstations; database archive systems are based on a distributed computing environment using Common Object Request Broker Architecture (CORBA) middleware protocols. The underlying telecommunications network is an ATM-based backbone network that connects the RMC regional networks and PACS networks at medical centers and RMC clinics. This project is a collaborative effort between Army, university, and industry centers with expertise in teleradiology and Global PACS applications. This paper describes a model and simulation of the USAVRE for performance evaluation purposes. As a first step the results of a Technology Assessment and Requirements Analysis (TARA) -- an analysis of the workload in Army radiology departments, their equipment and their staffing. Using the TARA data and other workload information, we have developed a very detailed analysis of the workload and workflow patterns of our Medical Treatment Facilities. We are embarking on modeling and simulation strategies, which will form the foundation for the VRE network. The workload analysis is performed for each radiology modality in a RMC site. The workload consists of the number of examinations per modality, type of images per exam, number of images per exam, and size of images. The frequency for store and forward cases, second readings, and interactive consultation cases are also determined. These parameters are translated into the model described below. The model for the USAVRE is hierarchical in nature. There are three levels to the model: (1) Network model of the Cable Bundling Initiative (CBI) network and base networks (CUITIN), (2) Protocol model, including network, transport, and middleware protocols, such TCP/IP and Common Object Request Broker Architecture (CORBA) protocols, and (3) USAVRE Application layer model, including database archive systems, acquisition equipment, viewing workstations, and operations and management. The Network layer of the model contains the ATM-based backbone network provided by the CBI, interfaces into the RMC regional networks and the PACS networks at the medical centers and RMC sites. The CBI network currently is a DS-3 (45 Mbps) backbone consisting of three major hubs, at Ft. Leavenworth, KS, Ft. Belvoir, VA, and Ft. McPherson, GA. The medical center PACS networks are 100 Mbps and 1 Gbps networks. The RMC site networks are 100 Mbps speeds. The model is very beneficial in studying the multimedia transfer and operations characteristics of the USAVRE before it is completely built and deployed.

The U.S. Army Medical Command, lead by the Brooke Army Medical Center, has embarked on a futuristic project that will revolutionize the practice of teleradiology in the DoD. The U.S. Army Virtual Radiology Environment (USAVRE) is a CONUS- based network that connects all the Army's major medical centers and Regional Medical Commands (RMC). The purpose of the USAVRE is to improve the quality, access, and cost of radiology services in the Army via the use of state-of-the-art medical imaging, computer, and networking technologies. The USAVRE contains multimedia-viewing workstations for static and dynamic modality cases. The storage and archiving systems are based on a distributed computing environment using Common Object Request Broker Architecture (CORBA) middleware protocols. Collaborations between archive centers and viewing workstations are managed by CORBA functions and multimedia object streams. The underlying telecommunications network is an ATM-based backbone network that connects to the RMC regional networks and PACS networks at medical centers and RMC clinics. The U.S. Army Information Systems Engineering Command (USAISEC) at Ft. Huachuca, AZ is responsible for the ATM backbone network to the RMC sites. This project is a collaborative effort between Army, university, and industry centers with expertise in teleradiology and Global PACS applications. This paper describes the telecommunications backbone network architecture for the Cable Bundling Initiative (CBI) being deployed by the Technology Integration Center. The U.S. Army U.S. Army Information Systems Engineering Command (USAISEC) at Ft. Huachuca, AZ has begun deployment of a 45 Mbps backbone network called the Cable Bundling Initiative (CBI). The CBI backbone is upgradable to OC-3 (155 Mbps) and OC-12 (622 Mbps) transmission links. A goal of CBI is to bundle the dedicated networks and support all MACOMs in the transition to a DISN ATM-based network in the future. The CBI is a candidate backbone for the USAVRE. The current connections of the CBI backbone include a triangle topology between Ft. Leavenworth, KS, Ft. Belvoir, VA, and Ft. McPherson, GA. There are T1 star topology connections to these three hubs to other Army bases in the hub regions. CBI is poised to upgrade to ATM-based technology and 155 Mbps (OC-3) links and higher. A requirements analysis is being conducted to determine which Army Medical Treatment Facilities need to connect to the CBI topology to exchange teleradiology services. An alternative architecture for the VRE backbone is MEDNet, a network of partially interconnected T1 links to Army bases. This paper describes the ATM technology components and fiber optic transmission links of the CBI backbone and the base-level ATM-based sub-networks. The Technology Integration Center (TIC) at USAISEC has been deploying the CBI backbone since 1995. Demonstrations of multimedia teleradiology will be been conducted over the CBI network. This paper presents data on the planned demonstrations.

The Digital Imaging Network -- Picture Archive and Communication System (DIN-PACS) procurement is the Department of Defense's (DoD) effort to bring military medical treatment facilities into the twenty-first century with nearly filmless digital radiology departments. The DIN-PACS procurement is unique from most of the previous PACS acquisitions in that the Request for Proposals (RFP) required extensive benchmark testing prior to contract award. The strategy for benchmark testing was a reflection of the DoD's previous PACS and teleradiology experiences. The DIN-PACS Technical Evaluation Panel (TEP) consisted of DoD and civilian radiology professionals with unique clinical and technical PACS expertise. The TEP considered nine items, key functional requirements to the DIN-PACS acquisition: (1) DICOM Conformance, (2) System Storage and Archive, (3) Workstation Performance, (4) Network Performance, (5) Radiology Information System (RIS) functionality, (6) Hospital Information System (HIS)/RIS Interface, (7) Teleradiology, (8) Quality Control, and (9) System Reliability. The development of a benchmark test to properly evaluate these key requirements would require the TEP to make technical, operational, and functional decisions that had not been part of a previous PACS acquisition. Developing test procedures and scenarios that simulated inputs from radiology modalities and outputs to soft copy workstations, film processors, and film printers would be a major undertaking. The goals of the TEP were to fairly assess each vendor's proposed system and to provide an accurate evaluation of each system's capabilities to the source selection authority, so the DoD could purchase a PACS that met the requirements in the RFP.