Performance evaluation and trade-off analysis are the central issues in the design of communication networks. Simulation plays an important role in computer-aided design and analysis of communication networks and related systems, allowing testing of numerous architectural configurations and fault scenarios. We are using the Block Oriented Network Simulator (BONeS, Comdisco, Foster City, CA) software package to perform discrete, event- driven Monte Carlo simulations in capacity planning, tradeoff analysis and evaluation of alternate architectures for a high-speed, high-resolution teleradiology project. A queuing network model of the teleradiology system has been devise, simulations executed and results analyzed. The wide area network link uses a switched, dial-up N X 56 kbps inverting multiplexer where the number of digital voice-grade lines (N) can vary from one (DS-0) through 24 (DS-1). The proposed goal of such a system is 200 films (2048 X 2048 X 12-bit) transferred between a remote and local site in an eight hour period with a mean delay time less than five minutes. It is found that: (1) the DS-1 service limit is around 100 films per eight hour period with a mean delay time of 412 +/- 39 seconds, short of the goal stipulated above; (2) compressed video teleconferencing can be run simultaneously with image data transfer over the DS-1 wide area network link without impacting the performance of the described teleradiology system; (3) there is little sense in upgrading to a higher bandwidth WAN link like DS-2 or DS-3 for the current system; and (4) the goal of transmitting 200 films in an eight hour period with a mean delay time less than five minutes can be achieved simply if the laser printer interface is updated from the current DR-11W interface to a much faster SCSI interface.

We have experienced that system reliability and performance become primary concerns when we moved our PACS from a research and development laboratory to a clinical environment. Providing consistent uptime builds user confidence and fast service accelerates system acceptance. The fault-tolerant mechanism implemented in our PACS has minimized system downtime. This paper considers the performance issues of an on-line multimodal PACS. The intent of conducting performance measurements on individual PACS components such as the radiological imaging devices, archive devices, host computers, and communication networks is to compare the throughput rates of these components in a real-time clinical environment as a basis of evaluating the overall throughput of a PACS system. Our primary goal is to identify the major factors that degrade the performance of a PACS, and to establish new strategies so that fast service with minimal delay is provided to the clinic. Approximately 200-Gbyte data transactions including CT, MR and CR images from our PACS were analyzed. Results showed that PACS throughput was limited by two major factors: (1) low-speed data interface used in the radiological imaging devices and archive devices; (2) network degradation due to heavy network traffic. We concluded that PACS throughput could be improved by well- designed network architecture and image-routing strategy. However, device-dependent low- speed data interface limited PACS performance.

The accurate prediction of image throughput is a critical issue in planning for and acquisition of any successful Picture Archiving and Communication System (PACS). Bottlenecks or design flaws can render an expensive PACS implementation useless. This manuscript presents a method for accurately predicting and measuring image throughput of a PACS design. To create the simulation model of the planned or implemented PACS, it must first be decomposed into principal tasks. We have decomposed the entire PACS image management chain into eight subsystems. These subsystems include network transfers over three different networks (Ethernet, FDDI and UltraNet) and five software programs and/or queues: (1) transfer of image data from the imaging modality computer to the image acquisition/reformatting computer; (2) reformatting the image data into a standard image format; (3) transferring the image data from the acquisition/reformatting computer to the image archive computer; (4) updating a relational database management system over the network; (5) image processing-- rotation and optimal gray-scale lookup table calculation; (6) request that the image be archived; (7) image transfer from the image archive computer to a designated image display workstation; and (8) update the local database on the image display station, separate the image header from the image data and store the image data on a parallel disk array. Through development of an event logging facility and implementation of a network management package we have acquired throughput data for each subsystem in the PACS chain. In addition, from our PACS relational database management system, we have distilled the traffic generation patterns (temporal, file size and destination) of our imaging modality devices. This data has been input into a simulation modeling package (Block Oriented Network Simulator-- BONeS) to estimate the characteristics of the modeled PACS, e.g., the throughput rates and delay time. This simulation provides estimates on what a radiological department can expect from a PACS in terms of throughput and delay.

Even though considerable progresses have been made with communication technology, one of the more difficult problems facing in installing a comprehensive clinically effective Image Management and Communication (IMAC) system for a hospital is the communication problem. Most existing systems are based on Ethernet or Token-ring net. Some of the newer systems are being installed using FDDL. All these systems have inherent problems like communication speed, control of bandwidth usage, or/and poor performance under heavy traffic. In order to overcome these difficulties, we are designing a complete IMAC system based on a novel network known as TeraNet, being developed at Center for Telecommunication Research, Columbia University.

The research and development efforts of several university and industry groups have brought digital imaging technologies into the practice of medicine. Radiographic images based on a digital data set can now be acquired, stored, communicated and presented for both primary interpretation and access by the referring physician. Moreover, conferences between a specialist and a primary care physician can be supported with audio and video links. A demonstration project at Washington University in collaboration with Southwestern Bell and NEC-America provides a testbed for deployment of ATM (Asynchronous Transfer Mode) broadband network technology supporting both LAN and WAN experiments in multimedia medical communications. A network based on four geographically dispersed ATM switches supports rapid display of high-resolution medical images, patient information, digital video and digitized real-time physiological signals at channel rates of 100 Mb/s. A prototype configuration of an Inquiry station is based on the NeXT computer with auxiliary displays for the medical images. Observations and preliminary performance results will be presented.

The technology for digital imaging in radiology and Picture Archiving and Communications Systems (PACS) has advanced at a rapid rate in the last five years. Local area fiber optic PACS networks which transfer digital images at over 200 Mega-bits per second are available now. Imaging equipment and viewing workstations technology currently provide excellent quality images for diagnosis and analysis for several modalities. In addition, the technology for network and telecommunications systems which transfer the images across country is also progressing rapidly. The opportunity exists to merge these technologies into a global PACS environment for digital radiology. This paper investigates the concepts of a global PACS and presents the user scenarios and system requirements for such systems. The concepts are described using user scenarios which describe diagnosis from remote workstations, interactive consultations when the experts are at remote locations, and conferencing between several experts located across country. The paper also describes the use of the new ACR-NEMA Version 3.0 standard for global PACS and the concept of national archive centers for research and education. The global PACS requirements are presented in operational, functional, and performance formats. A NSF sponsored project to develop a global PACS prototype system using the 45 Mbps NSFNET backbone network is summarized. The impact of such a system on digital radiology is explored.

A teleradiology system acquires radiographic images from one location and transmits them to one or more distant sites where they are displayed and/or converted to hardcopy film recordings. The long term goal of this research is to demonstrate that teleradiology systems can provide diagnostically equivalent results when compared to conventional radiographic film interpretation. If this hypothesis is proven, the following radiology tasks will be improved: (1) providing for primary interpretation of radiological images for patients in under served areas as well as other medical facilities; (2) integration of radiological services for multi- hospital/clinic health care provides consortiums (HMOs); (3) improving emergency service and intensive care unit coverage; (4) offering consulting-at-a-distance with sub-speciality radiologists; and (5) providing radiologists in the community or in rural areas immediate access to large academic centers for help in the interpretation of difficult and problematic cases. We are designing a high-speed, high-resolution teleradiology system between our level I medical center and several outlying medical centers within the metropolitan area. CT, MR and screen-film examinations will be digitized to 2 K or 4 K at the remote sites, transmitted to the central referral facility and sent to a laser film printer, reproducing the original film. The film can then be used for primary diagnosis, overreading/consultative purposes or for emergency room preparation. Inherently digital modality data (e.g. MR and CT) can be sent without digitization of the multi-format film is desired. A teleradiology system using a Wide Area Network (WAN) is to be connected to the following sites: (1) Olive View Medical Center; (2) Harbor General Medical Center; (3) UCLA Department of Radiological Sciences; and (4) two radiologist''s private residences. The wide area network (WAN) consists of a local carrier (GTE California Incorporated) and an inter-exchange carrier (US Sprint). Each affiliated hospital site is equipped with: (1) a radiographic film digitizer; (2) an interactive grayscale display workstation; (3) a computer system with teleradiology application software; (4) a local area network (LAN); (5) a LAN-WAN router; and (6) and dial-up (multiple switched N X 56 kbps lines) DS-1 WAN interfaces. The UCLA site is equipped with a local area image management network (PACS) for archiving, displaying, laser printed film hardcopy recordings, and WAN interfaces. The radiologists private residence is equipped with a grayscale station and a DS-0 56 kbps modem. We estimate the hardware costs at each remote site to be $DLR160,000, $DLR200,000 for the central referring facility, and under $DLR20,000 for the radiologist private residence.

There are 17 U.S. military medical treatment facilities scattered in the Republic of Korea to provide health care for approximately 40,000 U.S. troops and their dependents. Only three Army and one Air Force radiologist are assigned for radiological diagnosis. In some areas, report turn-around time can be as long as two weeks. An extensive teleradiology network is planned as part of an MDIS implementation program to improve the quality of radiological service overcoming the distance and time barriers.

The integrated combination of images from different modalities (signal, data, and text) yields a qualitative finding or helps with therapy. The communication medium ISDN (Integrated Services Digital Network) supports the integration of different communication forms to handle integrated information remotely. ISDN is therefore well suited to joint processing possibilities in the field of groupware for remote diagnosis, collective surgical planning, joint archiving, remote expert-guided training and teaching. This paper reports the functionality of groupware and the implementation of standards in specific software and conceptional levels to realize manufacturer independence for workstations. For practical reasons and for conformance tests the concept of journaling is used.

A dial-up, wide bandwidth, digital teleradiology system was implemented between Irwin Army Community Hospital (Fort Riley, Kansas), Munson Army Community Hospital (Fort Leavenworth, Kansas), and the University of Kansas (KU) Medical Center (Kansas City, Kansas). A laser film digitizer and a gray scale display system were used at Irwin and Munson Army Community Hospitals to digitize radiographic films and display digital images. A laser film printer at KU Medical Center generates a film hardcopy of the transmitted digital data and an interactive gray scale display is used to review the digital image data. The communication system consists of dial-up, switched, multiple 56,000 bits per second digital channels, transmitting digital image data in parallel. Conventional radiographic films, multiformat camera films, and laser printed films from multimodality imaging systems-- computed tomography (CT), magnetic resonance (MR), ultrasound (US), nuclear medicine (NM), digital fluorography, and phosphor plate computed radiography (CR)--have been successfully digitized, transmitted, and laser film recorded or gray scale displayed. It is concluded that the implemented dial-up, wide bandwidth, multiple 56,000 bits per second digital teleradiology system provides clinically acceptable image quality reproductions.

An on-going development effort to write a set of class-based libraries in C++ for image I/O and image processing is described. These new libraries will provide the following features to application programs based on them: derivable image classes, machine-architecture independence of data storage, fast, user-customizable image I/O, unlimited number of descriptive data items, arbitrary number of image dimensions, user-definable data types, optional automatic type conversion and compression, and support for accessing images using reference files and pipes. This highly-flexible and powerful capability is designed to meet the needs of workstation-based image processing and computer vision research in a multi- departmental university environment.

A hospital wide PACS project is currently under development at the University Hospital of Geneva. The visualization and manipulation of images provided by different imaging modalities constitutes one of the most challenging components of a PACS. Because there are different requirements depending on the clinical usage, it was necessary for such a visualization software to be provided on different types of workstations in different sectors of the PACS. The user interface has to be the same independently of the underlying workstation. Beside, in addition to a standard set of image manipulation and processing tools there is a need for more specific clinical tools that should be easily adapted to specific medical requirements. To achieve operating and windowing systems: the standard Unix/X-11/OSF-Motif based workstations and the Macintosh family and should be easily ported on other systems. This paper describes the design of such a system and discusses the extra cost and efforts involved in the development of a portable and easily expandable software.

A computer generated phantom has been produced to compare observer performance in detection tasks using various display systems. Simulated lesions of various size and contrast can be generated over a uniform or sloping grey level background. Many such lesions can be generated for a single phantom, and in each case, for example, for the set of all regions within the phantom, two such lesions are generated, one at the center of the region and a second at one of four orientations at a fixed distance from the central lesion. The observer performance test is therefore, for the ensemble of lesions and regions, to indicate which orientation occurs for each region. This phantom is based on that proposed in mammography. In order to evaluate the use of this phantom to assess image quality in general, an observer performance study was performed to assess image quality after data compression using 2 different techniques, capable of compressing the data by ratios of up to 50:1. Preliminary results indicate that this type of phantom is relatively efficient in assessing image quality and that a relatively low number of images, and therefore limited amount of time, is required of the observers. This phantom is part of a larger European project (ISCAMI) where a series of physical phantoms for testing PACS networks are being developed.

The viewing of radiological images has peculiarities that must be taken into account in the design of a compression technique. The images may be manipulated on a workstation to change the contrast, to change the center of the brightness levels that are viewed, and even to invert the images. Because of the possible consequences of losing information in a medical application, bit preserving compression is used for the images used for diagnosis. However, for archiving the images may be compressed to 10 of their original size. A compression technique based on the Discrete Cosine Transform (DCT) takes the viewing factors into account by compressing the changes in the local brightness levels. The compression technique is a variation of the CCITT JPEG compression that suppresses the blocking of the DCT except in areas of very high contrast.

A distributed digital archive system configured with dual archive devices (two archive servers, two database servers and two 680-Gbyte optical libraries) that provides fault-tolerant image archival has been implemented for the Radiology Department at UCLA. Digital images from various radiologic imaging devices are transmitted via Ethernet and FDDI networks to archive servers, where images are archived to optical disks and distributed to remote display stations or the print station via 1-Gbit/sec high-speed UltraNet network. The dual configuration of the system provides non-interrupt archive operations in the event of failure of any of the archive components. Once a failed device is detected, the system automatically re-configures itself so that all images are routed to the second equivalent device and archived. The global Ethernet network serves as a backup for the FDDI and UltraNet networks. In the even of FDDI or UltraNet failure, all images can be transmitted across the Ethernet. The system archives 1.5 to 2.0 Gbytes of data per day and provides inter-sectional image referencing throughout the department.

The requirements of a PACS database system (PACS DBS), which has been characterized to have large-volume image data, multi-media data and cross-modality reference, suggest the adoption of a distributed approach to PACS DBS design. This paper presents the design and the implementation of such a distributed PACS DBS. The design issues of a distributed PACS DBS include distributed transaction management for concurrency/recovery/security control, catalogue management for archiving fragmentation/location/replication transparency, and strategies for the distributed query processing. A prototype distributed PACS DBS that has been implemented and some implications from its evaluation will be discussed.

Our PACS design consists of clusters of computerized devices. At the center of each cluster is an intelligent machine termed a cluster controller. A cluster controller performs the following tasks: (1) receives images from capture computers, (2) extracts descriptive text information describing the received studies, (3) updates a network-accessible database management system, (4) determines the destination workstations to forward the newly generated studies, (5) automatically retrieves necessary comparison images from a distributed optical archive, (6) automatically corrects the orientation of computed radiography images, (7) archives new studies onto optical disk, (8) deletes images stored on remote capture computers upon successful image archival, and (9) services archive retrieval requests from remote workstations and other cluster controllers. The cluster controller is the only class of PACS computers that must communicate with all other types of PACS nodes (acquisition, other cluster controllers, radiologist''s workstation, referring physician''s workstation, printing stations, and database servers). This paper presents details of the logical partitioning of computers into clusters and the central role of the cluster controller machines in image acquisition, image routing, optical archive management, and system reliability.

In today''s economy, it takes significant funds to establish a high-performance image archival and retrieval system for any image application. One cost effective approach is to build the system in multiple phases but there is concern that technology is advancing rapidly and the original system may not be able to take advantage of new features. The concept of an open- architecture modular, flexible and expandable system is an essential element to achieving a high-performance image archival and retrieval system within a realistic short period of time. This paper introduces a proposal for a modular, flexible, and expandable image archival and retrieval open-architecture system to stimulate discussion and thinking. It will cover the following areas: (1) data archival and retrieval requirements such as storage capacity and data management, (2) data communication and distribution requirements using local area networks and/or wide area networks, (3) the architectural requirements such as adopting industry standards for hardware and software, and (4) an example of such open-architecture system to demonstrate the feasibility of implementing a modular, flexible, and expandable high- performance image archival and retrieval system.

A fast, easy-to-use, manufacturer-non-specific archival display and filming system was assembled and tested in the clinical environment. In this system, character recognition software decodes patient identification information and the type of examination for the archival data base in a manner that is transparent to the operator. Images are stored on an optical disk jukebox and can be retrieved to soft display for review and/or reprinting. Preliminary clinical evaluations with the system connected to one and/or two CT scanners clearly indicate that such a concept can successfully replace and/or serve as a backup to conventional film libraries. It can easily be attached to multiple devices (currently three) that are manufactured by the same or different vendors. Technologist and physician responses to the system have been favorable.

A hospital-wide PACS project is currently under development at the University Hospital of Geneva. This system is based on an open architecture regrouping equipment from multiple vendors in a distributed topology. The image archival is organized in multiple locations geographically distributed in the hospital. These regional archives are logically linked together to provide a virtual access to all images generated from different imaging modalities. A standardized image storage format called PAPYRUS was designed based on ACR-NEMA definitions to unify the image storage and display functionalities. The PACS database is also fully integrated with the concurrent RIS and HIS. Images from different archive servers are hierarchically distributed to other temporary storage space on regional display servers. Clusters of workstations are regrouped around these local servers allowing a more efficient access to the images on local subnetworks. Special software tools were designed for the management of image distribution and maintenance of local storage space. In addition to pre- programmed and rule-based distribution of images to regional display servers, special requests can be posted by the users to initiate the transfer of additional image files from the archive servers to their local display server. The design and implementation of the system will be presented and methodological issues will be discussed. Results from preliminary simulations performed prior to the final implementation will be presented and compared to real measurements performed on the system in operation. The advantages and difficulties of implementing a distributed hierarchical storage of images will be reported.

Ten years of PACS experience have demonstrated that PACS needs film and film needs PACS, in order to fulfill all the imaging requirements of a contemporary hospital. This paper analyzes the major technological and functional differences between film and PACS and describes a hybrid PACS paradigm which combines the strengths of both film and PACS. A practical implementation of this hybrid PACS concept in an ultrasound department is briefly described.

One of the advantages which a picture archiving and communications system (PACS) is supposed to provide over film-based operation is improved performance in retrieving images. While is seems self-evident that this should be so, we wanted to verify this, and provide some time comparisons for the two methods. The experiment consisted of randomly selecting ultrasound (US) and computed tomography (CT) cases and determining how long it took to retrieve them at one of our PACS workstations, and to retrieve them in person from the film room. To simulate actual retrieval volumes, a total of 40 cases from current to six months old, 20 cases from the past year, and 10 cases over one year old, was selected. Results indicate that indeed, PACS retrieval can be faster than file room retrieval. However, the difference is less for recent cases than for older cases. For cases six or fewer months old, the workstation retrieval was about two and one-half minutes faster per case than the film file room. This time difference increased markedly when extended to the one year and over one year groups. This paper details the results of this study, and will also provide information about the reliability of the two archives.

In a radiological examination reading, radiologists usually compare a newly generated examination with previous examinations of the same patient. For this reason, the retrieval of old images is a critical design requirement of totally digital radiology using Picture Archiving and Communication Systems (PACS). To achieve the required performance in a PACS with a hierarchical and possibly distributed image archival system, pre-fetching of images from slower or remote storage devices to the local buffers of workstations is proposed. Image Retrieval Expert System (IRES) is a knowledge-based image retrieval system which will predict and then pre-fetch relevant old images. Previous work on IRES design focused on the knowledge acquisition phase and the development of an efficient modeling methodology and architecture. The goal of this paper is to evaluate the effectiveness of the current IRES design and to identify appropriate directions for exploring other design features and alternatives by means of a cognitive study and an associated survey study.

We describe the interfacing methods between our departments RIS (Radiological Information System) and PACS database. Sharing information has become an increasingly important issue for both systems. RIS study information should be available at the PACS workstation to complete the image description. Patient scheduling, admission and discharge information are critical to PACS for improving image routing and storage management algorithms. The design of the interface between the RIS and the PACS database involves the development of some standardization methods. These include the network communication protocol between the two database servers, the language mapping operations, the database result format and the error handling. The implementation of the interface is in two steps. In the first one, a subset of RIS data is automatically transmitted to the PACS database when some relevant events occur in the RIS. In a second phase, the user at the PACS workstation can directly initiate queries to the RIS. The PACS SQL server (UNIX based) sends SQL network requests as ASCII text to the RIS MUMPS database server (VAX based). The RIS server translates the SQL requests to MUMPS, executes the statements and returns ASCII results to the SQL server. We have been able to exchange information between the RIS and the PACS display programs. RIS text reports can be displayed on PACS workstations. RIS information can be used into pre-fetch algorithms.

Multidimensional image data are becoming increasingly common in biomedical imaging. Three-dimensional visualization and analysis techniques based on three-dimensional image data have become an established discipline in biomedicine. Some imaging problems generate image data of even higher dimensions. It often become necessary, rather than just convenient, to consider the higher-dimensional data as a whole to adequately answer the underlying imaging questions. In spite of this established need for convenient exchange of image and image-derived information, no exchange protocols are available that adequately meet the needs of multidimensional imaging systems. This paper describes an exchange protocol that has been designed after careful consideration of the common requirements of methodologies for visualization and analysis of multidimensional data. It is based on and is a generalization of the widely accepted ACR-NEMA standards specified for two-dimensional images. It is implemented and actively being used in a data-, application-, and machine-independent software environment, being developed in the authors'' department, for the visualization and analysis of multidimensional images.

The PACS of an image management and communications system (IMACS) must have a communications infrastructure that supports not just the transmission of images, but the transmission of images with minimal latency while utilizing the bandwidths available in emerging high-speed networks that provide transfer rates between 100 Mbps and 2.3 Gbps. At the Bowman Gray School of Medicine, the ACR/NEMA Digital Imaging and Communication (DICOM) Document has been modified to use standard networks and protocols in lieu of the lower layers of the ACR/NEMA document, providing valuable insight into the use of the ACR/NEMA upper layers. This paper compares the performance of three communications mechanisms that use TCP/IP and XTP for data transport. The first application uses a communication paradigm that has been proposed for the new version of ACR/NEMA. Even though DICOM version 3 does not support the emerging high-performance protocols, XTP is also used in this mechanism. The other two applications employ a communication paradigm used by PAPYRUS, in which the actual transfer syntax (often called ''ACR/NEMA logical'') is used with standard network protocols that move data within application programs. One utilizes transfers from an application program, and the other uses standard file system transfers.

Picture Archiving and Communication Systems (PACS) is an integration of digital image formation in a hospital, which encompasses various imaging equipment, image viewing workstations, image databases, and a high speed network. The integration requires a standardization of communication protocols to connect devices from different vendors. The American College of Radiology and the National Electrical Manufacturers Association (ACR- NEMA) standard Version 2.0 provides a point-to-point hardware interface, a set of software commands, and a consistent set of data formats for PACS. But, it is inadequate for PACS networking environments, because of its point-to-point nature and its inflexibility to allow other services and protocols in the future. Based on previous experience of PACS developments in The University of Arizona, a new communication protocol for PACS networks and an approach were proposed to ACR-NEMA Working Group VI. The defined PACS protocol is intended to facilitate the development of PACS''s capable of interfacing with other hospital information systems. Also, it is intended to allow the creation of diagnostic information data bases which can be interrogated by a variety of distributed devices. A particularly important goal is to support communications in a multivendor environment. The new protocol specifications are defined primarily as a combination of the International Organization for Standardization/Open Systems Interconnection (ISO/OSI), TCP/IP protocols, and the data format portion of ACR-NEMA standard. This paper addresses the specification and implementation of the ISO-based protocol into a PACS prototype. The protocol specification, which covers Presentation, Session, Transport, and Network layers, is summarized briefly. The protocol implementation is discussed based on our implementation efforts in the UNIX Operating System Environment. At the same time, results of performance comparison between the ISO and TCP/IP implementations are presented to demonstrate the implementation of defined protocol. The testing of performance analysis is done by prototyping PACS on available platforms, which are Micro VAX II, DECstation and SUN Workstation.

Currently available PACS networks rely on Ethernet and TCP/IP or a related standard for image transfer. We have implemented a hardware interface which is fully compliant with the ACR-NEMA 2.0 protocol, both hardware and software. The signalling bandwidth of the ACR-NEMA parallel interface is considerably higher than for Ethernet over coaxial cable. The relative performance of these two standards for a point to point link is of great interest in determining the feasibility of using the ACR-NEMA 2.0 standard for device interfacing to a larger PACS network. To compare the performance we have put together two systems, each of which supports both TCP/IP over Ethernet and ACR-NEMA 2.0. The first system is based on a PC/AT platform using the ISA bus. The second is a UNIX system based on a VMEbus platform. Each of these systems has a commercially available Ethernet interface running a commercially available TCP/IP implementation. Each system also has a DeJarnette Research Systems ACR-NEMA 2.0 interface (ANSIF). In addition, the ISA bus system also supports the Siemens PACSNet standard. Performance measurements were made for both Ethernet protocols and ACR-NEMA 2.0 between similar platforms and dissimilar platforms. Comparison of these throughput results will allow direct comparison of the competitiveness of the ACR-NEMA 2.0 standard with Ethernet. Results of the performance measurements are discussed in detail. The implications of these relative performance measurements for the future of the ACR-NEMA standard will be examined as well.

A summary of the CEN TC 251/4 ''Medical Imaging and Multi-Media'' activities will be given. CEN is the European standardization institute, TC 251 deals with medical informatics. Standardization is a condition for the wide scale use of health care and medical informatics and for the creation of a common market. In the last two years, three important categories-- namely, the Commission of the European Communities with their programs and the mandates, the medical informaticians through their European professional federation, and the national normalization institutes through the European committee--have shown to be aware of this problem and have taken actions. As a result, a number of AIM (Advanced Informatics in Medicine), CEC sponsored projects, the CEC mandates to CEN and EWOS, the EFMI working group on standardization, the technical committee of CEN, and the working groups and project teams of CEN and EWOS are working on the subject. On overview of the CEN TC 251/4 ''Medical Imaging and Multi-Media'' activities will be given, including their relation to other work.

Comparisons of imaging systems through observer performance studies involve the selection of images, including those with subtle abnormalities. The selection of subtle abnormalities is typically performed in a subjective manner and is not quantitative. To assess the validity of such a selection process, we analyzed 163 ROC data sets. The difference in observer performance summary indices (Az) between ''typical'' and ''subtle'' cases were evaluated for each set. Average difference was found to be 0.10 (Az ''typical'' - Az ''subtle''), and in only 4 instances of 163 comparisons, observer performance for the subtle cases increased by 0.01 ((Delta) Az). We conclude that although it is not quantitative, this approach to the selection of subtle abnormalities is appropriate for imaging systems comparisons using ROC- type studies.

A number of user-centered methods for designing radiology workstations have been described by researchers at Carleton University (Ottawa), Georgetown University, George Washington University, and University of Arizona, among others. The approach described here differs in that it enriches standard human-factors practices with methods adapted from ethnography to study users (in this case, diagnostic radiologists) as members of a distinct culture. The overall approach combines several methods; the core method, based on ethnographic ''stream of behavior chronicles'' and their analysis, has four phases: (1) first, we gather the stream of behavior by videotaping a radiologist as he or she works; (2) we view the tape ourselves and formulate questions and hypothesis about the work; and then (3) in a second videotaped session, we show the radiologist the original tape and ask for a running commentary on the work, into which, at the appropriate points, we interject our questions for clarification. We then (4) categorize/index the behavior on the ''raw data'' tapes for various kinds of follow-on analysis. We describe and illustrate this method in detail, describe how we analyze the ''raw data'' videotapes and the commentary tapes, and explain how the method can be integrated into an overall user-centered design process based on standard human-factors techniques.

The transition from analog film-based to electronic imagery management in radiology departments and clinics requires accurate projection of storage and network capabilities. We previously developed a method of estimating from film usage, static storage requirements for the central archive systems. Planning requires us to project: (a) intermediate (e.g. magnetic disk) storage needs for local area networks and workstations serving clinical care areas, and (b) data transmission rates needed to deliver data to nodes on the network, given the expected dataflows empirically derived in our previous studies. The majority of medical imagery is currently stored on 14'' by 17'' film, each film representing about 6 Mbytes of storage at current standard digitization resolutions. With such applications, initial projections of data rates can be made using records of film usage available in most departments. However, it is also necessary to perform a survey of film users to determine usage of new imagery modalities and comparison imagery requirements in the areas to be served by the network.

The laborious PACS development process can be supported by means of modeling and simulation. In this paper we illustrate the usefulness of these techniques by means of a simulation study of the PACS as developed at the University Hospital of Geneva. The paper summarizes the characteristics of the Geneva PACS, discusses the assumptions made, and finally presents some first simulation results.

An extremely important aspect of implementing and maintaining a reliable clinical PACS is the formation of structured quality control and user education programs. Often when systems are deployed for clinical use, improvements and usage come about slowly. This is often due to a lack of sufficient clinical feedback and/or lack of enthusiasm by the developers to make recommended changes. The PACS quality control program at UCLA is comprised of six coordinators who keep in close contact with the clinical operation. Their responsibilities include: (1) checking the status of all system processes, (2) educating PACS users on protocols and program interaction, (3) conducting user surveys, (4) checking image quality, (5) checking workstation study availability, and (6) periodically generating error and performance reports. The quality control program has been invaluable in improving system functionality and reliability. Protocols have been developed which result in daily reports summarizing the acquisition, archive, and workstation forwarding status of all studies from 4 CT, 3 MR, and 2 CR systems. The reliability of the acquisition and archiving subsystems of our PACS is over 98 percent. The clinical coordinators also study the clinical behavioral effects of PACS on its users.

Instead of working on a full-scale PACS environment in one step, we have chosen for an approach of total upgradability. Different PACS islands are integrated. The focus of this paper is the ultrasound (US) management unit. PACS is focused as a communication service. The US management unit is based on a Mac Computer on which an object-oriented user interface is implemented. The database ensures the data consistency locally, but has also SQL query possibilities for other segments. The local architecture is based on a hierarchical storage including erasable optical disks. One or more screens allow an overview image display. Interaction with the RIS is realized through a serial link. The ultrasound units are connected by video signal. An evaluation for wireless connection between the ultrasound unit and PACS environment is taking place for the emergency and intensive care department. This unit in full clinical use is serving as a study base between the radiologist and new types of reporting possibilities offered by PACS. The system proved to be of greater clinical use and less expensive than the traditional system. The system is described with also the knowledge base used for the user interface handling, network, and storage handling.

A neuroradiology PACS module developed at UCLA was evaluated in the clinical environment from February 1990 to July 1991. The results of this clinical operation will be used to assess its clinical value. The overall evaluation focused on three aspects: (1) image delivery performance, (2) system availability, and (3) user acceptance. This study evaluates the image delivery performance by analyzing the elapsed time spent in each modularized task in both the film-based system and the PACS system. The system availability was examined by observing the down-time occurrence and up-time probability of individual hardware components in the PACS module. The user acceptance was verified through the display workstation.

PACS research and development efforts at the University of Florida, Department of Radiology have been directed solely toward solving clinical problems with an objective of incorporating successful products or improving the functionality, performance, and reliability of the system. This paper describes the current network and system, experiences with system upgrades and changed, quality control measures used to verify that the system is operational, and current works in progress.

With the proliferation of CT travel to distant scanners to review images before their patients can be released. We designed a fiber-optic broadband video system to transmit images from seven scanner consoles to fourteen remote monitoring stations in real time. This system has been used clinically by radiologists for over one years. We designed and conducted a user survey to categorize the levels of system use by section (Chest, GI, GU, Bone, Neuro, Peds, etc.), to measure operational utilization and acceptance of the system into the clinical environment, to clarify the system''s importance as a clinical tool for saving radiologists travel-time to distant CT the system''s performance and limitations as a diagnostic tool. The study was administered directly to radiologists using a printed survey form. The results of the survey''s compiled data show a high percentage of system usage by a wide spectrum of radiologists. Clearly, this system has been accepted into the clinical environment as a highly valued diagnostic tool in terms of time savings and functional flexibility.

As hospital radiology departments and ICUs make plans to use or expand the usage of digital data outside of radiology, the need to assess the requirements of the potential recipients in the ICUs has become more important. The present operations in an intensive care unit that uses digital viewing instead of film has been compared with a unit that does not. The difference in time between X-ray exposure and final image viewing was determined in both settings and compared. In this preliminary study significant differences were found between the two units. The odds of having exam results actually accessed by the ordering physician in an hour or less were 91/2 times greater for the unit with digital viewing capability than for the one without it.

After several years of continual operation, the utility of digital viewing stations was investigated by distributing questionnaires to past and present users. The results of the questionnaire indicated that the respondents preferred using the workstations over handling film. For evaluation of line placements, chest tubes and pleural effusions, softcopy display was preferred over hardcopy. However, for analysis of air space disease and pneumothorax, images displayed on the workstation were not perceived to be as useful as standard hardcopy.

If softcopy display systems are to be used for primary interpretation, there are a number of problems, in addition to image quality issues, that must be addressed. In order to evaluate these issues in a clinical setting, a system was developed to allow chest radiologists to routinely utilize softcopy displays in the interpretation of images from selected intensive care units. Computed radiography studies were acquired, stored on high-speed disk systems, and made available for review on high-resolution softcopy displays. Images from relevant prior studies were made available for comparison. Links to the department''s radiology information system were established to provide routine access to corollary information, including reports from related examinations. The initial results of this experiment are encouraging, but a number of problems and pitfalls were identified which must be addressed in the design of future systems.

In 1987, a prospective study was done in an intensive care unit comparing a digitized-film picture archiving and communication systems (PACS) with a conventional-film management and display system. At that time a 512 X 512 X 8 bit display station was used only for about 30 of the cases. The display was upgraded to a 2560 X 2048 X 12 bit unit and a second prospective study again showed utilization for 30 of the cases. Then the image resolution was reduced to 1024 X 1024 in order to decrease the display writing time from 12 to 3.5 seconds. The user interface was also simplified. A recent prospective study showed that the display was utilized for all of the cases. For certain diagnostic categories such as ''check tube position'' it was used for primary reading while for more complicated categories such as ''rule out pneumonia'' it was used mainly for review.

Graphical User Interfaced (GUI) workstations are now available from commercial vendors. We recently installed a GUI workstation in our nuclear medicine reading room for exclusive use of staff and resident physicians. The system is built upon a Macintosh platform and has been available as a DELTAmanager from MedImage and more recently as an ICON V from Siemens Medical Systems. The workstation provides only display functions and connects to our existing nuclear medicine imaging system via ethernet. The system has some processing capabilities to create oblique, sagittal and coronal views from transverse tomographic views. Hard copy output is via a screen save device and a thermal color printer. The DELTAmanager replaced a MicroDELTA workstation which had both process and view functions. The mouse activated GUI has made remarkable changes to physicians'' use of the nuclear medicine viewing system. Training time to view and review studies has been reduced from hours to about 30-minutes. Generation of oblique views and display of brain and heart tomographic studies has been reduced from about 30-minutes of technician''s time to about 5-minutes of physician''s time. Overall operator functionality has been increased so that resident physicians with little prior computer experience can access all images on the image server and display pertinent patient images when consulting with other staff.

Radiologist productivity has been a major factor limiting the acceptance of soft copy image displays. This situation is particularly acute in projection radiography. Three principal factors contribute to radiologist productivity: image quality, the user interface and system response time. These factors are not completely separable. In particular, it is difficult to optimize the user interface when system response times are a major limiting factor. A newly developed high speed multi-screen display system has enabled further refinement of the user interface into the subsecond regime. The present work focuses on optimizing control of zoom, roam and gray scale windowing. We describe the clinical optimization protocol and the user interface evolved therefrom. We also discuss the subjective impressions of the radiologists involved.

Software engineering principles suggest that complex software systems are best constructed from independent, self-contained modules, thereby maximizing the portability, maintainability and modifiability of the produced code. This principal is important in the design of medical imaging workstations, where further developments in technology (CPU, memory, interface devices, displays, network connections) are required for clinically acceptable workstations, and it is desirable to provide different hardware platforms with the ''same look and feel'' for the user. In addition, the set of desired functions is relatively well understood, but the optimal user interface for delivering these functions on a clinically acceptable workstation is still different depending on department, specialty, or individual preference. At the University of Washington, we are developing a viewing station based on the IBM RISC/6000 computer and on new technologies that are just becoming commercially available. These include advanced voice recognition systems and an ultra-high-speed network. We are developing a set of specifications and a conceptual design for the workstation, and will be producing a prototype. This paper presents our current concepts concerning the architecture and software system design of the future prototype. Our conceptual design specifies requirements for a Database Application Programming Interface (DBAPI) and for a User API (UAPI). The DBAPI consists of a set of subroutine calls that define the admissible transactions between the workstation and an image archive. The UAPI describes the requests a user interface program can make of the workstation. It incorporates basic display and image processing functions, yet is specifically designed to allow extensions to the basic set at the application level. We will discuss the fundamental elements of the two API''s and illustrate their application to workstation design.

In this paper, we describe the development strategies for integrating hard-copy output devices into a comprehensive picture archiving and communication systems (PACS). These strategies were created in response to a need for selective film-based hard copy in a filmless clinical ultrasound section.

An interface between a Radiology Information System (RIS) and a Picture Archiving and Communication System (PACS) has been designed in Victoria. A prototype has been developed for a Health Care Systems Inc. RIS and a Siemens PACS. The main design objective for this prototype was to create a generic RIS-PACS interface. The portability of the interface is ensured by its modularity, the utilization of a standardized language and communication protocols and the face that no changes were required to either RIS or PACS. In the case of the prototype the communication of data is unidirectional, i.e. 36 data elements are exported from RIS to PACS. Although full integration of these two information systems in the Medical Imaging department appears more desirable than interfacing, an efficient, slim RIS-PACS interface is usually the more feasibly option. Current PACS projects have to cope with the restrictions of today''s RIS and PACS, which are often not transaction-oriented, rarely use modern database models and have been designed neglecting interfacing considerations. The paper summarizes interconnection experiences of four other international projects, outlines the design of the Generic RIMS-PACS Interface and evaluates the experience with the Victoria prototype.

We developed a formal method to organize and to concisely document a radiological diagnoses from a PACS workstation which is acceptable to referring physicians and radiologists. In our PACS environment, images are viewed from a high resolution 2.5 K X 2 K workstation. Based on the requirements for report documentation, tools to easily build, organize and playback a radiological report were developed. A report typically includes multimedia data made up of images, text, graphics and voice. A voice server integrated to the workstation allows a very natural way to include voice reports. Full editing functions are available for dictation. The presentation of the diagnostic report to the referring physicians consists of replaying a script generated automatically from the radiologists actions at the workstation and include displaying the pertinent images with the associated image processing operations, graphics ROI and voice reports. A high level data representation model is defined to handle the representation of both multimedia data (relevant images, graphic pointers to regions of interest, window text and voice report) and the associated sequence of operations that will allow convenient presentation of a report at the workstation. The proposed documentation tool will improve the presentation of cases already viewed and is useful for both the referring physicians and for the radiologists when viewing future studies of the patient.

We have integrated several classes of intelligent software modules into our existing PACS. The first class can be categorized as general decision-making routines and include study routing, comparison image retrieval, study migration, storage management, and fault tolerance algorithms. These algorithms are rule based. The second class include user interface routines that allow radiologist to operate a PACS workstation within their own knowledge domain. Thus, window and level operations are translated into options to improve visualization of soft tissue or bone. The third class of intelligent modules are performance optimization algorithms and include optimizing file retrieval speed from a distributed PACS, optimizing T1-T2 contrast for various brain lesions in MR images, and automatic contrast and brightness adjustment of images displayed on PACS workstations. The effects of these intelligent algorithms on our PACS operation has been improved system reliability and study availability, and has made workstations easy to use. Currently, we are developing a general framework for the development of a knowledge-based multi-media distributed database management system for PACS which will support abstract and approximate query processing.

Our Medical Center is faced with the problem to design a cable wiring system today, install it by the middle 1990''s, and allow for upgrades and enhancements for the next ten to fifteen years. The cable plant must be able to support functions and activities which are poorly defined today, but will include the hospital information system (HIS), a future picture archiving and communication system (PACS), and possibly an electronic patient chart with integrated image data. The cable plant must also connect to a future campus wide Medical Network of the Indiana University Medical center with the first component being a positron emission tomography system (PET) located approximately one mile away. To meet these goals the proposed cable plant will be a structured wire cabling system following existing and proposed standards for building wiring architectures including the Electronics Industries Association (EIA) and the Telecommunications Industries Association (TIA) 568 Commercial Building Wiring Standard and the proposed Federal Telecommunication Standard 1090. The structured wiring system approached has been evaluated with emphasis on the present and future network topologies that can be implemented, the type and size of fiber optic cable to install, and the need to install fiber optic cable to individual workstations.

The implementation of Picture Archival and Communication Systems (PACS) within the contemporary radiology department is a complex procedure. Given the complexity of the total PACS environment with its advanced technology requirements, most sites will find it impractical to implement a filmless department with a total PACS solution at this time. Although many technical problems make a total PACS solution impractical in many situations, small, clinically useful, partial PACS (''miniPACS'') can be developed now and can provide experience for future development of more complete PACS. Planning for PACS and creating a supporting infrastructure are important and complex procedures. This paper describes an analysis performed as a neuroradiology PACS system was designed and implemented. Network bandwidth and image storage were evaluated; interfaces were specified; databases were designed; and plans were made to accommodate physical equipment requirements.

The Wright-Patterson USAF Medical Center (WPMC) is one of three DOD medical centers currently engaged in the installation of the ''Medical Diagnostic Imaging Support'' (MDIS) system. MDIS is a comprehensive hospital wide ''Picture Archiving and Communications System'' (PACS). In defining the requirements for the MDIS system it quickly became apparent that detailed operational information was necessary to adequately define the system requirements. This information included an understanding of the needs of our customers, workload data, current and expected future imaging capabilities, current problems encountered in meeting the needs of our customers, and the resources available to respond to these problems. Armed with this information the MDIS technical development team was able to define a detailed functional description for the MDIS system. This functional description was presented to industry in the form of a competitive solicitation with the subsequent submission of proposals by interested vendors. Following an extensive review of these proposals and the results of live benchmark evaluations, a contract was awarded to Loral Western Development Laboratory on September 27, 1991. This paper presents an operational overview of WPMC, a description of the deficiencies in our current image management system, and a discussion of the PACS to be installed this summer.

It is observed that the dataprotection aspects of PACS have hardly been explored, while for Hospital Information Systems (HIS) considerable attention has been paid to these aspects. In the dataprotection domain around HIS we distinguish the aspects data integrity (safety), usage integrity (confidentiality, privacy) and availability. A wide range of measures (dealing with hardware, software or organization/procedures) has been suggested to achieve adequate protection. For each system a selection of the suggested measures has to be made, where risk is balanced against costs. It seems that it is taken for granted that the usual measures as implemented for HIS will also be adequate for PACS. In this paper dataprotection aspects for PACS are discussed. Apart from similarities with the HIS situation also differences will be identified. In particular safeguarding the image databank requires special measured. Dataprotection should not be an add-on of the system, it should be taken into account in the design phase. As an illustration two alternatives for the organization of the image databank will be considered from the point of view of dataprotection. The one being consecutive historical storage of images, the other the clustering of images of the same patient on the same volume in the storage system.

Development of automated medical information systems requires knowledge about the utilization of the system both for design and as feedback during the development process. One of the significant advantages of computer systems is that they can provide an automatic log of system utilization. This paper reports the first study of utilization data collected automatically in an integrated image environment. Images are routinely used in clinical care. Selected images are shown repeatedly for teaching demonstration and reference at locations throughout the hospital. Images are viewed both by the collecting service when producing procedure reports and consultations as well as by the treating physicians from other hospital services. Medical departments have developed image collection policies to meet their internal needs and the needs of treating physicians.

Images are subjects to various types of noises. Noise carries no useful information and deteriorates image quality. Most of the studies on noise are for projectional radiography; few are on the effect of 3-D images. In this paper we studied the effect of noise on the 3-D surface display and compared three techniques to smooth out noise. The three smoothing techniques used are averaging, median filter, and Butterworth filter. The performances of these techniques are compared. For averaging and median filter techniques, smoothing is performed on each slice of the 3-D data set. The Butterworth filter is performed on the 3-D cosine transform of the data. Then the 3-D surface is reconstructed from these smoothed data. It was found that all of the three techniques can smooth out the noise effectively. The effect of noise on the 3-D image is enhancement of the slicing artifacts. The 3-D Butterworth filter can eliminate the slicing artifacts. The blurring artifact which is prevalent in 2-D smoothing do not show up in the 3-D display.

Clinical acceptance of PACS is predicated upon maximum uptime. Upon component failure, detection, diagnosis, reconfiguration and repair must occur immediately. Our current PACS network is large, heterogeneous, complex and wide-spread geographically. The overwhelming number of network devices, computers and software processes involved in a departmental or inter-institutional PACS makes development of tools for network and resource management critical. The authors have developed and implemented a comprehensive solution (PACS Network-Resource Manager) using the OSI Network Management Framework with network element agents that respond to queries and commands for network management stations. Managed resources include: communication protocol layers for Ethernet, FDDI and UltraNet; network devices; computer and operating system resources; and application, database and network services. The Network-Resource Manager is currently being used for warning, fault, security violation and configuration modification event notification. Analysis, automation and control applications have been added so that PACS resources can be dynamically reconfigured and so that users are notified when active involvement is required. Custom data and error logging have been implemented that allow statistics for each PACS subsystem to be charted for performance data. The Network-Resource Manager allows our departmental PACS system to be monitored continuously and thoroughly, with a minimal amount of personal involvement and time.

As presentation of medical radiographic images on soft-copy displays (cathode ray tubes) becomes increasingly prevalent in electronic radiography, methods of quality assurance must be developed to ensure that radiologists can effectively transfer film-based reading skills. Luminance measurements provide the basis for evaluating the state of soft-copy displays. An integrated approach has been implemented at Mallinckrodt Institute of Radiology (MIR) which facilitates measurement of geographically distributed soft-copy displays with centralized data logging, performance tracking and calibration. MIR''s central radiology image manager (RIM) exercises the display station which drives the monitor, harvests the measurement data, stores the results and submits the resulting data for additional processing. The luminance measurements are collected by a small, portable photometric instrument designed at MIR that includes a serial port which is accessed via local area terminal service (LAT) supported by the RIM. The design details of the photometric instrument and example luminance characteristics of several soft-copy displays used at the Mallinckrodt Institute are presented in this paper.

The importance of remote access to both radiological images and medical information has stimulated many demonstration projects utilizing a variety of telecommunications providers'' offerings. Teleradiology over modest cost channels can achieve adequate response times using a combination of narrow-band ISDN and data compression. A demonstration project, developed in collaboration with Southwestern Bell Technology Resources, Inc., utilizes the aggregate bandwidth of two B channels (achieving a rate of 120 kb/s) and a compression/ decompression implementation based on the JPEG block-oriented DCT approach. System response measurements for an Inquiry and Display Station accessing the Mallinckrodt Institute of Radiology''s (MIR) Radiology Image and Information Management (RIM) Testbed via the N-ISDN connection shows response times to be within 20 seconds. Viewing applications have been demonstrated at sites within St. Louis and at RSNA-90 in Chicago. Image quality metrics and observations are presented for a sample digitized film chest radiograph and computed radiographs of portable x-ray examinations vis a demonstration at the poster session.

Currently a large European project in the area of PACS research and prototyping is being performed. The EurIPACS project is subdivided into a dozen topics. One of these topics is involved in the integration of PACS, RIS, and HIS on conceptual, functional, and technical levels. Within the scope of this topic, which is called HIPIN (HIs/ris-Pacs INtegration), a prototype PACS-RIS/HIS interface will be specified, designed, realized, clinically implemented, and evaluated. The interface software, together with the computer configuration on which it runs, is called the HIPIN box. This HIPIN box will be designed in such a way that it can be applied (with minimal adaptations) to interface any RIS/HIS with any PACS in a multi-vendor environment. Two incarnations of the HIPIN box will be realized and will be implemented in the clinical routines of a Belgian and a German hospital. The functionality of the interface will be focused to support the clinical requirements in these hospitals. The HIPIN box will support communication facilities, integration of the textural databases, and data- protection measures. Furthermore, it will contain intelligent modules to control image file archival and migration (prefetching).

The University of Maryland Medical System (UMMS) is currently building a PACS network to link all its digital modalities (CT, MR, DSA and Ultrasound) together to provide four basic services: laser imager sharing, archiving, teleradiology and centralized resident review of all after hours cases. The objective is to provide the full digital contrast range for all modalities. Like most large hospitals UMMS has digital equipment from a variety of vendors. We have designed a system to connect modalities from Siemens and General Electric together with a variety of ultrasound manufacturers and several 3M laser imagers into a coherent network. The principal obstacles to such an effort are the lack of standardized network interfaces and the use of proprietary image data storage formats. Our goal is to use the ACR-NEMA file format and interface standard to the maximum extent practicable to resolve these difficulties. Since UMMS doe not have the financial resources or incentives to purchase a complete commercial PACS system we have opted to develop a lower performance but more affordable alternative to accomplish the above objectives. We will discuss the obstacles to interfacing with each manufacturer''s system(s), the solutions agreed upon and the resulting overall design.

Model-based reasoning techniques complement traditional image processing and pattern recognition to compensate for the ambiguities inherent in medical imagery data. In model- based reasoning, higher-order image features, such as inferred surface boundaries, are matched against parametric models of anatomy. The instantiated model can be used to predict approximate locations of other anatomical image features for further segmentation processing. The inferential power of this approach is limited by the accuracy with which anatomical models capture population statistics. One possible problem with a model-based approach is that it requires a large training samples to develop models. We show that selection of quasi- invariant features greatly reduces the population sample required to accurately model population distribution of features. Quasi-invariant features include certain ratios, angles and other functions of directly measurable image features that constrain each others values. When one observable is measured in the image, the likely values and locations of others are relatively constrained. For example, the ratios of lengths of phalanges has small variance across the population of all humans compared to the variance of each bone length individually. An experiment is presented on a population of 90 radiographs that supports this approach to segmentation for hand anatomy measures.

We have developed a system for routine teleradiology between a local clinic and a central hospital, where the dictation is digital, and sound data is transmitted in real-time over a digital 64 kbit/s network. The need to stock, transfer, and account for analog tapes is eliminated. The dictation is integrated in a program for viewing the patient images, so that the audio data is automatically connected with the current patient. The doctor uses a commercial workstation running UNIX, with image display on multiple X windows servers. The dictation is transcribed by the secretary who uses an ordinary PC equipped with a popular sound board. Sound data is transmitted over the teleradiology network during playback, and the text file containing the transcription is also connected with the patient.

This paper describes the design of a distributed database for a picture archiving and communications system (PACS). This system consists of several relatively cheap workstations and a large optical archive and distributed database modules. The communication between the modules uses the same network as used to connect the system to the clients. Existing, working, and available standards are used wherever possible. This makes the system open and accessible in a multivendor environment such as hospitals. The distributed approach makes the total system flexible, cheaper, faster, and when designed carefully, more reliable.

The unusual nature of sites chosen for hospital-wide PACS implementations and the very small number of proposed implementations make evaluation a complex task. The UK Department of Health is funding both the evaluation and implementation of a hospital-wide PACS. The Brunel University evaluation of the Hammersmith Hospital PACS has two main components: an economic evaluation of the costs and benefits of hospital-wide PACS installations and an exercise in monitoring the implementation process. This paper concentrates on the economic component.

The Medical Diagnostic Imaging Support (MDIS) system contract for federal medical treatment facilities was awarded to Loral/Siemens in the Fall of 1991. This contract places ''filmless'' imaging in a variety of situations from small clients to large medical centers. The MDIS system approach is a ''turn-key'', performance based specification driven by clinical requirements.

The Picture Archive and Communication System developed by Loral Western Development Laboratories and Siemens Gammasonics Incorporated utilizes an advanced, high speed, fault tolerant image file server or Working Storage Unit (WSU) combined with 100 Mbit per second fiber optic data links. This central shared file server is capable of supporting the needs of more than one hundred workstations and acquisition devices at interactive rates. If additional performance is required, additional working storage units may be configured in a hyper-star topology. Specialized processing and display hardware is used to enhance Apple Macintosh personal computers to provide a family of low cost, easy to use, yet extremely powerful medical image workstations. The Siemens LiteboxTM application software provides a consistent look and feel to the user interface of all workstation in the family. Modern database and wide area communications technologies combine to support not only large hospital PACS but also outlying clinics and smaller facilities. Basic RIS functionality is integrated into the PACS database for convenience and data integrity.

Siemens and Loral Western Development Labs have designed a Picture Archiving and Communication System capable of supporting a large, fully digital hospital. Its functions include the management, storage and retrieval of medical images. The system may be modeled as a heterogeneous network of processing elements, transfer devices and storage units. Several discrete event simulation models have been designed to investigate different levels of the design. These models include the System Model, focusing on the flow of image traffic throughout the system, the Workstation Models, focusing on the internal processing in the different types of workstations, and the Communication Network Model, focusing on the control communication and host computer processing. The first two of these models are addressed here, with reference being made to a separate paper regarding the Communication Network Model. This paper describes some of the issues addressed with the models, the modeling techniques used and the performance results from the simulations. Important parameters of interest include: time to retrieve images from different possible storage locations and the utilization levels of the transfer devices and other key hardware components. To understand system performance under fully loaded conditions, the proposed system for the Madigan Army Medical Center was modeled in detail, as part of the Medical Diagnostic Imaging Support System (MDIS) proposal.

Due to the expense and time required to configure and evaluate large scale PACS rapid modeling and simulation of system configurations is critical. The results of the analysis can be used to drive the design of both hardware and software. System designers can use the models to help them during the actual system integration. This paper will show how the LANNET 11. 5 and NE1''WORK 11. 5 modeling tools can be used hierarchically to model and simulate large PACS. The detailed description of the Communication Network model which is one of three models used for the Medical Diagnostic Imaging Support System (MDIS) design analysis will be presented. The paper will conclude with future issues in the modeling of MDIS and other large heterogeneous networks of computers and workstations. The way that the models might be used throughout the system life cycle to reduce the operation and maintenance costs of the system is explained.

The American College of Radiology and the National Electrical Manufacturers Association published the ACR-NEMA Digital Imaging and Communications Standard in 1985. Implementations are just now becoming available. Working groups of the committee have been very active. An expanded version of the Standard was published in 1988 and in a third version, to be known as Digital Imaging and Communications in Medicine (DICOM), is being prepared for publication in 1992. This paper briefly reviews the history of the Standard, describes the participation of the committee in international radiological imaging standards activities, and outlines the extensions planned for the DICOM Standard.