Many of the commercial applications for teleradiology have involved the transmission of reduced resolution x-ray images over modest bandwidth telecommunications lines for the purpose of making a preliminary diagnosis. In order to study the technical and operational requirements for future teleradiology applications, the authors have focused on the demanding requirements for teleradiology within the hospital and medical center. Applications within the hospital often require x-ray images of primary diagnostic quality transmitted with a minimum of delay. An experimental, high-resolution film scan/print system designed by Health Sciences Division, Eastman Kodak Company, has been developed for installation in a working clinical environment. Images scanned at a spatial resolution of 4K X 5K can be delivered over a fiber optic link to a laser film printer at a rate of two films per minute. Preliminary plans to install this device in a variety of clinical settings have led to rethinking the requirements for automatic film loading, film and patient identification, throughput requirements, and image display formats. As an initial implementation, and application is being developed which allows chest radiographs taken in the admission area to be interpreted at a remote site within the hospital. Images can be viewed on high resolution monitors, or film replicates can be produced on a nearby laser printer. Tight coupling with a radiology information system provides access to relevant diagnostic information including prior radiology reports, and prompt electronic reporting and signature can be accomplished.

The installation of a local area network of microcomputers, workstations, archival storage devices, and digital acquisition modalities is the first step taken by the Radiology Department at the University of Florida toward a comprehensive picture archival and communications system (PACS). The system meets the following requirements, identified during the initial planning and design stage: (1) It allows the connection of heterogeneous equipment from multiple vendors. (2) It is modular and can be extended and modified. (3) It uses current technology with the option to upgrade when improved methodologies are available. (4) It maintains a database of all patient studies/images archived, including those on optical disks not currently mounted in a disk drive. (5) The database can be customized to match specific requirements. (6) ACR-NEMA commands and headers combined with TCP/IP protocols allow access to images by different types of workstations for display and evaluation. In addition to the clinical operation, the network allows communication among microcomputers and research workstations, allowing extended access to images on the archive. The authors discuss their early experience with the system, including design requirements and performance measures.

Teleradiology can be defined as the remote transmission of radiographic images for clinical use or expert interpretation. This definition indicates that there is a physical distance that impedes patient care between the interpreting expert and the primary physician, which can be overcome through electronic communications. The major benefit of such a system is faster communication of images with expert interpretation to remote sites. Depending on the application, teleradiology can extend the usefulness of the radiologist or make the primary physician's job much less time consuming by saving trips to radiology. In addition, patient interaction can be improved by eliminating the interval between the study and the availability of the images and report. It has not been satisfactorily determined that this more rapid system will lead to improved patient care but most students of the current delivery system recognize its limitations and the promise of electronic communications. The authors confine their remarks to the hospital and immediate clinics, leaving the wider area networks to the other presentations in this seminar, and they draw on the experience of the group at the University of Florida in establishing teleradiology to all the intensive care units two years ago and several other more limited, point-to-point electronic communication links. They have, during the past year, worked very hard at establishing several local area networks with digital archiving capability within their institution. This borders on the notion of picture archiving and communication systems (PACS), but has not reached that full potential. The authors find it is useful to concentrate on the teleradiology component because a number of projects can be undertaken without the need of a complete PACS environment. An extensive bibliography, compiled from select sources, is included.

Preliminary results of a large-scale ROC study evaluating the diagnostic performance of digital hardcopy film and 2K X 2K softcopy display for pediatric chest radiographs are presented. The pediatric disease categories studied were pneumothorax, linear atelectasis, air bronchograms, and interstitial disease. Digital images were obtained directly from a computed radiography system. Results from the readings of 239 chest radiographs by 4 radiologists show no significant difference between viewing images on film and softcopy display for the disease categories pneumothorax and air bronchograms. A slight performance edge for softcopy was seen for the disease categories of interstitial disease and linear atelectasis.

A high-definition, high-performance display station for reading and review of digital radiological images is introduced. The station is based on a Sun SPARC Station 4 and employs X window system for display and manipulation of images. A mouse-operated graphic user interface is implemented utilizing Motif-style tools. The system supports up to four MegaScan gray-scale 2560 X 2048 monitors. A special configuration of frame and video buffer yields a data transfer of 50 M pixels/s. A magnetic disk array supplies a storage capacity of 2 GB with a data transfer rate of 4-6 MB/s. The system has access to the central archive through an ultrahigh-speed fiber-optic network and patient studies are automatically transferred to the local disk. The available image processing functions include change of lookup table, zoom and pan, and cine. Future enhancements will provide for manual contour tracing, length, area, and density measurements, text and graphic overlay, as well as composition of selected images. Additional preprocessing procedures under development will optimize the initial lookup table and adjust the images to a standard orientation.

An interpretation report, generated with an electronic viewbox, is affected by two factors: image quality, which encompasses what can be seen on the display, and computer human interaction (CHI), which accounts for the cognitive load effect of locating, moving, and manipulating images with the workstation controls. While a number of subject experiments have considered image quality, only recently has the affect of CHI on total interpretation quality been measured. This paper presents the results of a pilot study conducted to evaluate the total interpretation quality of the FilmPlane2.2 radiology workstation for patient folders containing single forty-slice CT studies. First, radiologists interpreted cases and dictated reports using FilmPlane2.2. Requisition forms were provided. Film interpretation was provided by the original clinical report and interpretation forms generated from a previous experiment. Second, an evaluator developed a list of findings for each case based on those listed in all the reports for each case and then evaluated each report for its response on each finding. Third, the reports were compared to determine how well they agreed with one another. Interpretation speed and observation data was also gathered.

The requirements for the rapid display of radiographic images exceed the capabilities of widely available display, computer, and communications technologies. Computed radiography captures data with a resolution of about four megapixels. Large-format displays are available that can present over four megapixels. One megapixel displays are practical for use in combination with large-format displays and in areas where the viewing task does not require primary diagnosis. This paper describes an electronic radiology system that approximates the highest quality systems, but through the use of several interesting techniques allows the possibility of its widespread installation throughout hospitals. The techniques used can be grouped under three major system concepts: a local, high-speed image server, one or more physician's workstations each with one or more high-performance auxiliary displays specialized to the radiology viewing task, and dedicated, high-speed communication links between the server and the displays. This approach is enhanced by the use of a progressive transmission scheme to decrease the latency for viewing four megapixel images. The system includes an image server with storage for over 600 4-megapixel images and a high-speed link. A subsampled megapixel image is fetched from disk and transmitted to the display in about one second followed by the full resolution 4-megapixel image in about 2.5 seconds. Other system components include a megapixel display with a 6-megapixel display memory space and frame-rate update of image roam, zoom, and contrast. Plans for clinical use are presented.

The Mayo Clinic and IBM (at Rochester, Minnesota) have jointly developed a picture archiving system for use with Mayo's MRI and Neuro CT imaging modalities. The communications backbone of the PACS is a portion of the Mayo institutional network: a series of 4-Mbps token rings interconnected by bridges and fiber optic extensions. The performance characteristics of this system are important to understand because they affect the response time a PACS user can expect, and the response time for non-PACS users competing for resources on the institutional network. The performance characteristics of each component and the average load levels of the network were measured for various load distributions. These data were used to quantify the response characteristics of the existing system and to tune a model developed by North Dakota State University Department of Computer Science for predicting response times of more complex topologies.

The authors have implemented a communication network connecting multiple buildings for their picture archiving and communication system (PACS) in the Radiology Department at UCLA. The network consists of three types of local area networks (LANs) and a 1.0-km fiber-optic link connecting the outpatient and inpatient facilities. Images from radiologic imaging devices (4 CT scanners, 5 MR scanners, 4 CR units and 5 film digitizers) are transmitted to the acquisition computers via the Ethernet LAN. The fiber distributed data interface (FDDI) LAN then provides data communication among the cluster controllers, the acquisition computers, and the database servers. A 1-gigabit UltraNet LAN is used to route images from the cluster controllers to remote display workstations. All inter-building connections are through fiber-optic cables. Among these multiple networks, Ethernet offers multi-access to the multimodal PACS in image acquisition, FDDI controls a fast data flow so that all acquired images have a shorter residence time on local disks, and UltraNet provides high-speed transfer of images from the cluster controllers to the display workstations. The three-tiered functionality of Ethernet, FDDI, and UltraNet eliminates network traffic bottlenecks and hence provides high performance in image communication. The delay time of a 2K X 2K X 8-bit CR image (4 MBytes) from acquisition to display is less than 5 minutes. In addition, the standard Ethernet serves as a backup to guarantee network connectivity of the entire PACS.

There is a need for a consolidated picture archival and communications system (PACS) in hospitals. At the Bowman Gray School of Medicine of Wake Forest University (BGSM), the authors are enhancing the ACR/NEMA Version 2 protocol using UNIX sockets and TCP/IP to greatly improve connectivity. Initially, nuclear medicine studies using gamma cameras are to be sent to PACS. The ACR/NEMA Version 2 protocol provides the functionality of the upper three layers of the open system interconnection (OSI) model in this implementation. The images, imaging equipment information, and patient information are then sent in ACR/NEMA format to a software socket. From there it is handed to the TCP/IP protocol, which provides the transport and network service. TCP/IP, in turn, uses the services of IEEE 802.3 (Ethernet) to complete the connectivity. The advantage of this implementation is threefold: (1) Only one I/O port is consumed by numerous nuclear medicine cameras, instead of a physical port for each camera. (2) Standard protocols are used which maximize interoperability with ACR/NEMA compliant PACSs. (3) The use of sockets allows a migration path to the transport and networking services of OSIs TP4 and connectionless network service as well as the high-performance protocol being considered by the American National Standards Institute (ANSI) and the International Standards Organization (ISO) -- the Xpress Transfer Protocol (XTP). The use of sockets also gives access to ANSI's Fiber Distributed Data Interface (FDDI) as well as other high-speed network standards.

During the last decade, the concept of picture archiving and communication systems (PACS) has evolved and matured to integrate digital image information in a hospital. PACS integrates various imaging devices, database archive systems, and image viewing workstations. One of the most difficult problems for integration is the standardization of communication protocols required to connect devices from different vendors. When the ACR-NEMA standard was announced for communications in digital radiology in 1985, it solved some of the problems for PACS with a hardware interface, a set of software commands, and a consistent set of data formats. However, it has been found to be inadequate for networked PACS environments, not only because of its point-to-point nature and lack of a network layer, but also because of its inflexibility to allow other services and protocols to be added in the future. Based on previous experience, the case is made for a new standard for PACS networks and a suggested approach for new protocol is presented. This paper addresses the development of a new standard for PACS protocol layers and attempts to define the protocols from a layered-systems approach. The basis for the new protocol definition is primarily the International Organization for Standardization/Open Systems Interconnection (ISO/OSI) protocols and the data format portion of ACR-NEMA standard. The scope and rationale of the proposed protocol, background, and relationship to PACS technology are discussed. The proposed PACS protocol covers the protocol service definition and specification for the application, presentation, session, transport, and network layers. The proposed PACS protocol is intended to facilitate the development of PACSs by several vendors capable of interfacing with each other. The new PACS protocols will also support a global PACS environment.

This paper examines the effect that transmission errors have on the quality of radiologic images. Since the data compression of digitized images is explored as a means for reduction in storage requirements and transfer time, the authors examine the image quality of both compressed and uncompressed images when bit errors occur. This is necessary because a wide variety of communication media and protocols can be used for image transfer, each with different reliability levels. The trade-off between compression, image quality and reliability is explored as a first step for the examination of the issues related to the transfer of images between geographically remote sites. Thus, the main objective of this paper is to identify issues and parameters related to the interconnection of geographically dispersed picture archiving and communication systems.

PACnet is a 200-500 Mbps dual-ring fiber optic network designed to implement a picture archiving and communication system (PACS) in a hospital environment. The network consists of three channels: an image transfer channel, a command and control channel, and a real-time data channel. An initial network interface unit (NIU) design for PACnet consisted of a functional description of the protocols and NIU major components. In order to develop a demonstration prototype, additional definition of protocol algorithms of each channel is necessary. Using the International Standards Organization/Open Systems Interconnection (ISO/OSI) reference model as a guide, the definition of the data link layer is extended. This definition covers interface service specifications for the two constituent sublayers: logical link control (LLC) and medium access control (MAC). Furthermore, it describes procedures for data transfer, mechanisms of error detection and fault recovery. A performance evaluation study was then made to determine how the network performs under various application scenarios. The performance evaluation study was performed using stochastic activity networks, which can formally describe the network behavior. The results of the study demonstrate the feasibility of PACnet as an integrated image, data, and voice network for PACS.

Three difficult problems in making picture archiving and communication systems (PACS) a clinical reality in radiology are image archiving, very high-resolution display stations, and high-speed networking. This paper considers high-speed image transmission through a high- capacity network. Several commercially available high-speed networks were tested over the past year. Only one of these networks (UltraNet) has adequate throughput and capacity potential necessary for the PACS used in the test. The focus of this experiment is to determine the throughput and capacity characteristics of this star topology networking scheme as relates to the operation of a PACS in the clinical environment. A large-scale test was performed to gauge network memory-to-memory performance for three networking configurations modeling those in a PACS: duplex, parallel and relay. Ten computers used in the PACS (Sun 3 and 4 class computers) were connected with UltraNet for the test. For point-to-point throughput (half-duplex model) the network delivers up to 3.1 megabytes/second (MBps) for Sun 3/computers and 4.7 MBps for the Sun SparcServer 490. As regards capacity considerations (parallel model), five parallel image transfer processes generated a maximum of 13.9 MBps through the network. Only a slight degradation in individual process throughput was observed (1.4%). With regard to shared access to high-contention resources on the PACS network (e.g., archive servers), this network demonstrated equal sharing of server networking capacity between various client computers (relay model). For disk-to-disk performance measurement under loaded clinical conditions between two SparcServer 490s, the overall average transfer rate was found to be 1.125 +/- 0.257 MBps, while the average network transfer rate of 3.632 +/- 1.542 MBps was determined. This compares to an average overall transfer rate of 0.389 +/- 0.061 MBps and average network transfer rate under similar conditions of 0.568 +/- 0.060 MBps using Ethernet. For disk-to-memory transfer from a parallel transfer disk (PTD) on a Sun 4/470 to the 2K display frame buffer on a Sun 4/370, the PTD portion of the transfer required 1.2 seconds (6.7 MBps) and the network portion required 1.04 seconds (7.7 MBps) for an overall transfer rate of 3.6 MBps.

The development of PACS image databases has long been thought of as a major technological challenge, due to the amount of data to be managed. On the contrary, the authors think that despite major improvements in storage technology, new data management techniques must be proposed to make image databases medically useful in PACS environments. More precisely, image databases must contain not only images per se, but also the description of all objects used in medical imaging, in order to permit the remote processing, analysis, and interpretation of images. In several other papers, the authors explain why they adopted an object-oriented approach to model information in medical imaging. In this paper, the focus is on the inventory of objects manipulated in medical imaging from a qualitative viewpoint. For this purpose, a large number of representative imaging procedures were selected. The authors characterized how they are asked for by clinicians, realized in imaging departments, and consumed by requiring physicians and surgeons, in three French university hospitals. On the basis of this inventory, a set of image data -- i.e. of objects used in medical imaging -- was defined to show that this set must evolve with advances in medical imaging, and to point out that relational DBMS concepts cannot represent all image data.

This paper is concerned with the main underlying concepts of a comprehensive data model for the medical image database (MIDB), which was developed by the NRV-PACS group. This model is based on semantic and object-oriented model theory and describes not only an image and its environment, as it is the case for standard models, but all meaningful informations of medical image environment, such as acquired or processed data set structures. This model can also take into account changes in image production processes (new kinds of acquisition or new image processing techniques) because it does not describe the entities themselves but how they are created.

Retrieving medical images from a PACS database presents formidable requirements in terms of communications bandwidth and end-user response time. In the Mayo/IBM PAC System, these requirements are even more challenging due to the size and topology of the Mayo campus and the number of images being handled. Accordingly, early on in the project it was decided to distribute the server function among the local rings that make up the installation. Entities called MIDS Servers -- high-end IBM PS/2 computers with a considerable amount of fast DASD -- were designed to attach to a local ring and provide medical images to the medical review stations attached to that ring. In keeping with the cooperative processing philosophy of the overall system, MIDS Servers are semi-autonomous units that are responsible for their own integrity and communicate with the other components of the system as needed to perform their function. Through a strategy that tries to ensure that medical images reside on the MIDS Servers most likely to require them, the authors attempt to minimize the need to retrieve images from the optical archive and reduce inter-ring image traffic. Distributing the server function this way should provide a total server bandwidth greater than the 4Mb/sec capacity of the Token-Ring, as well as faster end-user response time.

The Radiology Department at the University of Pennsylvania is in the process of clinically testing its PACS implementation. The PACS implementation has been built around a Vortech Image Archival and Retrieval System (IARS) with a 140-platter optical jukebox. The Vortech IARS provides archival services only. A set of software modules have been developed in-house that allow the system to function as a digital film library. The current implementation allows connectivity to a RIS (DECrad), supports the routing of images to two intensive care units, and allows image acquisition from a Du Pont FD2000 laser scanner and two GE SIGNA MR units. All process-to-process communication follows the ACR/NEMA 2.0 protocol. The proposed folder extensions to ACR/NEMA 2.0 are being utilized for sending information to the display nodes. The system has been running clinically for about three months. Details of the design, implementation, and functionality of the PACS are presented.

In order to exchange digital imaging data between two independent PACS modules, the ISAC (Image Save And Carry: JPACS/MEDIS-DC) formatted 130-mm magneto-optical disk (MOD) was used as an interchangeable off-line data transferring and recording medium. This system can provide the means for connecting distributed image databases and is useful for clinical studies.

A multi-media magneto-optical disk filing system which stores the voice of image interpretation and order of image diagnosis issued at physicians departments has been developed. The advantages of these 5.25 in., 600 MB disks are rewritability and compactness; they are applicable to circulation media as well as filing media inside hospitals. In this paper, the status of magneto-optical disk (MOD) application to medical images involving standardization activity is briefly introduced. Results of system analysis of the hospital at Osaka University School of Medicine are presented, and the design of HIS/RIS/PACS and role of MOD through HIS/RIS/PACS are described. The result of the experiments -- oral reporting accommodation, together with medical images, interchangeability check, and hand carry acceptance check -- are also reported. Technology assessment steps and future development are further discussed in terms of speed, security, file integration, and operability.

The Radiology Department at the University of Pennsylvania is installing a clinical picture archiving and communication system (PACS). The first phase of the PACS will support two MRI scanners, one film digitizer, and will provide display stations for four intensive care units and the MRI section. The software design for the image display stations has taken the form of two software modules: a display process (DP) and a worstation folder manager (WFM). This layered approach will ease the addition of new display stations into the PACS. The two processes communicate directly and through a folder database (FDB) that resides on the display node. The DP will allow the user to view and perform spatial and grayscale manipulation on multimodality images displayed on two high-resolution (2560 H X 2048 V) grayscale monitors. The WFM is responsible for processing folder requests from the DP, receiving folders from the main archive folder manager (MAIN_FM), fetching images from the image archive and retrieval system (IARS), and maintaining the FDB. All communication between the WFM and both the MAIN_FM and IARS is performed using ACR-NEMA style messages. The WFM has the ability to receive unsolicited folders in addition to folders on request. The WFM can also generate return folders for the MAIN_FM that contain physician reports of viewed images.

The Mayo Clinic and IBM Rochester have jointly developed a picture archive and control system (PACS) for use with Mayo's MRI and Neuro-CT imaging modalities. One of the challenges of developing a useful PACS involves integrating the diagnostic reports with the electronic images so they can be displayed simultaneously. By the time a diagnostic report is generated for a particular case, its images have already been captured and archived by the PACS. To integrate the report with the images, the authors have developed an IBM Personal System/2 computer (PS/2) based diagnostic report acquisition unit (RAU). A typed copy of the report is transmitted via facsimile to the RAU where it is stacked electronically with other reports that have been sent previously but not yet processed. By processing these reports at the RAU, the information they contain is integrated with the image database and a copy of the report is archived electronically on an IBM Application System/400 computer (AS/400). When a user requests a set of images for viewing, the report is automatically integrated with the image data. By using a hot key, the user can toggle on/off the report on the display screen. This report describes process, hardware, and software employed to integrate the diagnostic report information into the PACS, including how the report images are captured, transmitted, and entered into the AS/400 database. Also described is how the archived reports and their associated medical images are located and merged for retrieval and display. The methods used to detect and process error conditions are also discussed.

From the user's point of view -- and that should be the starting point for the development of any information system -- a full-scale PACS has to do at least two things: (a) offer access to all available patient data, both the pictorial and the alphanumeric data, from the workstation, and (b) the response time related to the retrieval and display of these data must at least be acceptable. In order to be able to meet both requirements, the PACS must be linked with the HIS/RIS. For the first requirement, the HIS/RIS will in general be the main source for supplying alphanumeric patient data. The second demand can only be fulfilled if the PACS is able to manage its image database properly. A strategy which provides a way to optimize PACS image management using information generated elsewhere (e.g., by the HIS/RIS) is called prefetching. In this paper, a model of prefetching is presented, and a number of possible parameters for the prefetching algorithm are discussed. Finally, a method for evaluating the effectiveness of the algorithm is proposed and illustrated with the first results of a pilot study carried out in the Leiden University Hospital.

The Radiology Department at the University of Pennsylvania is in the process of clinically testing the first phase of its PACS implementation. The first phase has been designed to support two Intensive Care Units, two GE SIGNA MR units and a Du Pont FD 2000 Film Digitizer. The entire system has been running clinically for about three months. Any PACS effort involves integrating software and hardware from different manufactures. Some of the equipment on a PACS network are often 'closed' and provide limited access. In addition, the entire PACS system is expected to provide uninterrupted service, have adequate throughput and ensure data integrity. By and large, these goals have been met. Clinical testing has provided valuable insights about system integration and highlighted areas of improvement in the system design. A detailed discussion of these results is presented.

The authors have developed a PACS module for genito-urinary radiology. This module is based on image acquisition subsystem, database and storage server/cluster controllers, communication networks, display workstation and local database, and dedicated digitizer and printer. The design guideline for this system is generality and flexibility. As such this module serves as a prototype for future PACS module designs.

The introduction of the "Picture Archiving and Communications System (known as PACS)," provides many benefits, including the application of C.A.D., (Computer Aided Diagnosis). Clinically, this allows for the measurement and design of an operation to be easily completed with the CRT monitors of PACS rather than with film, as has been customary in the past. Under the leadership of the Department of Neurosurgery, Akita University School of Medicine, and Southern Tohoku Research Institute for Neuroscience, Koriyama, new computer aided functions with EFPACS (Fuji Electric's PACS) have been developed for use in clinical neurosurgery. This image processing is composed of three parts as follows: (1) Automatic mapping of small lesions depicted on Magnetic Resonance (or MR) images on the brain atlas. (2) Superimposition of two angiographic films onto a single synthesized image. (3) Automatic mapping of the lesion's position (as shown on the. CT images) on the processing image referred to in the foregoing clause 2. The processing in the clause (1) provides a reference for anatomical estimation. The processing in the clause (2) is used for general analysis of the condition of a disease. The processing in the clause (3) is used to design the operation. This image processing is currently being used with good results.

The Mayo Clinic and IBM/Rochester have jointly developed a picture archiving system (PACS) for use with Mayo's MRI and Neuro-CT imaging modalities. The system was developed to replace the imaging system's vendor-supplied magnetic tape archiving capability. The system consists of seven MR imagers and nine CT scanners, each interfaced to the PACS via IBM Personal System/2(tm) (PS/2) computers, which act as gateways from the imaging modality to the PACS network. The PAC system operates on the token-ring component of Mayo's city-wide local area network. Also on the PACS network are four optical storage subsystems used for image archival, three optical subsystems used for image retrieval, an IBM Application System/400(tm) (AS/400) computer used for database management and multiple PS/2-based image display systems and their image servers.

The relationship between subjective judgments of image quality for the performance of specific detection tasks and radiologists' confidence level in arriving at correct diagnoses was investigated in two studies in which 12 readers, using a total of three different display environments, interpreted a series of 300 PA chest images. The modalities used were conventional films, laser-printed films, and high-resolution CRT display of digitized images. For the detection of interstitial disease, nodules, and pneumothoraces, there was no statistically significant correlation (Spearman rho) between subjective ratings of quality and radiologists' confidence in detecting these abnormalities. However, in each study, for all modalities and all readers but one, a small but statistically significant correlation was found between the radiologists' ability to correctly and confidently rule out interstitial disease and their subjective ratings of image quality.

An increasing practical problem in the evaluation of the accuracy of new imaging systems, as well as the effect of modifications in the display of current imaging systems, is the effort entailed in performing the necessary readings. Although some constraints are dictated by the specific evaluation being conducted, some aspects of the experiment can be determined by the investigator. These include, but are not limited to, the method used to select the cases (selected, stratified, or random) and whether continuous variables that are being evaluated (i.e., pixel size, brightness, contrast) are grouped into discrete categories. The selection of the experimental design has an impact on the sample size required to answer the study question and thus impacts on the cost and effort required to do the study.

In a series of large ROC studies, the authors analyzed the time radiologists took to diagnose PA chest images as a function of observer performance indices (Az), display environments, and difficulty of cases. Board-certified radiologists interpreted at least 600 images each for the presence or absence of one or more of the following abnormalities: interstitial disease, nodule, and pneumothorax. Results indicated that there exists a large inter- reader variability in the time required to diagnose PA chest images. There is no correlation between a reader's specific median reading time and his/her performance. Time generally increases as the number of abnormalities on a single image increases and for cases with subtle abnormalities. Results also indicated that, in general, the longer the time for interpretation of a specific case (within reader), the further the observer's confidence ratings were from the truth. These findings were found to hold true regardless of the display mode. These results may have implications with regards to the appropriate methodology that should be used for imaging systems evaluations and for measurements of productivity for radiologists.

For the last four years, the UNC FilmPlane project has focused on constructing a radiology workstation facilitating CT interpretations equivalent to those with film and viewbox. Interpretation of multiple CT studies was originally chosen because handling such large numbers of images was considered to be one of the most difficult tasks that could be performed with a workstation. The authors extend the FilmPlane design to address mammography. The high resolution and contrast demands coupled with the number of images often cross- compared make mammography a difficult challenge for the workstation designer. This paper presents the results of preliminary work with workstation interpretation of mammography. Background material is presented to justify why the authors believe electronic mammographic workstations could improve health care delivery. The results of several observation sessions and a preliminary eyetracker study of multiple-study mammography interpretations are described. Finally, tentative conclusions of what a mammographic workstation might look like and how it would meet clinical demand to be effective are presented.

Physician ratings of the importance of rapidly viewing images were compared with the frequency of using a 512 X 512 X 8 display station instead of film for the first viewing of chest images made in a medical intensive care unit. The overall utilization of the display station was 35%. Although there was a statistically significant correlation between utilization and the importance rating, it accounted for only a small part of the difference in the utilization of the display station and film.

Image delivery performance of a CT/MR PACS module was evaluated through a year-long clinical operation. Performance comparisons between the current film-based management system and the PACS module were conducted. By analyzing the elapsed time spent in each modularized task in both systems, the study results indicate the image delivery performance of the PACS system is acceptable in clinical application.

Current and future operational requirements that support a picture archiving and communication system that manages filmless digital modality images are described.

The Department of Veterans Affairs has undertaken a project to integrate image management functionality within its hospital information system. One of the goals of this integrated image system is to provide high-quality image data from cardiology, pulmonary and gastrointestinal endoscopy, pathology, radiology, hematology, and nuclear medicine to clinicians throughout the hospital. Images are presented to clinicians on high-resolution workstations simultaneously with associated text data. This DHCP Integrated Imaging System is currently being installed and tested at the Washington DC VA Medical Center. Initial use has revealed a number of expected as well as unanticipated benefits for patient care and medical teaching.

The PACS under development at the University Hospital of Geneva is a hospital-wide image management system for radiological as well as non-radiological medical images which is part of one of the widest hospital information systems (HIS) in Switzerland (Diogene system). It is based on a multi-vendor open architecture and a set of widely available industry standards, namely: Unix as the operating system, TCP-IP as network protocol and an SQL-based distributed database (INGRES) that handles both the PACS and the HIS. The PACS is based on a distributed architecture of servers of two types: the archive servers connected to the sources of images and equipped with large optical disk libraries (jukeboxes) and display servers distributed over the hospital. A standard image storage format was developed based on the ACR-NEMA standard. This file format (the PAPYRUS format) allows storage of sets of images as a sequence of ACR-NEMA messages in an 'encapsulated' file structure. In order to provide a more uniform user interface on a variety of different workstations, a common platform for image display and manipulation called OSIRIS is developed based on X-11 windowing system and OSF/Motif extension. Such a platform is designed to be portable to any computer running Unix and equipped with a graphic display system running X-11. Because this software is written in the object-oriented language C++, it is easily expandable and easily adaptable to different needs and requirements.

The size of a typical data transfer in a PACS environment is several orders of magnitude larger than a typical data transfer in a local area computer network. PACS is also moving to a multi-vendor environment. To ensure reasonable response times to users, it is necessary to manage the image traffic on the PACS network explicitly, and for the heterogeneous network environment, tools for monitoring and maintenance are necessary. Modular network architectures, capable of reacting to growth and changes in traffic patterns, are essential. Knowledge about patient-radiologist and image flow which is available in the hospital environment can be exploited for careful planning and scheduling of network traffic. The authors designed and implemented a prototype network management system that makes the domain knowledge and the traffic management strategy explicit. In general, the term network management covers a wide range of tasks. Network design and configuration, for example, are network management tasks situated in the pre-operational phase. Network monitoring and control, including error detection, fault diagnosis, dynamic routing, etc., are network management tasks with a real-time aspect. This PACS is hierarchically organized. Each sub- department has its own local PACS. The paper focuses on the results of the implementation of the PACS architecture and, more precisely, on the network management aspects.

At the department of nuclear medicine of the University Hospital Utrecht a single-modality PACS has been operational since mid-1990. After one year of operation the functionality, the organizational and economical consequences, and the acceptability of the PACS were evaluated. The functional aspects reviewed were: viewing facilities, patient data management, connectivity, reporting facilities, archiving, privacy, and security. It was concluded that the improved quality of diagnostic viewing and the potential integration with diagnosis, reporting, and archiving are highly appreciated. The many problems that have occurred during the transition period, however, greatly influence the appreciation and acceptability of the PACS. Overall, it is felt that in the long term there will be a positive effect on the quality and efficiency of the work.

This paper gives an overview of several European PACS activities that are financially supported by the European Community (EC). The activities share as a characteristic that they all work in the direction of an open systems architecture. Current and future activities in the European PACS field, where BAZIS does or will contribute actively, are described.

A simulator has been developed to evaluate the performance of the Mayo/IBM PACS. The simulation model includes both the hardware and the software of the entire system. Using this model, the simulator is able to accurately predict the performance of alternative network configurations, device characteristics, and usage patterns. The significance of this capability is that performance-related issues can be conveniently explored prior to the design, installation, or modification of the PACS. Information about potential bottlenecks, resource utilization, and expected response times can be obtained while changes are still relatively easy to make. This paper presents the objectives and design of the Mayo/IBM PACS simulator. A simple archive task is used to illustrate the major features of this unique tool.

The authors report on the Signa Tutor, a magnetic resonance imaging (MRI) simulator. Their goal was to cost-effectively teach MRI scanning techniques to technologists and physicians using the simulator, off-line from the General Electric (GE) Signa scanner. They implemented the scanner's user-interface for image acquisition, using HyperCard together with an Oracle database to maintain images and scan-parameters. The software operates on a standard Apple Macintosh-II computer. Two aspects of the project are reported in this paper. First, enhancements to the Signa Tutor, including Signa 4.0 pulse sequences and new images, are presented. Second, a time and motion study that compares image acquisition learning time using Signa Tutor verses traditional technologist training on the GE-Signa console is described. In this study, an expediency test was performed which compares the speed at which an operator performs a single scan sequence on both devices.

Institutions involved in the research and development of picture archiving and communication systems (PACS) have formulated numerous models that can be used to design the optimal PACS. The models can be classified into one of three categories: engineering, operational, or cost-effectiveness. This paper concentrates on one result of the engineering models: the need for a large network bandwidth in a clinical PACS. The authors discuss a software solution that they have implemented which can cost-effectively achieve the required bandwidths for the majority of the transactions. The software operates on commercially available hardware. The scenarios under which the implementation may fail are also presented.

The development of picture archiving and communication systems benefits from modeling and simulation. In order to capture the full complexity of PACS in a simulation model, and to take full advantage of simulation as a design tool, the authors have started the development of a new modeling method. This paper illustrates the modeling method by means of simulation models of the Geneva PACS.

Traditionally, radiologists have collected and saved interesting cases in their film formats to teach medical students, residents, and physicians. These cases are classified according to various coding schemes, although current schemes alone are insufficient to meet today's educational needs. Teaching methods and cases also vary among institutions, along with the manner in which instructors present information to their students. In order to address this problem, the authors developed a standardized radiology teaching knowledge database known as the Digital Radiology Image Learning Library (DRILL). DRILL is a relational image knowledge database providing access to standard mammography cases in digital image format along with a pool of clinical and radiological information on a per-case basis. The development platform chosen is a standard Apple Macintosh-II computer and the Oracle database environment. The data entry and query interfaces are implemented in HyperCard. Images are stored on magnetic disk but could be stored on optical media. Since the personal computer platform was chosen, a wide variety of course building tools are available through which a teacher can construct a course, such as authoring and multi-media systems for building computer based courses, or word processors for writing course outlines tests. The interface also provides image conversion tools which convert images into PC-compatible formats.

Picture archiving and communication systems (PACS) have evolved at a sporadic rate over the past ten years. Significant technological advances have transformed these systems from a theoretical concept into functional reality. The Proceedings of the first SPIE/PACS conference held in 1982 served as a 'starting point' from which progress in PACS may be measured. The panel discussion at this conference provided a valuable forum for recognized leaders from industry as well as the medical community to present their opinions. Participation from this diverse group ensured that the entire spectrum of PACS related topics was discussed. This paper retrospectively reviews the perceptions and predictions about the future of PACS as expressed at the 1982 SPIE/PACS conference. The major predictions are summarized and compared with the current state-of-the-art in PACS.

Radiological diagnoses include tasks of detection and interpretation. Information from previous studies of observer performance was used to investigate the correlation between the detection and characterization of imaging features for five groups of observers who reviewed 300 posteroanterior (PA) chest images using each of the following modes: conventional films, digitized images displayed on laser-printed films, and high-resolution soft display with and without edge enhancement. Feature detection tasks for each image included the presence or absence of septal lines and nodules; interpretation tasks included the diagnosis of interstitial disease and the characterization of nodules as benign or malignant. In this study, edge enhancement resulted in an increase of both true- and false-positive detection of septal lines. Although limited by the small number of pathologically classified nodules (31), there was no significant correlation between a reader's ability to detect nodules and characterize them as benign or malignant, regardless of the mode of image presentation.

Currently, three GE CT9800 and three GE Signa Advanced scanners are connected to the picture archiving and communication system (PACS) at the University of California Los Angeles (UCLA) Medical Center. While the MRI systems are equipped with a standard Ethernet interface, special development efforts were necessary in order to utilize the data port of the CT system for an automatic data transmission into the PACS network. For both modalities designated computer stations accommodate the acquisition of the image data through a special protocol and transmit the data in UCLA PACS standard format to the central controller for archiving and distribution to the display stations. The MRI images are acquired as they are generated through a capture process of the acquisition station that has direct access to the internal database of the MRI system. The transmission of the CT image data, however, is governed by the CT system computer and is limited to a low priority process in between clinical operations. Average MRI studies become available for review within thirty minutes, whereas CT studies are delayed up to two to three hours depending on the load of the CT system.

A joint project between Mayo Clinic and IBM to develop a picture archival and communications system has been under development for three years. This project began as a potential solution to a pressing archival problem in magnetic resonance imaging. The project has grown to encompass a much larger sphere of activity including workstations, image retrieval, and report archival. This report focuses on the clinical aspects involved in the design, development, and implementation of such a system. In particular, emphasis is placed on the clinical impact of the system both inside and outside of the radiology department. The primary concerns have centered on fidelity of archival data, ease of use, and diagnostic efficacy. The project to date has been limited to neuroradiology practice. This group consists of nine staff radiologists and fellows. Administrative policy decisions regarding the accessibility and available of digital data in the clinical environment have been much more difficult and complex than originally conceived. Based on the observations thus far, the authors believe the system will become a useful and valuable adjunct to clinical practice of radiology.

This joint project began in 1988 and was motivated by the need to develop an alternative to the archival process in place at that time (magnetic tape) for magnetic resonance imaging and neurological computed tomography. In addition, this project was felt to be an important step in gaining the necessary clinical experience for the future implementation of various aspects of electronic imaging. The initial phase of the project was conceived and developed to prove the concept, test the fundamental components, and produce performance measurements for future work. The key functions of this phase centered on attachment of imaging equipment (GE Signa) and archival processes using a non-dedicated (institutionally supplied) local area network (LAN). Attachment of imaging equipment to the LAN was performed using commercially available devices (Ethernet, PS/2, Token Ring). Image data were converted to ACR/NEMA format with retention of the vendor specific header information. Performance measurements were encouraging and led to the design of following projects. The second phase has recently been concluded. The major features of this phase have been to greatly expand the network, put the network into clinical use, establish an efficient and useful viewing station, include diagnostic reports in the archive data, provide wide area network (WAN) capability via ISDN, and establish two-way real-time video between remote sites. This phase has heightened both departmental and institutional thought regarding various issues raised by electronic imaging. Much discussion regarding both present as well as future archival processes has occurred. The use of institutional LAN resources has proven to be adequate for the archival function examined thus far. Experiments to date have shown that use of dedicated resources will be necessary for retrieval activities at even a basic level. This report presents an overview of the background present status and future directions of the project.

As PACS mature and their integration in hospitals can be expected in the not too distant future, insight into the costs and benefits of these systems is needed. A major problem in the assessment of the costs and benefits of PACS is the lack of international consensus on, e.g., the functional requirements of PACS, on the costs of some major components of the system, and on the impact of the implementation of PACS in a hospital on logistics and patient care. The authors have developed and distributed the software package, CAPACITY, which supports cost and benefit analyses and facilitates the exchange of data. This paper presents an overview of some opinions on the costs and benefits of PACS. It is based on literature data and on data which have been assembled through CAPACITY. Most experts agree that a hospital-wide PACS is more expensive than film, at least now, but the opinions on the costs and the required functionality of PACS components show a wide variation. The costs of PACS however must be weighed against their benefits. PACS has the potential of improving the hospital logistics and patient care. More quantified data on the effect of PACS on, e.g., the availability of images and on the length of patient stay is needed.

Reliability is an increasing concern when moving PACS from the experimental laboratory to the clinical environment. Any system downtime may seriously affect patient care. The authors report on the several classes of errors encountered during the pre-clinical release of the PACS during the past several months and present the solutions implemented to handle them. The reliability issues discussed include: (1) environmental precautions, (2) database backups, (3) monitor routines of critical resources and processes, (4) hardware redundancy (networks, archives), and (5) development of a PACS quality control program.

Successful implementation of a teleradiology system in a clinical setting places stringent requirements on the system configuration and the performance of individual components. To evaluate possible approaches for high-quality teleradiology, a research system has been developed and is being tested at the Mallinckrodt Institute of Radiology. It consists of a high- performance laser-based film digitizer linked by fiber optics to a remote controlling node that can produce film replicas at a rate of one per minute. In this paper, the design considerations and performance of the research system are presented.

The computerized analysis and interpretation of three-dimensional medical images is of significant interest for diagnosis as well as for studying pathological processes. Knowledge-based image analysis and interpretation of radiological images can provide a tool for identifying and labeling each part of the image. The authors have developed a knowledge-based biomedical image analysis system for interpreting medical images using an anatomical knowledge base of the appropriate organs. In this paper, the structure of the biomedical image analysis system, along with results from the analysis of images of the human chest cavity, are presented. This approach utilizes an image analysis system with the capability of analyzing the data in both bottom-up (or data driven) and top-down (or model driven) modes to improve the recognition process. After an initial identification is achieved, segmented regions are aggregated and features for these aggregates are recomputed and matched to the model. This process continues until a 'best' match is found for the subject model region. Initial results are encouraging; however, much work remains to be done.

The Mayo Clinic and IBM Rochester, Minnesota, have jointly developed a picture archiving, distribution and viewing system for use with Mayo's CT and MRI imaging modalities. Images are retrieved from the modalities and sent over the Mayo city-wide token ring network to optical storage subsystems for archiving, and to server subsystems for viewing on image review stations. Images may also be retrieved from archive and transmitted back to the modalities. The subsystems that interface to the modalities and communicate to the other components of the system are termed Image Acquisition Units (LAUs). The IAUs are IBM Personal System/2 (PS/2) computers with specially developed software. They operate independently in a network of cooperative subsystems and communicate with the modalities, archive subsystems, image review server subsystems, and a central subsystem that maintains information about the content and location of images. This paper provides a detailed description of the function and design of the Image Acquisition Units.

One aspect of the Georgetown image management and communications system (IMACS or PACS) is a built-in capability to support teleradiology. Unlike many dedicated teleradiology systems, the support of this capability as a part of PACS means that any acquired images are remotely accessible, not just those specifically input for transmission. Over the past one and one-half years, two radiologists (SCH, BSG) in the abdominal imaging division of the department of radiology have been accumulating experience with teleradiology for on-call support of emergency abdominal imaging, chiefly in ultrasound. As of the time of this writing, use of the system during on-call (one of these attending radiologists primarily responsible) or back-up call (the attending responsible for the Fellow on primary call) has resulted in a marked reduction in the number of times one of them has to drive to the hospital at night or over the weekend. Approximately 80% of the time, use of the teleradiology system obviates having to go in to review a case. The remainder of the time, the radiologist has to perform a procedure (e.g., abscess drainage) or a scan (e.g., complex Doppler study) himself. This paper reviews the system used for teleradiology, how it is electronically and operationally integrated with the PACS, the clinical benefits and disadvantages of this use, and radiologist and referring physician acceptance.

The ACR-NEMA standard allows different medical devices to communicate compatibly in a picture archiving and communication system (PACS) environment. The performance of a PACS, incorporating the ACR-NEMA standard, has been previously analyzed. It was shown that the ACR-NEMA protocol results in excessive response time values for image viewing requests. In this paper, the authors propose and study the performance effects of a number of enhancements to the ACR-NEMA standard. The objective of such enhancements is to make the ACR-NEMA protocol better suited for transmission of large images. It is shown that such enhancements, indeed, result in a much improved system performance.

During a nine-month period, ultrasound reading was switched between a PACS workstation (CommViewR, AT&T/Philips) and a film alternator (PanoramascopeR RADX). The two radiologists who participated in this study were well acquainted with the operation of both systems. A total of 430 cases were read, and whether on film (n equals 292) or from the workstation (n equals 138), a routine mix of cases was interpreted. A timing study was performed which involved recording the time it took to read and dictate the cases. A previous study had concentrated on other facets of the total time, so these were not repeated. Over all cases read, the workstation readings took longer by an average of approximately 33 seconds per case. A review of the reports generated showed no differences in the average length, so that the time difference was thought to be due to other factors. The majority of the difference is the result of the longer time it takes to display all images at full resolution on the workstation. On the alternator, the average number of films per case can fit on one alternator panel, so to see all images, there is no need to page to the next panel. The workstation, on the other hand, can only display about half of the total images per study at once, so that paging must be done for almost all cases. This paper presents the detailed findings of this study and discusses the implications for the daily use of a workstation for reading.