Medicine is a visual discipline. In the Practice of medicine, physicians require many forms of visual information to successfully conduct their tasks of diagnosing the presence or absence of disease; evaluating the progression or remission of disease; developing strategies for individual patient treatment planning; and in educating their peers and students. Thus, today's hospitals must provide an effective management strategy for a variety of medical images. This management strategy includes the functions of the acquisition of patient images, the archiving of patient images, and the storage of patient images. In a hospital, each medical specialty generates a class of visual images from which information is extracted for use by the patient's physician. This paper will address four issues in the use of medical images in today's hospitals. First, an estimate of the sources and utilization of clinical images in a hospital will be presented. Second, estimates will be provided regarding the magnitude of each of these images sources. Third, current management strategies for dealing with these images will be reviewed. Fourth, several potential solutions will be described for improving the management and archiving of these image sources in a hospital environment.

The cost of recording and archiving digital diagnostic imaging data is presented for a Radiology Department serving a 614 bed University Hospital with a large outpatient population. Digital diagnostic imaging modalities include computed tomography, nuclear medicine, ultrasound, and digital radiography. The archiving media include multiformat video film recordings, magnetic tapes, and disc storage. The estimated cost per patient for the archiving of digital diagnostic imaging data is presented.

A survey is presented of recent changes in digital image acquisition for diagnostic radiology. New digital image archiving, communication and management strategies are proposed and discussed. A possible scenario for the development of digital Picture Archiving and Communication Systems ("digital PACS") in diagnostic radiology is presented, starting with a research/teaching environment and ending with a large scale clinical environment. This scenario is presented as a sequence of five strategies with increasing levels of digital archiving and communication and requires five key concepts. We briefly discuss each key concept and present possible implementations, given today's technology. A preliminary evaluation of proposed image archiving strategies is made. The paper concludes that the development of a PACS that provides partial on-line digital storage is feasible today.

The revolution in data handling associated with digital systems has excited the medical world into considering replacing existing methods. The integration of digital with analogue video systems into a complete electronic communications network holds tremendous promise. One area where these changes will have significant impact is in medical imaging. The potential of more and better information being,obtained from current studies, more rapid access to studies, intercomparison of images from different modalities, savings in space and equipment, reduced procedure time, improved communication in a distributed department, remote image consultation, and timely data base for management are only some of the advantages that can accrue from proper use of electronic systems. In the hospital, the traditional radiology department is becoming a medical imaging de-partment. To effect a change from the ubiquitous silver halide film base system into a more electronic department will necessitate much experimentation. At the University of North Carolina we are attempting to involve the entire department in these endeavors as well as others such as biomedical engineering, computer science and traditional clinical areas. To facilitate this, a management system has been developed and is being used. The system, our architectural draft of a future department and the initial problems and questions which have arisen from implementing a prototype system will be the subject of this report.

Current technological developments suggest that in the very near future all medical imaging modalities will have been converted to digitally based systems requiring no film or other intermediate media for either data recording or information storage. However, the full potential of medical imaging to diagnostic medicine will be realized only with the integration of the various imaging modalities into a single unified imaging system. Clearly the computer, and in particular, distributed computer systems, will play a central and unifying role in all future medical imaging systems. This paper outlines the architecture of a particular distributed system for the acquisition, processing, and filing of medical images produced by several different imaging modalities. Although the system is based in part on digital optical disc technology, other storage technologies could also be adopted. Methods for the standardization and unification of the various imaging modalities will be described. This unification is centered on key computer, microprocessor, and communication elements. The picture or image base is managed by a superior data base which permits user-oriented hierarchical access to data and related pictures. Separated hardware levels are provided for management, control, and signal processing. The processing level is equipped with image processors based on unified modular architecture. This paper also describes a possible historical scenario indicating how the integration of the various imaging modalities might be accomplished. The scenario begins with ultrasound, continues with CT, NMR, and nuclear medicine imaging, and concludes with digital radiography.

A PACS is a digital system for acquiring, storing, moving and displaying picture or image information. It is an alternative to film jackets that has been made possible by recent breakthroughs in computer technology: telecommunications, local area nets and optical disks. The fundamental concept of the digital representation of image information is introduced. It is shown that freeing images from a material representation on film or paper leads to a dramatic increase in flexibility in our use of the images. The ultimate goal of a medical PACS system is a radiology department without film jackets. The inherent nature of digital images and the power of the computer allow instant free "copies" of images to be made and thrown away. These copies can be transmitted to distant sites in seconds, without the "original" ever leaving the archives of the radiology department. The result is a radiology department with much freer access to patient images and greater protection against lost or misplaced image information. Finally, images in digital form can be treated as data for the computer in image processing, which includes enhancement, reconstruction and even computer-aided analysis.

Digital optical recording (DOR) technology lends itself particularly well to handling information at high data rates and to storage of large amounts of information per disc, thereby making this technology especially useful for present and future medical appli cations. Recent advances in high density modulations, parallel recording, and aisc for matting have increased the utility of this technology.

The optical disc has become a new technique for mass digital data storage of X-ray images from examinations and films in todays hospitals. Up to 36,000 X-ray images can be stored on one side of a 12-inch disc by melting holes 0.015 mils in size in an ablative material such as tellerium with a laser beam. This unique characteristic makes the disc suitable for storage and retrieval of X-rays in a record and playback system in either a single disc or multiple disc "jukebox" configuration. Doctors, nurses, technicians and other hospital personnel can call up a particular X-ray in less than 0.6 of a second in an on-line single disc system and up to less than 6 seconds in an on-line "jukebox" system. The jukebox is configured to hold up to 100 discs, thus storing 3,600,000 X-rays in hospitals with a bed size of greater than 500. The estimated exposed films on file in those hospitals is 327,400,000 and the estimated annual X-ray exams are 44,300. Thus, a single disc system could be used for an all electronic X-ray scanning system for annual X-ray exams. The jukebox configuration, which has expansion capability for servicing multiple simultaneous user request, can be applied to large archival mass storage. These systems could store the existing exposed films in hospitals with bed size greater than 500 at record and playback data rates of 50 Mb/s with access times of less than 15 seconds.

Shortly, picture archiving will be done with digital optical recording systems. In our approach, such systems employ Philips Air Sandwich' recording discs for aensities up to about 10¹¹ bits for a cost as low as $10.00, while for higher aensities cartriage designs have been fabricated. Archival life with a data integrity of about 1 in 10¹¹ is expected to be at least 10 years. A by brio pregroovea technology allows the upaate of files through proper data/disc organization.

Emphasizing the goals of low cost, high image quality, and reliability, a real-time magnetic digital video disc recording system has been developed for applications in on-line digital imaging systems and off-line fast access image storage and retrieval buffers. The disc recorder uses new high density recording technology and Winchester computer drive technology in a unique peripheral configuration which is fully synchronized to video system timing, provides for flexible formatting, achieves fast random access to a large video image file, and eliminates the need for complex data controllers. The paper includes a discussion of electronic radiography image memory requirements, and the role of a magnetic digital video disc recorder in various digital radiography system applications.

Historically, instrumentation tape recording systems have maintained state-of-the-art magnetic performance parameters in the areas of storage density, storage capacity, data throughput rates and cost-per-bit. In order to compare instrumentation equipment with other competing technologies, this paper presents important performance parameters for several data storage systems which directly compares the current status of instrumentation, disc and solid state storage devices, A discussion of the potential applications of instrumentation systems to medical imagery is also included. In addition, the technology base supporting instrumentation recorders is examined and anticipated improvements in performance are projected over the next several years. Advances in rotating head recorders are expected to show particularly exciting improvements in storage density and capacity.

Over the past 15 to 20 years moving head magnetic disks have been the primary device used on computer systems for mass storage. During this time magnetic disks have increased in performance and storage capacity while experiencing a significant drop in cost.

A number of recently developed optical data storage systems are reviewed in this paper. Each system utilizes some laser beam recording method as a means of permanently storing digital data on optical storage medium. The technical approaches and storage media employed by these systems vary considerably, ranging from ablative pit-creation on tellurium disc, or metalized glass slide to optical spot recording on silver halide film. Several performance characteristics and parameters of these optical data storage systems relevant to their applications in picture archiving are discussed. Also addressed are the developments of other types of optical data storage media such as bubble forming medium, dye film, and magneto-optic disc.

This paper discusses a fiber optic communication system linking ultrasound devices, Computerized tomography scanners, Nuclear Medicine computer system, and a digital fluoro-graphic system to a central radiology research computer. These centrally archived images are available for near instantaneous recall at various display consoles. When a suitable laser optical disk is available for mass storage, more extensive image archiving will be added to the network including digitized images of standard radiographs for comparison purposes and for remote display in such areas as the intensive care units, the operating room, and selected outpatient departments. This fiber optic system allows for a transfer of high resolution images in less than a second over distances exceeding 2,000 feet. The advantages of using fiber optic cables instead of typical parallel or serial communication techniques will be described. The switching methodology and communication protocols will also be discussed.

Fiber optics promises to be an attractive transmission medium for communication between the computer and its I/O devices and for local networks. Its attractiveness results from its potential for providing high data transfer rate and long transmission distance. Within the central computer complex, serial transmission on fiber optics can also eliminate many of the problems associated with today's bulky parallel cables and connectors. However to obtain the maximum benefits of fiber optics, it will be necesary to make changes to the I/O interface architecture as well as the transmission medium. Changes may also be required in the way applications utilize I/O in order to take advantage of the high data rate potential. This paper discusses some of the features of fiber optics as applied to the computer system and also indicates possible architecture and application approaches for deriving the greatest benefit from fiber optics.

The paper will discuss individual parameters of fiber optic broadband communications system design and will review systems being installed or projected around the world, regarding the feasibilities of fiber optics. Emphasis will be given to the application of distributed control switching. Finally the experimental communications system of the Heinrich-Hertz-Institut, Berlin, a fiber optic system with distributed control at high bandwidth will be presented and discussed.

This paper explains the theory and operation of local area networks which employ broadband coaxial cable communication technology. Such networks combine data, voice and T.V. communications on a single one-half inch diameter coaxial cable. The 300 Megahertz (MHZ) bandwidth of the network also provides multiple high speed (multi-million bit per second) data channels required for medical image processing applications. The theory is then applied to the development of the total facility (hospital complex) Information Network providing services which include security, facilities management, data processing, and emergency paging as well as medical image data transmissions.

A 160 million bit per second (Mbps) data bus system is presented. This system provides a high speed data path between distributed subsystems through the use of microprocessor controlled access ports called nodes. Data bus system operation, node functions and high level node design are described. The system is modularly designed to allow data transmission rates from 20 to 160 Mbps in 20 Mbps steps. Each step is achieved by adding a 20 Mbps cable to the bus system, up to the maximum of eight cables. This modularity yields flexibility in tailoring the bus system transfer rate to the character of a particular application or to provide redundant cables for backup capability.

Clinical trials have been conducted at several CT scanner sites that form an imaae transmission network in operation since June 1981. Network connection provides a special image reformatting service to aid CT diagnosis of difficult lumbar/cervical spine, petrous bone and orbit cases. Important side effects of the network service include improved scanner maintenance, higher patient throughput and scanner system utilization. The overall network concept is introduced, network products are illustrated, early results are discussed, and new directions are indicated.

A hierarchical systems architecture is being described, which introduces separated levels for picture data processing, control, and management, respectively. The system is distributed with respect to the picture signal level and the control level: Workstations for picture generation, evaluation, or processing contain control functions (conventional mini/micro-computers) and picture processing functions (e.g. specialized hardware modules). They communicate with the central optical disc picture base via standard channels on the control level, and via high-speed channels on the picture data level. The management level is, in contrast, centralized. Any existing data base system may be used to manage, by means of minor extensions, the physically separated picture base, and thus to link pictures to non-pictorial information.

Digital radiography, nuclear medicine, radiation therapy planning and diagnostic ultrasound share common requirements for image acquisition, storage, recall and processing. An architecture which has unified the design, manufacture, installation and servicing of systems for these medical imaging applications is described in this paper. Bounds on system performance imposed by physical constraints which have proven useful to aid in the system designs are also described.

Even in its most primitive mode of operation - acquisition, filing, and retrieval - picture information systems will need high-speed picture processors. The tens of megabits of each picture have to be transferred, buffered, and coded in a user-friendly time of only a few seconds. The required throughput rate will be about 10 Mpel/s, which is far beyond the capability of conventional computer hardware. In order to avoid a proliferation of special purpose hardware designs, we have developed a picture processor concept, which is based on a modular architecture. A register of full (or multiple) picture size is connected via a high-speed bus structure to one or several processing elements. Each processing element manages its own data stream in order to synchronize data transfer and processing. Processing elements may be hardwired (fast but specialized), or programmable on different levels. Several modes of multiprocessing are possible (SIMD, pipeline, etc.). Any local sub-system of a distributed picture information system may be assembled from a number of standardized modules. Local sub-systems may easily be optimized with respect to the most frequently used operations, but will still be flexible to allow any extension towards more sophisticated processing.

A preliminary design study has been carried out for a distributed picture archiving and communication system for the Mallinckrodt Institute of Radiology. The study develops design equations for three layers of a picture network and examines the estimated flow of digital images between a multiplicity of picture sources, picture archives and picture viewing stations. Application of these data to the design equations leads to some preliminary conclusions. One network architecture consistent with these conclusions is discussed.

Image systems design is currently undergoing a metamorphosis from the conventional computing systems of the past into a new generation of special purpose designs. This change is motivated by several factors, notably among which is the increased opportunity for high performance with low cost offered by advances in semiconductor technology. Another key issue is a maturing in understanding of problems and the applicability of digital processing techniques. These factors allow the design of cost-effective systems that are functionally dedicated to specific applications and used in a utilitarian fashion. Following an overview of the above stated issues, the paper presents a top-down approach to the design of networked image analysis systems. The requirements for such a system are presented, with orientation toward the hospital environment. The three main areas are image data base management, viewing of image data and image data processing. This is followed by a survey of the current state of the art, covering image display systems, data base techniques, communications networks and software systems control. The paper concludes with a description of the functional subystems and architectural framework for networked image analysis in a production environment.

Photoelectronic imaging offers the advantages of low operating cost, ease of use, and compatibility of media throughout the system. Until recently, most diagnostic images were full-size x-ray film which serves both as image receptor, storage medium, and display. The disadvantages associated with film are a relatively long preparation time, virtually no possibility to interact with the recorded information, and incompatibility with high-speed storage and retrieval systems. Digital imaging modalities such as CT and DVI are posing new image management requirements. The contrast resolution of these systems is larger than can be represented on CRT monitors and multiformat films. Therefore, it is necessary to use digital media to store the full image data. Also, for fluorography digital formats offer advantages. The image data can be transmitted without adding noise and image processing can improve diagnostic viewing. Techniques of electronic image management and optical disk storage provide a unified and flexible means for integrating dissimilar imaging devices and non-imaging data in a single system.

This paper will be published in Picture Archiving and Communication Systems (PACS) for Medical Applications Part II.

We discuss recent developments in the construction of intelligent systems, and suggest that such an "expert system" would be useful in diagnostic imaging as the central core of a distributed Pictorial Information System Architecture supporting interactive acquisition, archiving, retrieval and analysis of images from various modalities. We review the elements of such systems and the technical, economic and political problems which must be solved before their construction becomes feasible.

Picture archiving and communication by computers needs not only support by the appropriate hardware, but also requires sophisticated software structures, if the needs of a clinical user are to be met in a consistent and flexible way. In the system CA/1 the problem oriented language PROFI-ll is used to implement the necessary structures for data management and communication with the user. This paper gives a short description of the system CA/l, the language PROFI-ll, the implemented data and communication structures and as an example its application for analysis and management of image sequences in digital radiography.

The MITRE Corporation has designed a system for gathering, storing, transmitting over telephone lines, subsequent restoring and recalling radiological cases consisting of multiple images and textual case data. This system has been implemented at four satellite sites and one central reading site. An evaluation is being conducted to determine the accuracy, reliability, user acceptance, and utility of the system. This paper describes the context of the demonstration, describes the system operation and its principal components, emphasizing the technical characteristics of image gathering, processing, transmission, storage and display.

Testing was conducted to obtain qualitative and quantitative (statistical) data on radiology performance using the Remote Medical Diagnosis System (RMDS) Advanced Development Models (ADMs)1. Based upon data collected during testing with professional radiologists, this analysis addresses the clinical utility of radiographic images transferred through six possible RMDS transmission modes. These radiographs were also viewed under closed-circuit television (CCTV) and lightbox conditions to provide a basis for comparison. The analysis indicates that the RMDS ADM terminals (with a system video resolution of 525 x 256 x 6) would provide satisfactory radiographic images for radiology consultations in emergency cases with gross pathological disorders. However, in cases involving more subtle findings, a system video resolution of 525 x 512 x 8 would be preferable.

The high cost of computed tomography scanners has limited their availability to secondary and tertiary hospitals only, although they are very desirable at all medical centers. We are developing techniques which should enable reconstructive tomography to be carried out from a set of radiographs taken at different angles using standard, overhead radiographic units which are available at many primary health care units. By adding tele-communication links between health care centers at remote sites and a central medical computing facility, it would be possible, for example, to provide computed tomographic facilities for the 50 such sites in Manitoba at a capital cost equal to that of a single commercial scanner.

It is indeed a pleasure to be included among the distinguished scientists and radiologists in this First International Conference and Workshop on Picture Archiving and Communication Systems for Medical Applications. Just a few years ago a dedicated computer system together with provisions for an image archival system having ability to transmit images from the radiology department to various parts of the medical facilities was proposed at the Walter Reed Army Medical Center. Unfortunately, that proposal was referred to committee because acquisition of computers had to first go through the Computer Systems Command and the procurement of an image archival system required approval of the Audiovisual Command. Nonetheless, in a very short time, the imaging in radiology has matured to the point that serious movements are underway to automate, process, display, and archive as well as report radiological images and thereby reduce the image problem to a manageable form. Although all phases of diagnostic radiology and parallel imaging modalities have improved monumentally, we remain in a "horse and buggy" era with regard to film display, storage, and conferencing procedures so that too often the image we need for patient care and management is not "where it is needed, when it is needed."

The design and implementation of a distributed diagnostic imaging management system is new. The goal of these distributed systems is to integrate digital diagnostic imaging modalities through a local area network. This paper provides an estimate of the current utilization of digital diagnostic image data and provides the user logical functions required for each node of the distributed system.

System designs for PE-DI-R are being evaluated relative to the optimum configuration to meet the needs of diagnostic radiology at the University of Arizona. It is clear that the system can be examined in terms of sub-systems broadly delineated as image acquisition, processing, display and archival storage. Its architecture is influenced by the diverse imaging requirements of different radiological exams; the extent of processing desired in terms of complexity and speed; the degree of sophistication for displays serving a reading room as compared to clinics, conference rooms and offices; and the instrumentation available for archival storage. Issues have risen pertaining to system architecture, particularly in questioning the advantages of systems stressing a strong computer center versus a distribution of "stand alone" systems utilizing microprocessors. These appear to be resolvable at least in part on the basis of hospital size. The hospital carrying out in excess of 200,000 procedures per year has different requirements than those of a hospital performing only 10,000 exams/year, The University of Arizona Health Sciences Center is a 300 bed hospital which performs 67,000 radiological procedures per year. Its size and activity suggest a mix of computer and microprocessors will be needed to carry out its multiplicity of tasks, We anticipate completing the design and essentially install most components of the system over the next four years. Features of the system that have evolved to date and deemed most pertinent to the interests of the participants will be discussed at the symposium.

The full capabilities of digital image producing modalities are rarely utilized routinely. Only a small fraction of the available information may be used in the initial diagnosis, and subsequent review is often severely limited by the use of film recordings which may display less than 5% of the original data. An alternative is presented, as a unified digital image distribution and processing system, linking various digital image sources through a high speed data link and a common image format. The system allows for viewing and processing of all images produced within the complex, from viewing stations at any number of convenient locations. The physical handling of storage media is totally eliminated. Complete archiving, file maintenance, and large scale processing capabilities are provided by a central file server. Provisions for hard copy, and the input and output of external image data are also included. The cost of such a system and its operation for ten years is shown to be approximately $0.20 per image.

A prototype picture information system should provide a means to gain practical experience in the use of video communications within a medical center, as well as provide a mechanism to accommodate sophisticated video applications in conjunction with clinical procedures. Such a system, being installed at the University of North Carolina, will be described. Installation of the system is initially within the radiology area; however it will be expanded to include other medical center locations once an appropriate system has been fully developed and verified. Various equipment interconnections will be studied, such as T.V. camera to remote monitor and recorder, recorder to multiple remote monitors, T.V. camera to remote image processor and returned for local display, and computer to computer. Technical objectives for the system will be briefly reviewed. A distributive broad band coaxial cable network will provide the flexibility to relocate picture equipment modules at any location within the radiology department as desired. Versatile communications protocols, to include both analog and digital based channels, will allow for the proper interface of image processing, storage and display devices. The architecture of the system will be described.

The use of X-ray film in the diagnostic imaging process is changing rapidly. As electronic archiving and retrieval systems evolve, there will still be demand for photographic hard copy. Consequently, X-ray films are being modified for use as the photographic medium in CRT-to-film recorders. The critical performance criteria of a film recorder are given and the components of the system that are most crucial to performance are discussed.

Distributed picture archiving and communication systems require electronic displays, today probably video displays. An obvious restriction with video displays, especially when multiple images are viewed, is on resolution both along image lines and due to the video raster. The effect of this restriction and the needs for improvement will be briefly reviewed. Perhaps a less obvious restriction of video displays is on contrast, as compared to film. Only limited grey-scale contrast is provided, and the fact that multiple images must be presented near each other on a single display means that the contents of one image will affect the perception of another. The perceptual and display mechanisms causing these effects will be described, means of specifying these contrast effects will be presented, and quantitative measures of the effectiveness of various display systems will be given. Pseudocolor scales provide a means on video displays of lessening the restriction on contrast. However, these scales bring with them problems of associability and variation in relative sensitivity across the scale. A linearization method will be presented for avoiding the second problem and thus allowing the comparison of different scales. Ppproachs for choosing sensitive pseudocolor scales without associability difficulties or contour artifacts will be presented, and specific scales superior to grey-scale will be recommended.

Obtaining hard copy images from computerized medical imaging devices generally involves photographing video displays, often with a multi-image camera. Appropriate film selection for video photography with a negative type film requires considerations of the characteristics of the cathode ray tube (CRT) being photographed and the properties of a video image. Of particular importance is the contrast transfer characteristic of the CRT. In this paper we discuss the CRT characteristics and their effect on film selection. The significant film parameters are reviewed and their effects on film image quality are discussed. Our results indicate that a film for video photography should have high contrast (G = 2.5 to 3.0), relatively high speed, be orthochromatically sensitized and be single emulsion. Appropriate camera adjustment and quality control techniques are of major importance in achieving optimum film images, the former is briefly discussed.

With the continuing trend towards electronic image acquistion, archiving and communication there will be an on-going demand for high quality photographic output as the hard copy record. Film is transparent; durable, and capable of excellent detail and dynamic range for the human eye-brain response system. So, if film has a future, how will it be handled, exposed and processed? For the short range, one to three years, the CRT will still be the best price/performance approach for converting the electronic image into the light energy that exposes the film. With the implementation of coherent light sources in other products such as computer output printing onto plain paper for the office/business environment, the price of lasers and light emitting diodes will be reduced to a point where they can be considered as an alternative to the CRT.

Digital image processing offers an extensive repertoire of acquisition, enhancement and archiving operations to the medical researcher and diagnostician. Real-time picture acquisition and display, the ability to enhance and register pictures, rapid access to archives and fast, reliable picture communication are the more important benefits of digital processing technology for the medical community. Until recently, however, the high cost of digital processing equipment and the specialized skills required for its operation had severely limited its application in the medical environment.

This paper will cover the techniques available today for storing large numbers of digital imagery and also large data bases needed for medical image analysis. Also covered is how real-time roam and zoom can be used to investigate large image data viewed through a display window. A technique for fast communication links, which is in development will be discussed. Finally, the paper discusses what is seen as areas of exploration for storage mediums and communication.

This paper describes a system for rendering objects in three dimensional space using medical data. The combination of a variable focal length vibrating mirror and computer electronics will be described and also the software system required to process the data. A brief review of several promising medical applications will allow visualization as to how such a display can be used most effectively. The display of data in three dimensions overcomes the ambiguity often found in two dimensional presentations. It allows a truly objective examination of the display data while two dimensional displays require a subjective interpretation of what might exist in the Z direction. The basic technologies employed in SpaceGraph were developed at Bolt Beranek & Newman, Inc. in Cambridge, Massachusetts and were proven in a laboratory prototype. Usefulness for a wide variety of applications was also proven by BB&N by obtaining data from many fields of science and creating displays of these data. What we show here are early indications of promise in medical areas that require true three dimensional imaging capability. The display is a true three dimensional display that allows presentation of data in a volume filling manner. The data can occupy a volume of 20 X 25 X 30 centimeters.

Digital image processing systems necessarily consist of three components: acquisition, storage/retrieval and processing. The acquisition component requires the greatest data handling rates. By coupling together the acquisition witn some online hardwired processing, data rates and capacities for short term storage can be reduced. Furthermore, long term storage requirements can be reduced further by appropriate processing and editing of image data contained in short term memory. The net result could be reduced performance requirements for mass storage, processing and communication systems. Reduced amounts of data also snouid speed later data analysis and diagnostic decision making.

In discussing analog or non-digital images it is important to recognize that all images are generated by an interaction between quanta (either light, x or gamma rays, etc.) or particulate radiations, and a substance capable of responding to the interaction of these rays and that substance in such a way that an image can be generated. The intensity of the primary radiation and its spatial variation will then reflect the object being imaged. Therefore, the real difference between a digitized and an analog image lies in the method of recording this data. Many of the systems classified as digital make use of first-stage analog devices from which information is extracted and then transformed to a digitized format. In this paper we will discuss images which have not undergone such a transformation and which have been stored in analog form by an analog storage device. Such storage devices are usually photographic film, video disk, or video tape. It is from these analog storage media that subsequent digitization and/or processing by computer takes place. The choice of the methods of recording and digitization can greatly influence the signal to noise level and therefore the precision and accuracy of the computed results whether expressed as an image or in terms of numbers. Commonly used are video cameras, laser scanners, and photo-electronic scanners. The signal to noise level capabilities of these digitization modalities also influence both spatial and gray level resolutions achievable. In the case of radiological imaging most applications utilizing digitization have involved the use of video. The use of video for such applications has been reviewed recently.' In this paper we shall restrict our discussion to this modality.

One of the most rapidly expanding new imaging techniques in diagnostic radiology is computerized fluoroscopy. In this technique the video images obtained by using an X-ray image intensifier-TV system are processed by dedicated video image processors to enhance certain clinical aspects. The video images acquired by such systems may also be used to extract quantitative information from such images. The physical and technical requirements for obtaining accurate quantitative information will be discussed. The capabilities of the current systems in meeting these requirements will also be presented.

Data and hardware format standards play a fundamental role in determining the cost and versatility of large scale information systems. The use of digital images in medicine is embryonic (first trimester?) and developing rapidly. The time is ripe for consideration of standards. The prospective roles of the Federal agencies and professional groups having responsibilities and interests in this field and the potential for coordination are discussed.

This paper will be published in Picture Archiving and Communication Systems (PACS) for Medical Applications Part II.

This paper will discuss requirements for a protocol for exchanging digital image information between users of possibly dissimilar equipment. The need for simplicity, flexibility and ease of data access will be outlined and several examples will be presented which illustrate how practical issues such as data block sizes and hardware compatibility impact these requirements. An additional requirement for users who must exchange physical media, perhaps on a one time basis through the mail, is that the data format contain sufficient descriptive information so the data may be utilized without reference to a published document. A companion paper describes one implementation of a standard magnetic tape format which meets these requirements.

The increasing need for distribution of images between generating centers and the desirability of using the images within digital analysis systems and digital data bases has led to the need for standardized formats which self-document the image structure while minim-izing the overhead and maximizing the format clarity. For Landsat, such a format has been defined. It is currently in use in Canada, and will be initiated in the United States with the advent of Landsat-D in the fall of )982. Although initially defined for Landsat, the family has specifically been designed to accomodate images from any source, of any size, and with any internal logical structure. For example, the (implied) two dimensional matrix of data points may be alphameric records, outline drawings, point data or data tables as well as imagery. The format family is open ended in that future specific formats may be generated by the use of pre-specified organization rules. The required standard superstructure information provides the roadmap through the data. The format will accomodate multiple images per logical volume, multiple tapes per logical volume, and multiple logical volumes on a given tape.

This paper will describe the details of a standard magnetic tape format for digital image exchange; The topics which will be covered include: key-value pairs, required keys, comments, pixel location, and an overview of the software needed to utilize this standard. It will be illustrated by showing: How the standard would be used on a simple image; How additional descriptive information may be added to further annotate this image (the electronic lab notebook); How a tape is created, How images are added to an existing tape; and, How an image is selected and extracted for the tape. A separate paper will describe the philosophy for this standard, motivating the features of the standard, while this paper will focus on the information necessary for implementation of software to support it's use.

A standardized format is required in order to communicate image information between different systems. It is important that the standard be sufficiently flexible to allow for the encoding of the raw data that was used to form the image. Arguments are presented concerning the importance and value of storing the raw data, independent of its source. Future standards evaluation committees are asked to consider raw data storage flexibility as a selection criterion. An example of image/data storage flexibility in currently used formats are given.

Picture archiving and communication systems, especially those for medical applications, will offer the potential to integrate the various image sources of different nature. A major problem, however, is the incompatibility of the different matrix sizes and data formats. This may be overcome by a novel hierarchical coding process, which could lead to a unified picture format standard. A picture coding scheme is described, which decomposites a given (2n)2 picture matrix into a basic (2m)2 coarse information matrix (representing lower spatial frequencies) and a set of n-m detail matrices, containing information of increasing spatial resolution. Thus, the picture is described by an ordered set of data blocks rather than by a full resolution matrix of pixels. The blocks of data are transferred and stored using data formats, which have to be standardized throughout the system. Picture sources, which produce pictures of different resolution, will provide the coarse-matrix datablock and additionally only those detail matrices that correspond to their required resolution. Correspondingly, only those detail-matrix blocks need to be retrieved from the picture base, that are actually required for softcopy or hardcopy output. Thus, picture sources and retrieval terminals of diverse nature and retrieval processes for diverse purposes are easily made compatible. Furthermore this approach will yield an economic use of storage space and transmission capacity: In contrast to fixed formats, redundand data blocks are always skipped. The user will get a coarse representation even of a high-resolution picture almost instantaneously with gradually added details, and may abort transmission at any desired detail level. The coding scheme applies the S-transform, which is a simple add/substract algorithm basically derived from the Hadamard Transform. Thus, an additional data compression can easily be achieved especially for high-resolution pictures by applying appropriate non-linear and/or adaptive quantizing.

A laboratory prototype system for archiving X-ray CT pictures on digital optical discs is being built up, which employs a TOMOSCAN T-310 CT system for picture generation, a TOMOSCAN stand-alone viewing console (SAVC) for picture retrieval and evaluation, and a Philips digital optical recorder (DOR) for mass picture storage. The picture store is implemented as a separated subsystem equipped with its own minicomputer for administration and control ('picture base'). Communication is made by using a high-speed data link. A superior picture management is performed by employing the existing radiologic data base system RADOS implemented on a separate further computer system. Convenient search routines are thus offered for picture retrieval, and links are made to patient records.

Much of the information in real-time ultrasonic imaging examination is lost when only a few of the scan planes are recorded. Recording of all scan planes is particularly desirable in the examination of the breast, where there are few anatomical landmarks and discovery of small lesions leads to the best prognosis. We have developed a system, that allows such recording and provides for playback in a very flexible manner; the images can be reviewed at a rate from zero (static) to 30 frames per second, forward or backward, with 'instantaneous" reversal or stopping in the progression of images under review. The images are stored as two rings of 120 images each, one ring per breast, on a disc of photographic material. The developed disc is placed in a reader which spins it at a constant 60 rev/sec. By counting the timing marks recorded along with the images (obviating tight mechanical tolerances) any selected image can be flashed onto a TV camera by a strobe unit controlled by the timina-mark count. As there is no mechanical inertia, the rate of change in the selected image can proceed at any rate up to the monitor and is applicable for use with formatters, VTR, etc. , for further image recording. The recording disc is small (8 diameter.) and inexpensive (<$5) and provides an archival unit record for inclusion in patient files.

In this paper we will discuss a plan for an experimental application in clinical context of optical disc storage in diagnostic radiology. First computerized tomography (CT) and ultra-sound (US), later digital vascular imaging (DVI , intravenous angiography) and new digital device images such as digital radiography and nuclear magnetic resonance (NPR) will be stored permanently in a digital way. This image data base must be considered as an extension of the already operational hospital information system (ZIS). Replacing the costly, unique, bulky and operation-intensive film in a versatile, fast access- fast relocatable digital manner has obvious advantages and intrinsic,but limited, disadvantages. The aim of this experiment will be to evaluate easeof use, reliability and economic aspects of laser-disc based picture archiving and communication in a clinical experiment.

The Family of Sonic Mammographic Viewers (SKV-120* and SMV-50* Models, Technicare/Special Research group) contains a unique means of recording the large number of images required to evaluate the entire volume of the breast. The key element of the recording system is a disc of film containing two annuli (one for each breast), each containing 120 saggital images. By stroboscopic playout of the spinning disc, any one image or sequence of images may be rapidly obtained. The details of the Discam* camera used to produce these discs and Discplay* display device used to play them have been published by our colleagues'..

One of the persistent problems facing radiology departments is storage and timely retrieval of images. Usually prudence and the law require that old studies be kept for years, which can often be difficult and costly. Digital storage of images poses even more difficult problems. The cost of rapid access storage is likely to force the development of hierarchical storage of images. This paper examines the problem of which images to migrate to secondary or tertiary storage and the cost of different approaches both in terms of dollars and probability of delayed access. The model used is based on P. M. Morse's model of library book circulation. The Morse model attempts to determine the probability distribution of future use based the past use. It assumes that usage is a Markov process and that the probability that usage will be any particular value is dependent on the previous year's use. Each year's probability distribution is Poisson with the mean governed by a constant decay factor and by a factor which is a multiple of the actual use in the preceeding year. This model will be used as a basis for theoretical analysis to determine the cost of different methods of assigning images to secondary storage.

Since 1977, the Radiology Operations Management Computer System has been in operation at the Hospital of the University of Pennsylvania (HUP). This comprehensive system features registration and scheduling, file room management, automated interpretation reporting, management reporting, patient tracking, accounting and automated teaching file. In 1980, a group of non-profit institutions joined together in a consortium to develop the next generation radiology management system and to take advantage of the collective experience and knowledge of the group while using the HUP system as a model. Although the HUP Picture Archival and Communication System (PACS) was developed independently from this management system, their interdependence is important and will be discussed.

A fully automated and comprehensive Radiology Department system was implemented in the Fall of 1980, which highly integrates the multiple functions of a large Radiology Department in a major medical center. The major components include patient registration, film tracking, management statistics, patient flow control, radiologist reporting, pathology coding and billing. The highly integrated design allows sharing of critical files to reduce redundancy and errors in communication and allows rapid dissemination of information throughout the department. As one node of an integrated distributed hospital system, information from central hospital functions such as patient identification are incorporated into the system and reports and other information are available to other hospital systems. The system is implemented on a Data General Eclipse S/250 using the MIIS operating system. The management of a radiology department has become sufficiently complex that the application of computer techniques to the smooth operation of the department has become almost a necessity. This system provides statistics on room utilization, technologist productivity, and radiologist activity. Room utilization graphs are a valuable aid for staffing and scheduling of technologists, as well as analyzing appropriateness of radiologic equipment in a department. Daily reports summarize by radiology section exams not dictated. File room reports indicate which film borrowers are delinquent in returning films for 24 hours, 48 hours and one week. Letters to the offenders are automatically generated on the high speed line printer. Although all radiology departments have similar needs, customization is likely to be required to meet specific priorities and needs at any individual department. It is important in choosing a system vendor that such flexibility be available. If appropriately designed, a system will provide considerable improvements in efficiency and effectiveness.

Nuclear medicine presents a challenging set of requirements to the computer system designer and a unique opportunity to implement multi-user computer systems in diagnostic imaging. Data are presented to elucidate these requirements by comparison to the other imaging modalities and to demonstrate the growing need for distributed computer services. The efficacies of various system architectures for distributed processing in nuclear medicine are analyzed and an architectural hybrid between a network and multi-processor configuration is proposed. This architecture is analyzed against the current and future processing requirements of nuclear medicine.

This paper will be published in Picture Archiving and Communication Systems (PACS) for Medical Applications Part II.

The Computed Tomography System is aimed at availing the doctor of a non-invasive diagnostic tool to project an image of the patient's anatomy on a interactive display monitor. The total speed of data acquisition to first image is generally the perceived highest system requirement. The System need for archival storage of the images does produce a time limiting restraint on this need. The opportunity for advanced image analysis is at times ignored in deference to film due to the data acquisition impacts on the analysis within a single computer and the clinical needs of the attending physician. The presentation will address the benefits of film, tape and local area networking in the total picture of task distribution, archival storage, and improved analysis availability. It will delve into the dataset sizes and required technology in use within the CT system today. The concepts of multiscanner rooms feeding to a central archival storage facility and/or satellite image analysis stations will be analyzed as to the benefits, costs, and image analysis impacts. Then the projection toward the multi modality analysis system will be visited for benefits and needs. The question of wide area network access needs must ultimately be projected.

Nuclear magnetic resonance is a phenomenon which allows the generation of high quality medical images including the traditional transverse, coronal, and sagittal perspectives, while utilizing no ionizing radiation and without any known adverse biological effect. In addition, NNR imaging and diagnostic techniques allow the generation of basic chemical/pathophysiological measurements which have great promise to enhance the accuracy of medical diagnosis by providing more specific diagnostic information.

The advent of computed tomography (and other imaging techniques), the availability of interactive raster scan graphics display, and the rapidly diminishing cost of memory have made possible new approaches to long-standing problems in radiation therapy. These include: the secure definition (and display) of anatomy in all three dimensions; the synthesis of a variety of diagnostic studies; the identification of a target volume and of critical adjacent structures; the simulation (in three dimensions) of any available radiation field; the development of tools for guiding the treatment set-up and for its confirmation; and the comparison of alternative treatment strategies. Remote interactive planning is also desirable. We describe the implications for image archiving in providing these capabilities.

Electronic methods of imaging cancer patients can be superior to conventional radiographs in accuracy and speed of handling. Diagnostic video images can be displayed simultaneously with simulated treatment images or with actual treatment images to facilitate comparison and increase accuracy. Live display of patients undergoing treatment can be accomplished by using a therapy beam fluoroscreen. By simultaneously presenting images of the original treatment setup along with the live display, duplication of the original setup can be verified. These images can be stored on disk for permanent record. They can be played back by the physician at his convenience in order to check daily patient setups. Gamma camera images can be used to localize brachytherapy implants.

The ideal ultrasonic image communication and storage system must be flexible in order to optimize speed and minimize storage requirements. Various ultrasonic imaging modalities are quite different in data volume and speed requirements. Static imaging, for example B-Scanning, involves acquisition of a large amount of data that is averaged or accumulated in a desired manner. The image is then frozen in image memory before transfer and storage. Images are commonly a 512 x 512 point array, each point 6 bits deep. Transfer of such an image over a serial line at 9600 baud would require about three minutes. Faster transfer times are possible; for example, we have developed a parallel image transfer system using direct memory access (DMA) that reduces the time to 16 seconds. Data in this format requires 256K bytes for storage. Data compression can be utilized to reduce these requirements. Real-time imaging has much more stringent requirements for speed and storage. The amount of actual data per frame in real-time imaging is reduced due to physical limitations on ultrasound. For example, 100 scan lines (480 points long, 6 bits deep) can be acquired during a frame at a 30 per second rate. In order to transmit and save this data at a real-time rate requires a transfer rate of 8.6 Megabaud. A real-time archiving system would be complicated by the necessity of specialized hardware to interpolate between scan lines and perform desirable greyscale manipulation on recall. Image archiving for cardiology and radiology would require data transfer at this high rate to preserve temporal (cardiology) and spatial (radiology) information.

We have designed and implemented a system for the centralized acquisition, display, analysis and archiving of diagnostic cardiac medical images from x-ray fluoroscopy, two-dimensional ultrasonography and nuclear scintigraphy. Centered around a DLC PUP 11/34 minicomputer with an existing gamma camera interface, we have added a closed-circuit television system with a 256x512x8-bit video digitizer and image display controller to interface the video output of the fluoroscope and ultrasonograph. A video disc recorder (under computer control) is used as an input and playback buffer, allowing for data transfer to and from digital disc drives. Thus, real-time video digitization is possible for up to ten seconds of incoming RS-170-compatible video. The digitizer separates video fields at real-time into two 256x256x8-bit refresh memories, providing 60Hz temporal resolution. Generally, however, we choose to record at non-real-time rates to encompass more than ten seconds. In addition to I/O software controlling data acquisition ana playback, we have developed a versatile data analysis package (offering such capabilities as image algebra, Fourier analysis and convolutional filtering), as well as interactive data reduction subroutines (such as region-of-interest definition, profile plotting and regional extraction of statistical and probabilistic information). We have found the system useful for standard cardiac image analysis, for simultaneous display of images from the three modalities, for picture storage and retrieval, and as a research tool. future plans include the addition of intelligent terminals at each modality and progression to a 32-bit machine for the central processor.

The application of biomedical techniques developed in physics, chemistry, engineering and computer sciences has been the subject of this symposium. In general, these provide exciting new opportunities for participants in the health care field but they also represent challenges. This communication will discuss the basic principles of the laws of agency and evidence and how these might be applied to the specific technology discussed in this symposium. Since the resources necessary for these imaging devices and other improvements in data acquisition, information transmission and archiving are of such magnitude, public policies have been enacted to assure both public access to and protection from inappropriate acquisition and distribution of this technology. We will consider the combined effects of these initiative in relation to the present social changes providing the environment in which these policies will be enacted.

Emerging technologies such as inexpensive, powerful local computing, optical digital videodiscs, and the technologies of human-machine interaction are initiating a revolution in both image storage systems and image interaction systems. This paper will present a review of new approaches to computer media predicated upon three dimensional position sensing, speech recognition, and high density image storage. Examples will be shown such as the Spatial Data Management Systems wherein the free use of place results in intuitively clear retrieval systems and potentials for image association; the Movie-Map, wherein inherently static media generate dynamic views of data, and conferencing work-in-progress wherein joint processing is stressed. Application to medical imaging will be suggested, but the primary emphasis is on the general direction of imaging and reference systems. We are passing the age of simple possibility of computer graphics and image porcessing and entering the age of ready usability.

With the advent of digital radiography and the installed base of CT, Nuclear Medicine and Ultrasound Scanners numbering in the thousands and the potential of NMR, the market potential for the electronic management of digital images is perhaps one of the most exciting, fastest growing (and most ill defined) fields in medicine today. New technology in optical data storage, electronic transmission, image reproduction, microprocessing, automation and software development provide the promise of a whole new generation of products which will simplify and enhance the diagnostic process (thereby hopefully improving diagnostic accuracy), enable implementation of archival review in a practical sense, expand the availability of diagnostic data and lower the cost/case by at least an order of magnitude.

Increasingly sophisticated diagnostic imaging systems are being acquired by hospitals. The purpose of this paper is to identify the sources and types of clinical images in an academic, tertiary, acute-care, general hospital servicing a 600 to 700 bed population. An estimate is provided of the digital image information data that is being generated by these hospitals. The problems of digital archiving and area networks for successfully managing this large amount of image information will be difficult to achieve.

Current technological developments suggest that in the very near future all medical imaging modalities will have been converted to digitally based systems requiring no film or other intermediate media for either data recording or information storage. However, the full potential of medical imaging to diagnostic medicine will be realized only with the integration of the various imaging modalities into a single unified imaging system. Clearly the computer, and in particular, distributed computer systems, will play a central and unifying role in all future medical imaging systems. This paper outlines the architecture of a particular distributed system for the acquisition, processing, and filing of medical images produced by several different imaging modalities. Although the system is based in part on digital optical disc technology, other storage technologies could also be adopted. Methods for the standardization and unification of the various imaging modalities will be described. This unification is centered on key computer, microprocessor, and communication elements. The picture or image base is managed by a superior data base which permits user-oriented hierarchical access to data and related pictures. Separated hardware levels are provided for management, control, and signal processing. The processing level is equipped with image processors based on unified modular architecture. This paper also describes a possible historical scenario indicating how the integration of the various imaging modalities might be accomplished. The scenario begins with ultrasound, continues with CT, NMR, and nuclear medicine imaging, and concludes with digital radiography.

Management of film files has problems of slow access, large space requirements, loss, and manual administration. Digital imaging modalities such as CT and DVI are posing new image management requirements. The contrast resolution of these systems is larger than can be represented on CRT monitors and multi-format films. Therefore, it is necessary to use digital media to store the full image data.

This paper not available at publication time.

System designs for PE-DI-R are being evaluated relative to the optimum configuration to meet the needs of diagnostic radiology at the University of Arizona. It is clear that the system can be examined in terms of sub-systems broadly delineated as image acquisition, processing, display and archival storage. Its architecture is influenced by the diverse imaging requirements of different radiological exams; the extent of processing desired in terms of complexity and speed; the degree of sophistication for displays serving a reading room as compared to clinics, conference rooms and offices; and the instrumentation available for archival storage. Issues have risen pertaining to system architecture, particularly in questioning the advantages of systems stressing a strong computer center versus a distribution of "stand alone" systems utilizing microprocessors. These appear to be resolvable at least in part on the basis of hospital size. The hospital carrying out in excess of 200,000 procedures per year has different requirements than those of a hospital performing only 10,000 exams/year, The University of Arizona Health Sciences Center is a 300 bed hospital which performs 67,000 radiological procedures per year. Its size and activity suggest a mix of computer and microprocessors will be needed to carry out its multiplicity of tasks, We anticipate completing the design and essentially install most components of the system over the next four years. Features of the system that have evolved to date and deemed most pertinent to the interests of the participants will be discussed at the symposium.

A prototype electronic picture information and communication system (PACS) should provide a means to gain practical experience in the use of video communications within a medical center, as well as provide a mechanism to accommodate sophisticated video applications in conjunction with clinical procedures. Such a system under current development at The University of North Carolina, will be described. Installation of the system will initially be within the confines of the radiology area; however it will be expanded to include other medical center locations once an appropriate system has been fully developed and verified. Dr. Edward Staab, presented an overview, describing radiology departmental PACS philosophy and program development. This report will expound upon the development of the prototype PACS to include conceptual approaches for a departmental design technique, communications network development, and test programs and project progress to-date.

Andre Duerinckx, PACS '82 Chairman. I encourage everyone to really participate. The panel members will make short statements and then we would like to get some lively discussion.

The utilization of x-ray video images for providing quantitative information poses more strict requirements on such systems. In this paper, a discussion of some of the important factors is given. An understanding of some of the limitations places an upper bound on our expectations.

This paper not available at publication time.

"Digital Radiography" is a generic term which has evolved recently to aescribe a collection of novel radiographic modalities where the primary image recording is accompiisheo through the use of x-ray-sensitive electronic detectors, rather than by the conventional use of phosphor screens ano silver-halide-based films. The electronic aetectors provide occult electronic signals which must be processed to render them ultimately visible and thus accessible to the oiagnostician -- in the short term for oiagnosis, in the long term for reference. The type of imaging involved is metric both in the spatial ano the luminous dimensions, with strong requirements for imaging repeatability and image element signal integrity.

The application of biomedical engineering advances to health care delivery has provided both opportunities and challenges. One of these has been considerations of resource acquisition and distribution. Guidelines, agencies, modalities and devices have become increasingly apparent. Applications of the laws of agency and concepts of vicarious responsibility are particularly relevant. Certificate of Need legislation has led to antitrust considerations that are exceedingly complex. This presentation will attempt to evaluate some of the legal aspects resulting from the acquisition, allocation, and application of the many exciting technologies that are the subject of this symposium.

Andre Duerinckx, PACS '82 Chairman. I think we can start with the second panel discussion, which will be the last activity of this conference. The Chairman of this Panel is Dr. Bill Glenn from Multi-Planar Diagnostic Imaging Inc. I want to say that what you are about to hear or say will be recorded, so please do use the microphones; don't say anything without using microphones. The same goes, of course, for the panel members.