The other sign of evolution in PACS design is the move towards an open architecture with multivendor equipment based on industry standards for interconnection of imaging devices and other PACS components. This trend is made possible by a similar evolution among the manufacturers. Major medical manufacturers tend to support established standards and to be less protective about the integration of their equipment with other equipment allowing imaging data to be transferred to other computers or devices. The emergence of standards from a joint effort between the academic community, the manufacturers and the computer industry have played a major role in this evolution (1). This allows a more realistic approach in a real clinical environment where different image acquisition equipment is rarely purchased from a single vendor but, in most cases, radiological departments are equipped with heterogeneous devices from different vendors. PACS implementation must adapt to the established fact that most hospitals and clinics cannot commit to a single vendor. An integration effort is therefore required to connect all devices to a common image archival and distribution network.
Multimedia Medical Record
The rapid development of computerized information systems and medical informatics provide access to a variety of documents in different forms. In particular a large number of medical data are generated in the form of static or dynamic sets of images. Clinical investigations rely more and more on imaging modalities and the clinicians are faced with the difficult task of integrating very large amounts of data from increasingly complex imaging techniques. The user must be able to visualize and compare images from different imaging modalities and to correlate what can be observed on these images with the rest of the clinical and ancillary data. Multimodality workstations capable of handling all this data on a convenient and practical platform will soon become a necessary tool for practicing physicians. The first step toward such workstations is achieved through appropriate management of images from different imaging modalities together with the rest of the medical data. It has become widely recognized that the real added value of computerized image management and communication systems (IMAC) comes from the integration of images and image related data with the rest of the clinical data that constitutes the medical record (2, 3).
Figure 1: Schematic diagram showing the medical data flow around the patient record
(adapted from Lemke et al. (3))
The contents of a medical record consist of data provided from different investigation techniques as well as collections of notes, reports and observations (see figure 1). The medical record also contains chronological observations as well as treatment plans. The completeness of the medical record that should contain all the medical and therapeutic events of a patient's lifetime is essential for adequate patient care. A computerized patient record must therefore offer this completeness in order to fulfill the clinician's needs. It must handle multimedia data from different sources (text, sounds, physiological signals and images) but it must also offer a coherent link between all these data. The tools offered to the users to browse through these data and visualize them in a convenient way are as important as the quality of the data itself. Clinical users must be able to access this data in a natural and convenient way that matches the convenience of the paper-based medical record.
Most studies on PACS requirement showed that several types of workstations are necessary. For primary diagnoses the radiologist needs a workstation that provides the best possible performance with the highest resolution. For review and clinical conferences the requirements are less stringent and do not require the same high performance. Finally a third type of workstation can be envisaged for image processing and analysis. This third category should provide more specific tools for image analysis of certain types of images and provide the best possible user interface to allow easy access to these tools for non computer-oriented users.
In 1984, the Japan Industries association for Radiation Apparatus (JIRA) created a committee to establish a medical image transfer standard for Medical Image Processing System (MIPS), to be used as a standard for transfer of both image data and demographic data for medical imaging hardware. Soon afterwards, the ACR/NEMA published its own standard. The current version, is basically similar to the ACR/NEMA version 2.0 minimum requirement, with some minor modifications, mainly with regard to the handling of text data.
Two other committees were also founded in Japan: the committee for Personal Health Data (PHD) and the Image Save And Carry (ISAC) committee. Doctor Ikeda from the National Cancer Institute of Japan proposed in 1982 that every citizen should have his medical record on digital media. Dr. Ikeda and his colleagues proceeded with the development and evaluation of the PHD system between 1982 and 1987, supported by the Japanese Ministry of International Trade and Industry (MITI) and the Japanese Ministry of Welfare. The work done by this committee has been followed by the ISAC committee for image data handling.
The development of worksstations capable of handling medical data in different forms such as text, graphic, images and sounds requires careful user-interface design to be easily accessible by clinicians and non computer-oriented users. Effective management of medical images from different imaging modalities is the first step toward the integration of multimedia information. It is also the most challenging part because the amount of data obtained in the form of images is much larger than text, graphic or sound elements. The proportion of clinical investigations based on imaging modalities is also constantly increasing. Image display and manipulation on graphic workstations is certainly a technical challenge. The major difficulty comes from the large size of the image data and the high performance required to display and manipulate them in a convenient way. Two papers presented in this special issue have focused on this problematic aspect of workstation design. The project under development at the university of Berlin is defining and implementing the basis for multimedia workstation in a medical environment. The OSIRIS software developed at the Geneva University Hospital focuses on the management and presentation of medical images from different imaging modalities on a variety of workstations. It is being used as part of a hospital wide PACS directly integrated with a distributed HIS (7).
The components required for a multimodality workstation are becoming widely available on today's computer systems. Most systems can handle high resolution graphics, sounds and animations. The challenge is to use these capabilities properly. Software and hardware capable of handling animated images can be used in a large number of applications in medical imaging. The paper by Tessler et al. in this special issue clearly illustrates a practical utilization of such capabilities in displaying dynamic ultrasound images. Conversely, some multimedia capabilities available on most of today's computer systems are not always appropriate for handling all types of medical images. The particularity of medical images is that they require a higher dynamic range than most other, non-medical graphic applications. A typical medical image contains data that is 10, 12 or 16 bits deep corresponding to 1024, 4096 or 65536 grey levels. This dynamic range is not usually supported in most multimedia systems. These systems are more often geared toward video animation and tend to support images with a large number of colours (16 million colours on 24 bit RGB systems) but only 256 shades of grey. In order to properly visualize medical images special software (or hardware) tools are necessary to allow the user to interactively adjust the window of grey levels of an image that are displayed at a given time. This is particularly true for radiologists and other clinicians responsible of primary diagnosis tasks for which they need the full information contained in a medical image.
Another aspect of multimedia workstations in a clinical environment is their accessibility. Physicians and nurses tend to move frequently from one location to another and do not have a fixed workspace where they spend most of their time. Therefore, a single personal workstation on a desk is often not enough, and the clinical users need to be provided with workstations distributed in different locations of their work environment. In order to become really useful and efficient, access to the patient record must be possible hospital-wide all the way to the bed side. The wide distribution of workstations capable of accessing data from different distributed archive systems requires appropriate management of the data flow and an efficient networking system. Portable and movable workstations have often been proposed to allow users to place the workstations in different locations where and when they are needed. Hand-held systems as well as notebook computers connected to database servers through wireless networks are expected to become really handy in a busy clinical environment. Fixed workstations will remain, however, in areas where user interactions with the system require more complex and repetitive tasks. Primary interpretation of radiological studies as well as quantitative analysis of the images are bound to be performed in a fixed location on high resolution and preferment workstations especially tailored for these tasks.
As mentioned earlier, the difficulty of setting up a distributed HIS and PACS resides in the complex task of distributing the information to the users in multiple locations. This requires appropriate usage of high performance networks. The network topology and the effective throughput must be carefully planned to ensure adequate response time. Although high speed networks can provide very high data throughput, a wide distribution of large amounts of data cannot rely only on a single network. Multi-tiered networks combining different networks with different bandwidths are necessary (8). In order to obtain the best performance it is necessary to separate the data traffic into multiple local networks in different sections of a large institution. Incoming data must also be separated from data being distributed to remote consultation stations. In a distributed architecture the traffic between different database servers and file servers must be optimized to allow for a better communication between these servers independently from the data acquisition and distribution workload. It is often difficult to accurately predict the real needs in traffic throughput for a given system in a given environment. This is particularly difficult when dealing with images because the volume of data is several orders of magnitude higher than the one necessary for text and numerical data. Several computer simulation techniques have been proposed in an attempt to create realistic models of PACS networks (9). These models are however often simplified and limited to a very particular setting representing a subset of an institutional wide network. Besides, it is difficult to accurately predict the workload variability related to a nonuniform access to the data by large number of users at different times. Accurate planning of large medical networks requires a combination of both, computer simulations and experimental measurements. As far as images are concerned, recent observations tend to show that different network performance are required in different tasks of an IMAC system. Data acquisition and archives are usually fairly constant in data rate which can easily be handled by relatively low performance networks. Traffic between different database servers requires higher throughput that should not suffer from degradation of performance when the data rate increases. The most unpredictable parts of the network are the segments responsible for data distribution to display workstations especially in a multimedia environment where the exact proportion of images and text required at different times by different users can vary considerably. Depending on the clinical task they have to perform, users may have different requirements. In some instances the workload can be handled through relatively slow networks such as Ethernet and in others the data throughput must be one or two order(s) of magnitude higher to allow for rapid transmission of very large amounts of image data. A schematic representation of a simple multi-tier network combining different networks with different data throughput is shown in figure 2.
Figure 2: Example of a simple multi-tier network where different communication tasks are handled by different networks with different data throughput.
In combinations with high speed networks, rapid transmission of large amounts of data can also be achieved through data compression. The largest amount of data that can benefit from data compression in a patient record are the images. Images can be compressed either through reversible lossless compression or through lossy compression. Lossless compression algorithms can only achieve relatively low compression rates (ratios of 2:1 up to 4:1). Lossy compression algorithm can provide much higher compression rates. However, through lossy compression, images are slightly degraded. The challenge is to determine the amount of image degradation that is applicable without affecting the diagnostic quality of the images. Recent studies have demonstrated that large images such as chest X-rays could be compressed using full frame discrete cosine transform (DCT) at a rate of 10 to 1 without affecting the diagnostic quality of the images (10). An alteration in diagnostic quality was found around a ratio of 20 to 1 for these images. However, these compression techniques are rarely used because they are computationally intensive and require special hardware implementation to be applicable to large images in a clinical routine. Besides, these techniques require extensive clinical evaluation experiments to determine the degree of compression that is applicable without interference with the clinical diagnosis. To be statistically significant, these experiments must be carried out using a large number of images from different modalities with different clinical findings that must be evaluated by a large number of observers (11). While these clinical trials are still underway, several new developments occurred in the area of data compression. These developments come from a rapid evolution of the television and movie industry toward digital solutions. New compression algorithms have been developed and implemented with much higher compression ratios than the simple full frame DCT. Besides, combined progress in the computer industry and in digital video and data communication have lead to new hardware solutions based on these new compression algorithms. Some of these new concepts in image compression are discussed in the paper by Cicconi et al. in this special issue. It is predictable that some of these new developments in the area of image compression will rapidly be applicable to medical imaging systems.
The Internet revolution
The recent fantastic development of Internet and World Wide Web has opened new perspectives in the domain of data exchange and data communication around the world. The rapid evolution of the Web has extended the possibility of communications between a large number of users far beyond the academic community where it was initially confined. The major innovative aspect of the WWW is to offer standardized interfaces for viewing and linking multimedia documents including images, texts, sounds and animations through standard protocols. The wide availability of software programs that allow to display and manipulate such data on different hardware platforms made that evolution very attractive to the whole community. The hypertext concept that allows to link parts of a text document to other documents including images and sounds has opened a new prospective in the way one should navigate across the different texts of data. The WWW browsers have adopted a new and attractive concept of triggering different programs for displaying different texts of data. Every time a new type of document is referred to, corresponding software can be added to the pool of programs that are referred to often as help us that can be figured from the browsers to display the new type of data. This is obviously a very attractive mechanism for displaying data such as medical images of medical documents. By designing an appropriate tool for displaying and manipulating medical images and linking them to existing WWW browsers, it is possible to use the existing Web for communicating and remote consultations of medical documents. Such an approach was recently explored and tested at the Geneva University Hospital where the conventional PACS environment was replaced by a prototype of the WWW browser that directly triggers a specific program for displaying medical images from the conventional netscape or Mosaic browsers. The program that displays the medical images called OSIRIS is the one that is widely used in the Hospital for viewing images. The only difference is that instead of going to tedious query of data bases and file transfers to retrieve images from different patients, especially designed interface written HTML can be used from any conventional WWW browser on any platform. Also because the OSIRIS program is written for the different hardware platforms (Macintosh, PC Windows, Unix) it is ideally suited to be used as a document viewer in the conventional Web architecture.
Figure 3: Web server implemented at the University Hospital of Geneva for in-house access to radiological images and reports as well as laboratory procedures and results
Further developments were carried out at the Geneva University Hospital to insure appropriate confidentiality and access rights control before being able to use the Web environment to explore and display medical images and medical documents. The standard accessory verification mechanism was implemented into an HTML environment similar to what is being used in the more traditional hospital information system in Geneva. The same rules and reservations for accessing medical documents through the HIS system are applied in the corresponding Web environment. Furthermore, in order to protect any standard access, the WWW server for the medical document is only accessible inside the Hospital.
Figure 4: Two different steps that allow hospital users to access patient radiological files through an in-house web server that can be accessed from any platform through the Netscape interface. Appropriate access right control is performed before these pages can be displayed.
Beyond just using the Web browsers inside the Hospital for retrieving and viewing parts of the medical computerized record, the Internet network offers a much more attractive pass for exchanging medical documents on a much larger scale. Such a possibility was recently explored in Geneva through the implementation of a public server that would provide access to medical information and also possibilities of confidential exchange of medical documents. Such a server is made available to the whole medical community to be able to access through Internet to a more convenient and low-cost access to different information as well as date communication. A large number of relevant information and up-to-date medical data can be made available and retrieved on any computer that are nowadays widely available in private practitioners' surgeries. Besides, a special proprietary communication protocol was set up to allow confidential exchange of data such as lab results, reports, images and other graphics through exchange of the corresponding files that can be further consulted on any computer using the appropriate viewer linked to the WWW browser. This type of solution should replace advantageously pre-existing tedious and proprietary communication protocols that are often used to exchange data between different labs and clinics. The standardization efforts that are currently on the way in the medical community are also very important for allowing the exchange of documents across different platforms.
Figure 5: Example of a public server implemented in Geneva allowing private physicians and medical institutions to exchanges medical documents and computerized patient records through the Web. The page on the left shows the home page for accessing general information and references and the one on the right shows a personalized page that displays the files that were sent to a given physician (images, texts, reports and lab results)
Computer assisted diagnosis
The extension of computerized medical records and the access to clinical data through multimodality workstations will certainly have a significant impact on physicians' work habits and will change the way they practice medicine. Medical informatics tools will soon become part of everyday medical practice. It is therefore natural to foresee that efforts will be developed to provide the medical users with advanced tools to assist them in data analysis, knowledge finding and decision making. With the development of digital imaging techniques in medicine, quantitative analysis techniques have emerged in a large number of applications. Data extracted from the images as well as appropriate image processing techniques can assist radiologists and clinicians in providing more accurate, objective and reproducible image interpretation. Furthermore, the integration of data obtained from different imaging procedures as well as other clinical investigations can provide the user with a better synthetic view of the data. However, the increasing number of highly specialized investigation techniques, often leave healthcare providers with a very difficult task of matching all the results into a coherent clinical diagnosis. Computer assisted interpretation of these data can help the clinician by providing the appropriate interpretation criteria and by matching them with epidemiological and statistical data as well as relevant diagnostic rules obtained from knowledge databases.
Figure 6: Access to medical images in digital form and being able to process and analyze them on a workstation opens a whole new era of computer assistance to the physicians for image interpretation and diagnostic tasks.
Other domains that are extending beyond the diagnostic assistance are the computer simulation domain and the virtual reality domain. Computer simulation can help to analyze and predict physical and physiological phenomena. They have been widely and successfully used in areas such as biochemistry, microbiology, physiology and pathophysiology. Similar tools are expected to become very useful in clinical practice as well. They should help to predict the clinical evolution and effects of therapeutic operations based on statistical and rule-based models. Such tools are designed to help answering "what if ?" questions in complex situations based on concrete statistical data and sophisticated models. They are expected to become more widely used in every day practice as soon as the clinical data as well as the statistical and knowledge databases become directly available in computer-accessible form.
Similarly, computer graphic simulations also known as virtual reality tools can provide nowadays very powerful means for exploring three dimensional anatomical and physiological phenomena. Although we are still a few years away from being able to practice a complex surgical operation through computer simulation before performing it in real life, some preliminary developments have already shown very promising results. The paper presented in this special issue by Thalmann et al. clearly shows the trend of new applications in medicine for computer animation and virtual reality developments. These techniques are already being used in many technical applications for training individuals in highly skilled tasks such as airplane navigation, mechanics and industrial engine manipulation. Preliminary work will soon offer medical students virtual three-dimensional anatomical models that will help them to better visualize and understand the human anatomy. With the increasing sophistication of such models, it is expected that surgery practice, non-invasive operations and even diagnostic procedures such as echography will be possible on virtual models in the near future.
Recent studies demonstrate that the cost-benefit of PACS should not be evaluated as a resource of the radiology department alone but should extend to the entire hospital (2). The main benefit of PACS is expected outside the radiology department where clinicians will benefit from a more efficient access to the images. A combined access to the images and related data such as the radiological report, annotations and analysis results is a key feature of the beneficial effect of PACS on the clinical routine. Substantial added value also comes from the image manipulation and processing features that can be provided to the clinical users outside the radiology department. It is therefore important to allow for a flexible design of such tools for easy customization of the workstation software that must be adapted to the different users' needs.
The development of PACS in a clinical environment will allow physicians and radiologists to assess radiographic images directly through imaging workstations. The development of medical workstations was primarily oriented toward the development of a convenient tool for rapid display of images (12). There is, however, an increasing need for image processing and quantitative analysis. Therefore PACS workstations can be divided into two categories depending on their specific usage: 1) high resolution viewing stations: their function is to replace the conventional "viewbox" commonly used by physicians to review film-based images. These stations must provide a resolution sufficient to match the diagnostic quality of conventional films. The number of interactions and manipulations required from the users are usually reduced to a minimum to allow for a more efficient utilization in a busy clinical environment (13). 2) image review analysis workstations usually designed for remotely reviewing images as well as for the evaluation of specialized imaging modalities that require more quantitative analysis of the images. This type of workstation must therefore provide a large number of image processing and analysis tools. The design and performance of these two categories of workstations are quite different due to the difference in utilization. Furthermore, the second category will have different utilizations ranging from simple viewing stations in clinical wards to sophisticated image processing and analysis stations in specialized units such as cardiology or nuclear medicine.
During the past decade a large number of image analysis tools were developed to improve the diagnostic accuracy of different imaging techniques (14, 15). Most of these techniques required sophisticated software packages that were often only accessible to research centers or clinical setups that benefit from large computer and software support. The programs for image analysis were often implemented in complicated systems which required special training of the users. They were usually developed on special computers with software tools that were neither modular nor portable on other computer systems. With the development of PACS and the wider availability of imaging workstations it is our goal to make these tools more widely and easily available.
Most software applications rely nowadays on Graphical User Interfaces (GUI) where the functions are activated through icons and selectable menus, and triggered through mechanical devices such as a mouse. The goal of GUI is to create tools that rapidly become transparent to the user allowing him or her to achieve complex tasks with the minimal amount of technical training.
Is the future already here today ?
New technical developments in medical informatics, digital imaging, computer simulations and distributed database access lead me to believe that we are progressively moving towards an era where computer assistance is slowly becoming omnipresent in medicine and healthcare. Recent developments in the computer industry have taught us not to expect big revolutionary changes in technology but rather to look for progressive and constant improvement in performance and capabilities of computer systems. We can expect to move into tomorrow's world without really noticing it. What seems today to be a technological breakthrough such as being able to manipulate multimedia documents on a single workstation, will soon become part of our daily world. It will only make the usage of hospital information systems and clinical workstations more natural and user-friendly.
As a last remark I would like to state that multimedia communication and information systems will come into our daily life whether we want it or not. It is our task now to clearly identify the areas and applications where this new technology can help us in everyday practice. It is also important to adapt multimedia applications to the user's needs rather than watch the users progressively adapt themselves to a technology-driven evolution of their environment.
2. van Gennip E.M.S.J., van Poppel B.M., Bakker A.R., Ottes F.P. Comparison of worldwide opinions on the cost and benefits of PACS. Dwyer III S.J. ed. Medical Imaging V, San Jose, CA: Medical Imaging V: PACS System Design and Evaluation, SPIE: 1446:442; 1991.
3. Lemke H.U., Osteaux M. PACS and digital imaging-New directions (Editorial), Europ. J. Radiol. 71:1-2; 1993.
4. Duerinckx A.J. Picture Archiving and Communication Systems. Duerinckx A.J. ed. 1st Int. Conf. and Workshop Picture Archiving and Communication Systems (PACS) for Medical Applications, Bellingham, Wash.: Society of Photo-Optical Engineers; 1982.
5. Horii S. An Introduction to the ACR-NEMA standard. H.K. Huang O.R., A.B. Bakker, G. Witte ed. Picture Archiving and Communication System (PACS) in Medicine, Evian: NATO ASI Series, Springer-Verlag, Berlin, Heidelberg, 119-140; 1990.
6. NEMA. Digital Imaging and Communications. In: ACR-NEMA standard publication: 1988. (Invited Speaker)
7. Ratib O., Ligier Y., Girard C., Perrier R., Logean M., Vurlod J.-F. A picture archiving and communication system based on an open and distributed architecture. 14th Annual International Conference of IEEE Engineering in Medicine and Biology Society. Paris: IEEE. Inc,:1197; 1992.
8. Stewart B. Three tiered network architecture for PACS clusters. Picture Archiving and Communication System (PACS) in Medicine. Evian: NATO ASI Series, Spinger-Verlag, Berlin, Heidelberg, 113-119; 1990.
9. Stut W.J.J.Jr., van Steen M.R., Groenewegen L.P.J., Bakker A.R. Picture archiving and communication system design issues. The Importance of modelling and simulation. J. Dig. Imaging 3(4):246-253; 1990.
10. Chan K.K., Lou S.L., Huang H.K. Radiological image compression using full-frame cosine transform with adaptive bit-allocation. Com. Med. Imaging Graphics 13:153-159; 1989.
11. Harrington M.B. Some methodological questions concerning ROC analysis as a method for assessing image quali y in Radiology. J. Dig. Imaging 3(4): 211-218; 1990.
12. Huang H., Mankovich N., Taira R., et al. Picture Archiving and communication Systems (PACS) for Radiological Images: State of the Art. CRC Critical Reviews in Diagnostic Imaging 28(4):383-428; 1988.
13. Wendler T., Grewer R., Monnich K., Svensson H. Design Considerations for Multi Modality Medical Image Workstations. Hohne K. ed. Pictorial Information Systems in Medicine, NATO ASI Series: Springer-Verlag, Berlin, Heidelberg, 401-420; 1986.
14. Ratib O., Rutishauser W. Recent Developments of Cardiac Digital Angiography. Journal of Cardiac Imaging 1:29-40; 1985.
15. Ratib O. Computer Graphic Techniques Applied to Medical Image Analysis. In: Advances in Computer Graphics. Springer-Verlag, 546-554; 1987.
|http://www.hon.ch/Library/papers/ratib.html||Last modified: Wed Mar 12 1997|