Picture Archiving and Communication System
A picture archiving and communication system (PACS, about Bildablage- and communication system ) is in medicine a picture archiving and communication system based on digital computers and networks. The first PACS developments began in the 1970s. However, it did not find significant distribution in hospitals and doctor's offices until the late 1990s.
PACSs acquire digital image data of all modalities in radiology and nuclear medicine . In principle, images from other imaging processes, such as endoscopy , cardiology , pathology and microbiology , can also be used for PACS processing.
Individual computer systems that are permanently connected to a single diagnostic device and perform PACS tasks are called mini-PACS.
A PACS consists of the PACS server to which a short-term and a long-term archive is connected. The PACS server sends to the viewing and post-processing computers, but also communicates with the connected imaging modalities. In most cases, there is also a connection to the Radiology Information System (RIS) . Larger PACS installations can also consist of several u. There may be servers and archives coupled over long distances.
The DICOM and HL7 standards have been developed by international consortia to enable the various components to be integrated with one another and to enable PACS to be embedded in hospital information systems . The IHE (Integrating the Healthcare Enterprise) is an organization that combines various standards into so-called application profiles. A PACS can then correspond to one or more of these profiles.
The most important prerequisite for the establishment of PACSen was the DICOM standard, because uniform DICOM communication enables PACS servers and imaging devices to be used independently of the manufacturer and the connection of a device becomes simple and inexpensive. Modern large-scale medical imaging devices such as CTs , PET / CTs , MRs or SPECT cameras consistently deliver image data in digital form in accordance with the DICOM-3 standard. Similar to the Exchangeable Image File Format , an image or a series of images consists of two parts: In addition to the actual image, a wealth of information is stored in the so-called DICOM header. These are u. a. the identity of the patient, examination date and time, clinical question, type, type and manufacturer of the device used, but also the name and address of the examining institution. If images that are available as film recordings have to be captured, they are digitized using a scanner that receives the information for the DICOM header from the RIS. The image is then transferred to the PACS.
Unfortunately, the standard was sometimes not adhered to, especially with older devices, or not all fields were filled with information, which led to communication or storage errors. The device manufacturer often only offers the option of saving the image data in the DICOM standard (overpriced). To date (as of 2011) there are imaging devices such. B. (OR) microscopes or endoscopy devices that do not provide their image data in the DICOM standard. In this case, the image information can be captured via frame grabber cards and converted into DICOM format with the help of special software products.
The functionality of converting non-DICOM images into DICOM format or storing them without conversion is increasingly being offered by PACS manufacturers and mapped in the archive memory.
Server and storage
The core of every PACS is the PACS server. All modalities of a PACS environment deliver their images here and they are also saved here. In practically every hospital in industrialized countries today, all radiology image data is stored in the PACS. In a typical 400-bed hospital, this amount of data is around three to five terabytes per year. The radiology archive of a university hospital can therefore be several hundred terabytes in size. The exact size and quantity of the resulting images varies, however, depending on the type and number of connected modalities. A modern 64-line CT produces a multiple of the images that an older 4-line CT outputs. The amount of data in a mammography image is also considerably larger than that of a conventional X-ray image.
In the memory of the PACS archive, the image data are sometimes no longer in the DICOM data structure. In some systems the PACS server receives the DICOM data, separates the header and image data and saves both information - sometimes compressed - in a common database . In addition to the information from the DICOM header, further information such as changes or shifts in the image is stored there. If the images are called up by a remote station, they are converted back into DICOM format for transmission.
Since the PACS server makes the image data available to the entire institution, its failure means that no images are archived and - apart from the current images on the imaging device itself - cannot be viewed. The PACS therefore not only has to be designed with a high storage capacity, but also with a high level of reliability.
The image archive is usually divided into short-term and long-term storage. The long-term memory contains images that are older than the time specified by the system administrator - typically half a year to two years. In this way, the long-term storage can be implemented more cheaply than the short-term storage. The short-term memory is designed in such a way that these frequently accessed image data can be accessed very quickly. In older PACSs, very slow tape drives or CD jukeboxes were sometimes used for long-term storage. Today (2012), however, RAID archives are used in most cases for short-term storage and long-term archiving . The long-term archive is ideally designed as a RAID 61, while the short-term archive is often a RAID 51.
State-of-the-art technology today (2011) are RAID systems and HA clusters that are mirrored over two spatially separate locations and coupled via high-speed fiber optic connections (> approx. 4 Gbit / s) . Each individual mirror consists of a fully functional PACS, but is also designed redundantly. In the event of a hard disk failure, an immediately available spare disk is activated, the so-called hot spare . Network components such as B. switches or fiber optic cables are available twice. If the error is so serious that the system is no longer functional, the mirror server is activated without interruption.
The server and RAID controller have two to three power supply units so that the system continues to run even if one power supply unit or a circuit fails. Ideally, both the redundant components of a server and the mirror server are supplied by separately fused circuits and are connected to an uninterruptible power supply.
Errors occurring in the system are automatically triggered by e-mails to the system administrators, who can then initiate suitable measures to correct errors without delay. Despite these safety precautions, the image data are usually also backed up on tape drives so that in the very unlikely total failure of the storage system, backups of the images are still available.
With RAIDs, in which the hard disks are operated well beyond the time of two to three years used by the manufacturer for calculating the MTBF , there is the possibility of losing the data of a RAID despite the mechanisms described. Image data that is more than two years old is very rarely accessed. This means that the disk load on the long-term archive is usually very low. The disks of the RAID could therefore age or wear out without this showing up in a disk failure. This creates the false impression that the RAID disks are OK, although they may no longer be. If a disk fails, the RAID controller starts to rebuild the data stored on the failed disk with the help of the hot spare . The rebuilding of this data from a defective disk means, however, a very heavy burden on the remaining disks; thus the stress on the remaining functioning disks caused by the build-up can lead to a failure of another board before the rebuilding is complete. Since the data of a RAID 5 would be lost in such a case, long-term archives are usually implemented as RAID 6. A RAID 6 can also cope with the simultaneous failure of two hard drives. The rebuilding of data from defective disks can take 10 hours or more. When using a RAID 6, the risk of total data loss in the scenario described is small, but not completely negligible. This is one reason why it makes sense to swap the disks of the RAID of the long-term archive in good time, even if they are not defective, and to implement the long-term archive in the form of two spatially separated RAID 6s.
Workstation computer for image viewing and post-processing
The examinations are called up on special workstation computers. While the communication between the PACS archive and the medical device follows the DICOM standard, data between the workstation computer and the archive are sometimes transmitted in proprietary formats. The reason for this is that network communication via DICOM query / retrieve or store is not very effective; H. a lot of information is also transmitted, some of which is redundant and / or not relevant for the image display. Likewise, with a transmission according to DICOM 3 images are transmitted in a series one after the other, while with a transmission via e.g. B. HTTP any image in a series can be transmitted with priority, which shortens the loading time. Therefore, the software of the workstation computer and that of the archive are generally from the same manufacturer. The WADO standard has been around for a few years for HTTP-based transmission in accordance with the DICOM standard.
If necessary, images are digitally post-processed during display: Usually, the assignment of measured values (X-ray absorption, signal intensity, etc.) to gray values is manipulated ( windowing ) or subsequent structural measurements are carried out. After examining the images in the light of the medical history, the radiologist creates a written report. For this purpose, a RIS client and speech recognition software are usually installed on the PACS workstation computer.
Images and reports can also be viewed on less elaborately equipped workstation computers in the ward and polyclinic area. A conventional PC is usually suitable for this; the image viewing then takes place either with the help of a small special application or in the web browser.
Networking with other IT systems
Even with the first PACSs, it was considered to be closely networked with other IT systems, but this was initially only possible to a limited extent due to the lack of relevant communication standards. Due to the constant further development of DICOM and HL7, it is mostly possible today (2011) to closely interlink HIS , RIS and PACS.
Example: A radiological request that an employee registers in the HIS is passed on to the RIS, and the examination data is linked to the PACS studies. This means that today (2011) it is possible, for example, that the radiological findings stored in the RIS can be called up together with the PACS images in the HIS and thus in the entire hospital. The close RIS-PACS interlinking also enables PACS image data to be called up from HIS and RIS. The doctor only selects the patient and, if necessary, the examination, the PACS immediately displays the associated image data. Corrections of errors are also simplified by the close interlinking of the systems. If the patient's name is accidentally misspelled when registering in the HIS, a correction of the patient data in the HIS triggers an HL-7 message, which is forwarded to both the RIS and the PACS. A name correction "Maier zu Mayer" only has to be done once in the HIS, RIS and PACS are automatically synchronized.
A paradigm shift is currently taking place in the PACS architecture. While many PACS are implemented as deeply interlinked solutions with proprietary interfaces, users and experts consider the breakdown into the three components workflow ( RIS ), archive (VNA) and viewer to be advantageous. The operators put together the ideal PACS / RIS according to the best-of-breed strategy and link the components with standard interfaces ( DICOM , HL7 , IHE ).
In contrast to image documentation on paper or film carriers, PACSs work with digital image data. This opens up extensive possibilities for increasing the functionality and efficiency of work processes.
Thanks to digital storage, the quality of the recordings remains unchanged for many years. In the case of projection radiographic methods, digital recording enables a higher contrast range. Recordings are therefore more informative, repeated recordings after incorrect exposure less common than with film radiography.
For sectional image methods, there are expanded options for viewing and diagnosis. A series of cuts can be displayed as an animation or converted at any time into an MPR or into a 3D model or post-processed with special evaluation software. PACS also simplifies the documentation of moving recordings during ultrasound.
A major advantage is the simultaneous availability of images at several locations (even within a hospital) via a computer network , which completely eliminates the logistical effort for conventional image transport. Since the images can also be reproduced over greater distances, the assessment can be made more flexible in terms of time and space (see also teleradiology ). Recordings can be copied without loss. Cumbersome film archiving is no longer necessary. The risk of losing unique original recordings is minimized.
Savings in image media, transport costs and archiving space are further significant advantages.
In the first years of the PACS introduction, PACS environments often suffered from poor DICOM implementations, so that apparently DICOM-compatible devices could often not be connected or data could only be stored or read to a limited extent. The computer and network architectures of the first PACS servers and workstations were also unable to cope with the size of the image data, which led to excessively long transfer and loading times. A high maintenance effort, high system prices and sometimes low reliability led to the fact that PACSe and the PACS concept were often heavily criticized and the benefits of PACS were questioned.
In spite of great hardware expenditure and redundant servers, it is possible that the PACS cannot be accessed. Causes can be: physical interruption of network connections, failure of switches, incorrect configuration of, for example, newly added devices or the crash of server services. Such an interruption usually affects several, in the worst case all users of the PACS. As with other central, server-based IT systems, the resulting loss of productivity is therefore usually high.
Although medical image data could already be stored digitally in the 1970s, attempts to establish digital archives remained, in the absence of standardized data formats, proprietary isolated solutions. It was not until the development of the DICOM standard started in 1983 that manufacturer-independent PACS development was also possible. Since the first devices that could output their data in DICOM format did not come onto the market until 1993 with the adoption of DICOM 3.0, which is still valid today, PACSes were hardly to be found in hospitals even towards the end of the 20th century. In 2005, only an estimated 22% of all North American hospitals had a PACS.
When the DICOM standard was adopted in 1993, the data transfer rate of Ethernet was 10 Mbit / s. Even high-priced workstations had a few dozen megabytes of RAM and only a few hundred megabytes of hard drives. In 2011, the common transfer rate of Ethernet was 1 GBit / s: an inexpensive "off the shelf" PC had several gigabytes of RAM, hundreds of gigabytes of hard drives and computing power that was several orders of magnitude higher than computers from the 1990s. The amount of image data in radiology has also increased over the course of the twenty years of PACS development, but not by far as the performance of computers and network infrastructures increased. A chest x-ray is just as big today as it was 20 years ago. In combination with the constant drop in the price of IT systems as well as further developed and better adhered to communication standards, this led to a very widespread use of PACS. In 2011, the benefits of PACS (apparently) clearly outweighed the effort involved.
In 2016, Oleg Pianykh, professor of radiology at Harvard Medical School, published a study on unprotected PACS servers. At that time, he had found more than 2,700 open systems. In 2019 there are unprotected systems in 52 countries around the world with over 24 million data records and more than 700 million linked images. Of these, 400 million are actually downloadable.
Classification as a medical device
PACS software (e.g. for archiving and diagnosing image data) is generally marketed as a class IIa medical product throughout the European Union . If the PACS software has an influence on the mode of action of a medical device connected to it (e.g. for functionalities for radiation therapy planning), it can be assigned to class IIb. The conformity assessment procedure is subject to monitoring by a so-called notified body . Medical devices such as PACS may consist of several components that can be assessed individually. These components can therefore also have different classifications depending on their respective purpose.
- Dreyer, Keith J. [Eds.]; PACS (a guide to the digital revolution), ISBN 0-387-26010-2
- HK Huang; PACS and Imaging Informatics, ISBN 0-471-25123-2
- Imageeconomics.com The Prophet Motive: How PACS was developed and sold ( Memento of the original from October 5, 2011 in the Internet Archive ) Info: The archive link was automatically inserted and not yet checked. Please check the original and archive link according to the instructions and then remove this notice.
- An independent loan word has emerged from the English acronym in German usage . In this article, therefore, PACS is used as a noun with a capitalized first letter.
- The DICOM header
- Healthimaging.com The impending deconstruction of PACS
- PACS Administrator Blog Deconstructed PACS ( Memento of the original from February 12, 2015 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.
- Insecurely configured servers are leaking data from millions of patients heise.de, September 19, 2019