This activity may be available in multiple formats or from different sponsors. ARRT does not allow CE activities such as internet courses, home study programs or directed readings to be repeated for CE credit in the same or any subsequent biennium”

Objectives

Outline

Introduction

Section I: Network Architecture and Topology

Section II: Network Media

Section III: Archiving Storage and Retrieval Systems

Section IV: The cost effectiveness of PACS

Summary and References

Bottom

Objectives:

• Define what is meant by network architecture and topology.
• Discuss how cabling affects the speed and form of data transferred over the PACS network.
• Discuss network-attached-server archiving and SAN network archiving.
• Discuss the various aspects of protecting stored image data.
• Discuss how data is transmitted to the Web server.
• Discuss why open networking is not advisable for PACS, e.g. internet browser.
• Discuss the cost effectiveness of PACS and CR.

Outline

1. Network Topology and Architecture
   1. Main topologies used in LANs and PACS networking.
   2. Bus Ring, Star, and Tree designs.
   3. Example of a topology strategy.
2. The role of Network cabling in data transfer
   1. Network media
   2. Data Transmission
   3. Inherent characteristics of media
   4. Network cabling
   5. Network switches and media transfer
3. Archiving, storage, and retrieval
4. Cost effectiveness of PACS
5. Introduction to Digital Radiography

Introduction

In this unit we will discuss some of the finer points of the PACS network such as the network topology, cabling, storage devices and data checking as it is stored and retrieved from memory. By now is has become obvious that PACS is necessary to view soft copies of radiographic images. However, it is not necessary just because radiography equipment is digital. The cost of film and the related processing is quite expensive. One of the fundamental questions PACS team members frequently ask is: "What equipment should we have purchased for the needs we have now?" What is available to an institution now entering the PACS and digital imaging arena is the wealth of learned experience out there that can give concrete advice. Whether one is looking for a PACS for small, medium or a large institution the foundation of experience is well laid. The main goal for all imaging departments should be to become digital and PACS supported as soon as is possible. The fact remains that medical imaging has moved rapidly towards digital imaging as evidenced in current medical literature. This fact alone supports digital imaging as the new standard in medical imaging. Newly educated radiologists are almost all trained from digital radiographic images and how to read from a PACS workstation.

Currently digital imaging progress, including PACS archiving and viewing is so fast that using an analogy to human growth would be that for every year of PACS development has been equivalent to 15 human years. An institution that is 6-7 years into digital imaging supported by PACS is equivalent to growing from a baby to being a 75 year old. It is not the purpose of this self study to convince institutions that they should become digital and have a PACS network; however, it stands to reason that PACS is here to stay and its progression is not a parallel course along side screen-film imaging and darkroom technology. By and large, screen-film imaging has already become a near antiquated, but still popular imaging practice.

In this module, it is important to understand what makes the PACS network a tool that works smoothly together with many modalities. One of the most important benefits of having a PACS system is that the efficiency of the radiology department is multifold increased. Time technologists spend couriering image documents between physicians and departments, and the file room is time that is lost. This may be recovered by a well educated PACS team and knowledgeable technologists. Where so many institutions fail is the proper education of their technologists to all functions of the PACS system. It is important that the technologist grow in their understanding of digital imaging and PACS; not just "how to push the correct button." The institution should train the technologists fully in usage of the PACS components to achieve the full benefit from the system.

Section I: Network Architecture and Topology

Through much trial many PACS Administrators have designed their PACS system to arrive at an understanding of what may be the best way to set up a PACS network. An institution that is not heavily invested in PACS hardware may wish to have an independent analysis of their institution’s short and long term needs prior to settling on a particular vendors (such as Kodak, AGFA, or Fuji) network strategy, Basically, the wonderful problem with PACS is that a digital imaging department can increase its throughput 3-4 times to that of film processing departments. This means that there follows a needs demand for expanding imaging communications with time. Greater demand on a working PACS system such as wanting all wards of the hospital to have a clinical workstation, or wanting to go global through an intranet connection to various physicians’ offices, etc in a sense screams for network expansion. How the PACS network is designed and set-up from the beginning can affect the ease at which an institution may grow. There are so many variations on a PACS design that a consultation with an independent appraiser is worth the savings an error could cost.

There are four main topologies used in setting up a Local Area Network (LAN). Picture archiving and communication systems (PACS) are LANs and can be expanded to a Wide Area Network (WAN) through a Web based server system. The organization of network computers and their cabling connections causes it to take shape. The shape of the network as various nodes are connected to it is called the networks topology. The computers and peripheral components of a network that are connected to the cabling system are called nodes. There are four main topologies used in LANs: Ring, Star, Bus, and Tree.

Ring

The ring design is based on a closed loop in which each computer component is connected to two others. This is a very expensive network set-up and is difficult to install and maintain. Computers in a ring topology are somewhat protected from hackers since there is no outside connections.

In a ring formation, the network is connected to itself only. The nodes must take turns sending and receiving data with a token. The way data is sent is that a token and data are sent as a unit to the first node, which subtracts information addressed to it from the packet; it also adds any data it wishes to address to another component in the ring and sends it along with the token. All computers on the ring network must wait until they receive the token since only the node with the token may receive or transmit data. Once the first node is done with the token and data packet it passes them both to the next node. The second node performs its work on the data and then passes it along with the token to the next node, and so forth.

Star

The star is a computer set up in which a central hub connects each of the computers in the network. This type of topology is inefficient because data must travel through a hub, or central device, but it is easy to install. The hub does not perform any action on the data such as filtering or segmenting, it is simply a junction that joins all the different nodes together coordinating their communications.

Bus

The bus design uses a central network cabling system, such as Ethernet, to which component computers are connected so that they share their information and functions. This is the most common type of topology for PACS. Most PACS networks are designed as a bus-star topology on an Ethernet backbone; this provides a versatile network environment with growth potential. The ability to expand a PACS network, such as adding another node like a second MRI scanner is important to long term planning. This need is easily accommodated for by a bus design. The figure below demonstrates a bus Ethernet network topology.

Tree

The tree architecture is a combination of the bus and star topologies. Most PACS systems resemble something of a mixed topology design that emerges because the network is developed and expanded over time as more locations such as surgical suites, physician lounge, Web, and remote physician’s offices are added.

Now that we have looked at the four common types of network topologies used in local area networking, let’s look at an example of a real functional PACS system and relate its topology to its function of image transfer, then we will discuss the media that connects the various components. As it is with biological systems, so it is with PACS, structure is function. To understand the function of a network topology we will begin with the entry of data into the PACS network (image capture). Let’s assume our patient makes their first entry into our imaging system through a visit to the emergency room. As seen in the example below any modality can be a point of entry for image capture, CR, ddR, surgery C-arm, angiography, digital fluoroscopy, MRI, CT, etc. PACS must create a new folder for the patient that will be always be their medical imaging file.

The 1st step for the PACS server is to create an electronic file for the patient using their medical record number and pertinent text data inputted from the RIS/HIS server. Because the requisition containing the physician’s orders originated from the ER, the PACS Workflow server will route these images back to the ER-DICOM View Clinical Workstation, and also to the Matrix server, and to various memory locations within PACS collectively called storage.

Because PACS is linked to the RIS/HIS server, specific routing instructions are automatic because these fields are populated with preset instructions on what to do with the data. These functions are performed by workflow management software that is customized by the PACS team for institutional efficiency. The server must assign an electronic file to the data if one does not already exist. It does so by capturing the data from existing fields in the RIS-Server. Once a file is created the server must do something with the data. It will send it to the destination determined by the patient’s requisition origination location. In our example the patient is located in the emergency room; therefore the patient's images are routed to the ER workstation(s). Remember, the main PACS server can store a tremendous amount of imaging data on its short-term storage drive(s). So these images will reside there as well as be transferred directly to the DICOM view clinical workstations for viewing by ER physician(s). Simultaneously the images are sent to a long term storage (LTS) device such as the network attached server (NAS), optical jukebox, digital linear tape (DLT) systems, etc., and to the Matrix Server.

A matrix server is a separate optional server on to which images captured by the main server are copied. Because images are first received by the main server, then to the matrix server which is to itself a parallel network, any workstation that accesses image documents from a matrix server will have faster access to them because of direct routing to and from it. Generally speaking, only diagnostic workstations are connected to the matrix server so the radiologist has priority viewing of newly acquired images over clinical workstation viewers. Having a matrix server adds speed to the transfer of images to the diagnostic workstations because the server does not have to perform any other services within the PACS network. Thus, a matrix server is a dedicated server to the diagnostic workstations

The 2nd step is for the PACS server to do something with the data it receives. It will route it to storage, to the matrix server if there is one in the network, and to the preset workstation so images can be quickly accessed from a pull down menu worklist.

One of the problems with high volume PACS networks is that the main server is overworked because it pushes image documents to all network computers and receives from all base devices and network servers as well. In spite of high volume cabling and fast input/output (I/O) computer systems, image files are large computer documents and can be delayed in reaching the diagnostic workstation where radiologists view and interpret the images. For this reason many of the larger more established PACS dependent institutions are adding a matrix server to their PACS architecture. Basically what a matrix server accomplishes for the system is that data is copied from the main server to the matrix server. From the matrix server diagnostic workstations receive their images and control commands. It makes the overall network faster since the transfer of data is quicker between the matrix server and diagnostic workstations. As DICOM view clinical workstations get older they are being replaced with clinical workstations tied to a matrix server. Older systems relying on distributed DICOM view workstations; these receive images and data management from the main server, and must compete with all of the main server’s functions: capturing images from all modalities, transferring data, archival, and retrieval, and displaying images on workstations. It is ok for these images to take a little longer to reach clinical workstations where physicians view images for gross abnormalities. The radiologists on the other hand need images quickly in order to interpret images during life threatening trauma, and emergency “stat” reads. With a matrix server all radiology modality images are routed to diagnostic workstations in fewer steps since the matrix server is dedicated to workstation viewing, and not storage, printing, web browsing, and like trafficking. Prior studies can also be queried to the matrix server as a prefetched function. The process of prior images being queried to the PACS workflow server and distributed to diagnostic workstations is called Prefetching. The net result is faster throughput of images and therefore diagnostic readings.

It is important for data to be pre-routed because it avoids redundancies in processing. With pre-routing, images are sent to specified destinations so that when the requisition is shuttled between radiographer and radiologist the images are there. It makes for departmental efficiency, as data is where it needs to be timely manner. This includes prefetched study files. As crazy as it sounds, physicians who are accustomed to viewing softcopy PACS images have absolutely forgotten how long it takes for 300 CT images to printed, organized, and labeled. In addition, the film folder must be pulled from archive, the order requisition prepared, and films hung on viewboxes before they are ready for diagnostic interpretation. Only then, can general physician viewing begin in the film file room. Even with PACS, doctors want images faster. It is for this reason that matrix servers are now being installed, and radiologist love them because when the call come in concerning a patient’s image study they can view it on the workstation in a few seconds rather than minutes. Now let us revisit how our network topology looks and summarize why it works so well!

Just a glance at the diagram above makes it easy to see why seasoned PACS radiologists love the Matrix server set-up. Data travels faster to the diagnostic workstation because is out of the competition for the main server’s workflow of image transfer, storage, and archival/retrieval trafficking other nodes on the network must wait for.

Web Server

The last topic in our topology study is the WEB server. We have already alluded to the idea that a separate server is required for wide area networking by either Intranet or Internet customer servicing. The importance of a Web server is that the Internet uses a special Internet protocol for data transfer. This is an uploading protocol that prepares data to arrive at the recipient’s address in the order that it is sent, as data will take different routes on such a large network. Its kind of like going from designated location in New Hampshire to one in Alabama, there is several routes one can take on bus; and you need to arrive at each stop on time to make the next bus on your route. Data moving along on the Internet is sort of the same way. It will take different routes to your computer, but your computer knows the order in which to assemble the data so the document holds its integrity. It does this using a transmission protocol that is different than the DICOM protocol of PACS.

A special server that is formatted to receive and translate DICOM documents and to perform the requirements of Transmission Control Protocol/Internet Protocol (TCP/IP) is needed. Once the Web server has properly prepared the data it can be transmitted over the Internet. Another basic difference regarding a Web Server and the main server is memory capacity for storing image files. The server is best used as a device that formats data rather than an archiving device. Images can be stored on the server since a patient who is visiting their physician at a linked off campus site may view the images over several days or weeks during consultations. By storing image files on the Web server they are readily available for viewing without load time waiting. Web servers generally have limited memory pertaining to image document storage; therefore studies sent to it will be erased on a first-in-first-out basis or after a predetermined number of days on the server. The Web server at Regions Hospital, Saint Paul, Minnesota, is programmed to erase any image documents that are over 90 days on the server not accessed. However, a worklist registers the file so that if it is accessed at any time thereafter it can be retrieved from the main network through the Web server.

Section II: Network Media

The amount of data transmitted in a fixed amount of time is called the bandwidth. It is not enough that computers are capable of transmitting data at fast speed; all components on the network must also be capable of transmitting at at comparable speeds, this includes the cables over which data is interchanged. There are several intrinsic factors that affect the rate at which PACS images are transferred over the network, among these are: the bandwidth, component devices input speeds, and CPU output speed of each node. As a general rule the movement of data over the PACS network is only as fast as the slowest component transmitting at any given time; however, the reserved capacity for transmitting data is an inherent property of the type of cables connecting the nodes. What this means is that if a base device such as the CR reader is only capable of transmitting data at a rate of 100 kilobytes per second, and the media at 400 megabytes per second, and the PACS server inputs at 75 megabytes per second, then the network’s speed during CR transmission is only 100 kilobytes per second. The cables’ connecting the component nodes together is termed the media. If the MRI scanner transmits at 100 megabytes per second but its images follow the CR images, then when it begins to send images to the PACS server the network speed increases to 75 megabytes per second.

Understanding network cabling for PACS can be somewhat complex, but bear in mind that all we are talking about here is bidirectional movement of data over special cable lines. By comparing the cables to a road we can draw an analogy of how data traffics on network cables. It stands to reason that more lanes and wider lanes will traffic vehicles more efficiently than a single lane road. And it likewise follows that if the highway is divided so that bidirectional traffic does not interfere with each others passage there will be better average flow. In our case, the PACS network is capable of trafficking information “bits” at supersonic bidirectional speed depending on the type of media connecting the nodes.

Also important to data transmission is the input and output (I/O) speed of the central processing unit (CPU) of each computer on the network. For example, the CPU of the computed radiography reader will send images to the PACS server (output), and the CPU of the PACS server will receive the images and process them into the network for storage and display (input). The I/O speed of all devices on the network and the media between nodes are the transmitting factors that that affect how long it takes to route data. The topology of the network is also a factor in communication speeds between the nodes.

Most PACS topologies are Bus layout; therefore it is extremely important that I/O speed is relatively high from base unit devices to the PACS server and from the server to other network devices. This is another reason why the DICOM standard is so important, affecting component connectivity and workflow management between devices. Different vendors know that their devices must output and input data at fast network transfer rates. Data is transferred over the PACS network using several types of cables, mainly: twisted pair, coaxial cable, fiber optics, and fiber optic channel.

In spite of all the planning that goes into building a long term PACS network, unexpected growth patterns do occur as PACS become more popular within an institution. Four main problems are associated with network growth: scalability, latency, network failure, and collisions. In a network with hubs, limited bandwidth is available since it is shared among all the nodes; this property is called scalability. Sometimes entire networks must be redesigned to accommodate growth associated with poor bandwidth performance. Latency is the amount of time that it takes for a data packet to reach its addressed destination. For most transmitting protocols a node is required to wait until the network is clear before it may transmit. If a large data file, such as a MRI/MRA study is sending across the network, other nodes must wait for their opportunity to send. This is a primary reason that sometimes the radiologist waiting to read a “stat” image document may have to wait. There is competition for the CPU’s functions since the main server must also store and retrieve documents. A shared bandwidth that is sending a large MRI file may cause a delay for the CT scanner that is trying to send a few images of a trauma “Head.” And if a simple portable chest radiograph gets behind a large volume of CT images of a trauma C-spine, T-Spine, Chest, Abdomen, and Pelvis” studies, the delay may be noticeable. This could cause frustrations if the chest radiograph is also a “stat” read. But fortunately a delay is temporary and is measured in a few minutes rather than in half hours as with conventional imaging.

It is important that all devices on the PACS network have compatible input/output speeds. Network failure is a term to express incorrect speed setting of a device on the network. If the hub passes information a 10-Mbps, then each device must not send at a speed greater than the hub speed or network failure may occur. All device connectivity issues must be addressed to the PACS administrator before equipment is purchased.

Media (Cabling)

There are several types of cables used for PACS that are designed to have fast image data transfer speed. The relationship of cable size to image transfer rate is a lot like resistance in a circuit, the bigger the cable the more traffic it can handle. Standard Ethernet can traffic only 30-420 kilobits per second which is fine for a simple home computer. PACS requires FAST Ethernet which is 10x regular Ethernet throughput but it requires 100-Base-T interface and a hub. Other line such as phone company cables uses Asynchronous Transfer Mode (ATM) which is a switching protocol that transfers data in packets of fixed size. ATM uses a fixed route between nodes. Switched Ethernet on the other hand uses point to point dedicated Ethernet connection with one full line to each node. Data transferred over the WEB is sent in a series of packets over a packet switched networks. Now it is not important for us to distinguish these different types of cables, this is handled by the PACS administrator and the vendor. But we should understand that depending on how the network is designed and how far apart the various nodes are, the type of cable becomes increasingly more important. We can however, rely on principle of getting the largest cable ones budget will allow and grow into it.

Along with cabling considerations is the question of how does data respond when it reaches an intersection to avoid collision? There are several mechanisms in place which include switches, and transmission protocols. A switch is analogous to a train railing system in which the train is switched to another track to avoid collision and timing delays. The same is true of transmitted data in a packet switched network; a switch allows the data to have the full bandwidth to it. In a hub setting data shares all of the bandwidth so that if 10-Mbps is the bandwidth, then each node may only get a portion of it over which to transmit and receive. As you can begin to see, during the design of a PACS network carefully consideration must be given to planning long term growth.

Basics of Data Transmission

There are two methods of data transmission: Analog or Digital signals. For many years music, video, radio, telephone technologies and the like have used analog transmission. We still enjoy T.V and radio programs throughout our day because of analog signals that reach us. However, today the absolute gold standard for data transmission is digital signals. The best way to understand the difference between the two is to look at their amplitude as displayed by an oscilloscope. An analog signal is transmitted using variable voltage creating a continuous waveform. Analog waveform results in an inaccurate transmission whose peaks and valleys are not exact during transmission cycles reflecting variance in the data. (See figure below). These variances are not apparent to the ear listening to music or the eye watching a television show.

By comparison, digital signal pulses are not continuous like analog signals and their amplitude have discrete values of 0 or 1 which can be made consistently exact. We discussed in part III of this module that digital data is composed of binary code called "bits," which can have a value of either "0" or "1."

The exactness of data transmission is related to the frequency of transmission and noise from sources other than the network itself. The transmission frequency is expressed as a unit called the Hertz (Hz). The Hertz expresses the relative number of times the signal changes its amplitude over a period of time. The amplitude for digital signals is the change from zero to one, and for analog signals, from their lowest negative peak to their highest positive peak. As signals are propagated away from the source the signal’s strength diminishes slightly in quality due to noise distortion. Noise interferes with the exactness of the signal because it is an extrinsic force that distorts the signal. Therefore, along the transmission path the signal must be amplified to maintain its strength. Noise affects both analog and digital signals, which must be augmented periodically as its distance from the source increases. The effects of noise continue even though it is augmented so that the signal must be enhanced several times before reaching its intended reception device.

A unique property of a digital signal is that it is not amplified over its distance traveled; instead, it is replicated through a process of repetitive retransmissions of each digital signal. Like its analog counterpart, a digital signal is acted on by intrinsic and extrinsic forces as its distance from the source of the signal increases; however, because these signals are transmitted at discrete amplitudes of either zero or one, an exact signal can be regenerated and retransmitted by a device called a repeater. The signal is maintained through a process that replicates and transmits each signal along the network exactly as its original, having the same discrete amplitude of either "0" or "1."

Various modems like an analog–to-digital converters and the digital-to-analog converters also functions as modulator/demodulators between the transmitting and receiving ends of the transmission cycles to terminate the signal after its reception. Terminating digital signals within the PACS network after the server has stored them in memory is extremely important. These signals if allowed to circulate throughout the network would cause collisions, confusion, and network failure.

Inherent Characteristics of Media

Considerations such as the bandwidth and the throughput of the cables connecting the various nodes are also important to the PACS design. And there are cost, size, and scalability issues that must also be addressed when building a PACS network. The goal is to have reasonably fast transmission speeds and high noise immunity. The capacity of a media to transmit data in a given time period is its throughput, whilst the bandwidth measures the difference between the highest and lowest frequencies a media can transmit. Therefore, the range of frequencies a media can handle is directly related to its throughput. Consider the diagram below:

The PACS administrator and information systems engineers must decide on the what is the best type of cabling for their institutions productivity. It is generally advisable that the biggest cable possible should be installed, that is greatest throughput and bandwidth possible within institutional cost restraint. While there is a cost associated with the best media, maintenance, and support, the cost of lower transmission rate and obsolescence due to infrastructure problems is greater.

When considering a PACS system the size and scalability of the media is also important. Scalability refers to the maximum network length, maximum nodes per segment, and maximum segment length. These factors affect the latency of the network or delay between the transmission of a signal and its reception to addressed device. The cabling may need conduits to help provide noise immunity within the media network. Media cable must be protected from Electromagnetic Interference (EMI) and Radio Frequency Interference (RFI) that can affect signal transmission by creating noise.

Network Cabling

The type of network cabling is important to transmission, for example, with baseband transmission digital signals are pulsed across wire using direct current (DC), where as broadband transmission uses radio frequency (RF) transmission to modulate analog pulses of different frequency ranges. The media used in PACS must move images and text data in compatible form free of noise (noise immunity) and outside interference. The basic types of cables used in PACS networking are Coaxial cable, Thicknet 10Base5, Ethernet Thinnet (10base2), Twisted-Pair (TP), Category 3 cable, CAT 5 cable, and fiber optics to name a few. It is not important here to know the difference in the different types of cabling; however, it is important to understand that the speed at which data is transmitted is dependent on the output device’s speed, input device’s speed, cable transmission speed, and scalability of the network. The cables are like a highway over which data travels. It stands to reason that if there are many lanes, traffic can move faster with relatively free information flow; however, it the lane is single and a lane is occupied by a slow moving vehicle, there may be travel delay, especially when traffic is heavy like during rush hours. When deciding on the cabling for a picture archiving system one should go with the greatest cable capacity and bandwidth possible. Bear in mind that eventually all modalities will be merged into PACS. Modalities like MRI, CT and Angiography usually have large image files that will need high bandwidth in order to have a reasonably good throughput capacity.

Concerning networking media, the two pictures above provide an analogy of trafficking digital data throughout a network. Wide roads with several lanes can handle more traffic at a faster speed than can a single lane with some vehicles going at a slower pace and cannot be circumvented because traffic coming in the opposite direction as well.

Let's look at a few of the types of network cables and the characteristics of cabling. Cables used in PACS networking are twisted within a sheath and are usually paired, and are called twisted-pair (TP) wire. They consist of color-coded pairs of insulated copper wires twisted about each other and insulated in a plastic coating. Twisting helps to reduce the effects of crosstalk. Crosstalk is a term for interference caused by signals on other wires in the casing that infringe on another pair’s signals. There can also be alien crosstalk caused by signals from adjacent cables influencing another cable’s transmission. Twisted pair lines are also be shielded using a thin foil layer and a thick external sheath.

Twisting of the wire helps prevent crosstalk between adjacent wires in a sheath. This is important as image and text data is transmitted over the wire. The twist ratio is the number of twists per meter or by the foot in a twisted pair cable. A twisted pair casing can have two pairs or four pairs as in CAT5 cable.

To manage network cabling, it is necessary to be familiar with those standards used on networks, particularly as they apply to cabling types like fiber optics, category 3 (CAT3) and Category 5 (CAT5) cables, and Coaxial. Different types of cables are used for LAN and WAN and between the two types of networks a modem may be needed. For example, PACS data storage devices are sometimes arranged on a separate sub-network dedicated to storage devices only called a Storage Area Network (SAN). Because digital image files are large, it is important that if a SAN architecture is used that it supports high-speed data transfer. This will require fiber optic lines and a Fibre Channel Architecture in order to have full duplex (bidirectional) data transfer rates of up to 100 megabits per second. The structure of a fiber optic line is shown in the two diagrams below:

Fiber optic Cable (left) has a glass core surrounded by a glass shield called the cladding; these are enclosed in a protective jacket.

A fiber optic fiber can carry a single frequency light, or multiple frequencies can be transferred by a multimode fiber system. Bidirectional data and multiple frequencies are common with the different types of equipment used in radiographic imaging. The ability of the cables to handle this complex trafficking is part of the responsibilities of the PACS team. If you find this material interesting perhaps you should consider a career in PACS administrating.

Fiber optic channel and fiber-optic cable is necessary to transmit date effectively as duplex transfers if a SAN storage network is used for PACS. Fiber-optic lines can be single mode or multimode.

The other type of cable we should look at is the Coaxial cable, a type of cable that has been around for a long time and is commonly used to connect the household television to cable TV or to a satellite dish, etc. Coaxial cable has a central copper core surrounded by insulation made of PVC or Teflon. Covering the insulation is a layer of braiding, which is another layer of shielding from extrinsic signals. Coaxial cable is the foundation for Ethernet networks designed in the 1980s. It is still used to connect some older televisions, record players, and entertainment systems today.

The coaxial cable was the foundation for Ethernet networks of the early 1980s. Today there is fast Ethernet, thinwire Ethernet, thickwire, CATV, Baseband, and many others.

As we can now see, the network cable is a critical decision component of any PACS network. Cabling must be of the proper magnitude to handle the expected traffic and future trafficking concerns. It is equally important that those devices on the network have fast CPU output and input capabilities in order for the traffic to benefit from any cable media selected. So now let’s touch upon the reason we have put some time in understanding cabling, that is these media will interface at several hubs, connections, and switches as data is moved along their conductive paths. Other important considerations the PACS administrator must look at when designing and expanding networks is the total throughput, cost, connectors and switches, noise immunity, the size and scalability of the network’s nodes. Some media is designed to traffic data up to 295 feet between hubs; others will traffic across the Pacific Ocean. Thus the role of network media could not be overlooked in our study of PACS.

Network Switches and Media transfer

We have already stated that a PACS network is a group of specialized imaging computers that are connected together exchanging and storing data. Cables are rated in terms of the amount of data transmitted in units of megabits per second (Mbps). At the core of data routing is the transmission protocols by which devices send and receive, and the type of network (LAN or WAN). In this section we will briefly discuss the relationship of cable to networking. The picture above shows a corner of a PACS hardware room in which the network cabling comes together at the transfer switch apparatus. Ours is a brief discussion on the topic suitable for a general knowledge base for radiographers.

Every device, called a node, on the PACS network must have specific cable connections to the network backbone. The type of cable depends on the trafficking the network handles. The CR reader, CT scanner, MRI scanner, NM cameras, Interventional C-arms, etc., are all connected to the network in order for them to communicate with PACS. The picture below shows the cables coming together and connected to a HUBPORT switch that manages data transfer.

From the picture we see that cables from all digital imaging devices are connected to the PACS network main switching apparatus (white arrow). This point of entry into PACS is like the post office receiving lots of mail, as PACS receives from the various nodes attached on the Bus layout. Here primary imaging devices such as CT, MRI, CR, DR, and text data from IS server and paper scanner document input and the like enter the PACS network. Complex switching occurs here as data files from various nodes are routed to specific recipient address(s). It is here that the true topology of the network is laid out by switches and connections that shape the network architecture.

Full duplex Ethernet is possible using an elaborate system of switches. Those networks that are half duplex can only send information in one direction at a time. Using a switch allows multiple data communications simultaneously. Because signals are transferred over the network to all devices, it is important that the network contains a terminator device that ends data transmission once the intended device has received it so it does not echo throughout the system causing system overload and noise.

Section III: Archiving Storage and Retrieval Systems

The storage of image data is an important part of PACS, and even more important is that data is retrievable from archive accurately with each read of the data disk(s). Data must also be protected from disk failure since the loss of electronic data is more often than not, a permanent loss, and cannot be replaced unless it is protected by back-up systems. Furthermore, in most states radiology images are considered part of the patient’s medical record. These files and reports must be maintained for a minimum of 5-7 years for adults and longer for pediatric patients (until their 22nd birthday). PACS vendors offer several options for data back-up. In this section we will discuss a type of data storage called RAID (an acronym for Redundant Array of Independent Disk). We will also look at the use of digital linear tape (DLT) as a means of providing a disaster recovery system for PACS image documents.

The primary purpose of RAID is to distribute digital information within an electronic file across several computer hard drives to protect it from loss should a single drive failure occur. For most RAID arrays there is an added benefit of improved writing and read from the disks. This is because the archive server can write on multiple disks simultaneously as well as retrieve from multiple disks. This improves input and output speed of the server. The basic RAID levels combine three structural and arithmetic functions to improve read/write speed, and improve on the ability of the system to respond to an unexpected drive failure (fault tolerance). By combining mirroring, striping, and parity fault tolerance is raised.

Some institutions layer their protection of electronic data by using additional servers to back-up the data stored on the main PACS servers, using optical disk, and DLT. Other institutions use two or more servers to replicate data but locate one at a remote location outside the institution so that a disaster may not damage both servers. In both of these scenarios the back-up servers are used to access data as a routine part of archiving and retrieval so that their use is not limited to storage and back-up only. But what is important here in addition to having multiple servers is how the drives are arrayed to protect data. Let’s discuss some RAID concepts that afford protection to our image documents. RAID, redundant array of independent disk, refers to the design of multiple hard drives within the server on to which image files are stored. These disks are the locations where images are stored in the PACS network. RAID uses mirroring, striping, and parity to back-up and protect the memory of image documents.

The first and simplest level of RAID is called Level 0. RAID Level 0 uses data striping only which gives no protection to image data should a drive fail. Striping is a process of spreading data across multiple disk drives so that any one drive has only a portion of the data block (figure below). Data bytes of an image file could be distributed on disks in drives 1, 2, 3, 4 … and so forth.

RAID Level 0 - Data striping

Data striping serves only one purpose, data input and output time is reduced because several drives are writing and reading files rather than a single drive. Because the data is spread out over several drives if any one of the drives fail and cannot be restored all of the data is permanently lost unless it is backed-up on another server. Disk striping alone does not use as much disk space as drives with other arrays, and therefore cost less.

The next level of protection is RAID Level 1 called mirroring. We have already discussed disk mirroring in part 1, but briefly, drives are paired in their functions so that data is written on both. Should any drive in the array fail, the data is not permanently lost. The primary advantage of mirroring is that a damaged disk can be replaced and data from its duplicate can be transferred onto the new drive. This is an excellent form of data protection, especially when duplicate servers in the network are used; however, it is not an efficient use of precious disk capacity. If an image document is retrieved from a server that uses a mirroring the speed of retrieval is improved since half of the data can be read from each mirrored drive simultaneously. Speed is increased during data retrieval but is lost during data writing since data must be written onto each of the two drives of a mirrored pair.

RAID Level 1 - Disk is Mirrored

Concepts Parity checking of RAID Memory

RAID Level 3 and Level 5 both involve the use of a concept called parity. Parity is a simple counting process in which all bits of a data file are summed and the total is either an odd or an even number bit. This is called parity because it is checked against the original data during archiving and retrieval to verify the accuracy of data being transferred and stored. The nice thing about parity is that it does provide a level of data protection in that data distribution by striping can be check and can be repaired or replaced should a drive fail. Data is simply interpolated from parity counts and reasserted.

Remember that all data used in computing and in computerized imaging is ultimately broken down into combinations of ones and zeros. One of the advantages of digital information in the base-2 system is that we can use this to our advantage to check for data errors. Parity is a way of checking for data aberrations and repairing them. Basically parity works by adding up all of the bits in an image and data file, the total of which is either odd or even. The events that follow are complex, but the basic concept is that parity counts can be used to restore data if a disk fails. Now when data is transmitted over the PACS network from the Network Attached Server (NAS), the parity bits of the image data is checked against recorded numbers before transmitting. If the packet does not agree with the original parity bits the data is retransmitted and the first broadcast is said to have failed. Any transmitted data from storage must match the recorded parity counts in order for it to be transmitted.

Disk drives rarely fail; however, when they do there have to be a way to check for performance failure. Parity information is used in RAID arrays to provide this performance check and bestow data protection. Essentially data is striped across multiple drives and that parity is recorded. Should one of the drives in the array fail, a data check from the surviving data will not register the same parity throughout. In addition, the parity will tell us whether each missing bit was a zero or a one. When the disk is replaced missing bits and can be reconstructed.

RAID Level 3

Data striping and parity

The highest level of RAID data protection is RAID Level 5. Since parity is a tool used to protect data from drive failure and striping is a method of increasing I/O speed, Level 5 RAID combines striping and parity to achieve maximum data protection and involve several drives in the I/O process.

RAID Level 5

RAID Level 5 stripes data across multiple disks to improve I/O speed and uses parity to protect the data. In some institutions two servers with RAID Level 5 protection is used to back-up data and to get the benefit of two servers from which information can be outputted. This is a very expensive use of server space, but does provide an extreme amount of data protection.

As you can see storage concerns with digital and electronic imaging are quite different than with conventional hardcopy images. Medical institutions cannot afford to loose the information stored on a PACS server. RAID systems provide an excellent level of protection, and with back-up server, optical jukebox long term storage, and Digital Linear tape disaster storage PACS systems are completely secure. It should be understood that storage on DLT libraries and OJD as back-up is automated by PACS software; however, full back-up of data is performed weekly or monthly by the PACS team.

Section IV: The cost effectiveness of PACS

There are many articles written on the cost effectiveness of PACS; however, it stands to reason that the main reason so many institutions are slow to replace their current radiography systems with digital imaging is the cost. Conventional wisdom and experience has determined that the economics of PACS is that it is not a system designed to be a moneymaker as far as radiology science goes; however, the economics possible for having PACS is a production of savings and service to the medical community. According to the Journal of Digital Imaging, the average revenue for PACS sites was $38.2 million in 1999 and $13.5 million for non-PACS sites the same year. When we look at equipment savings, FTE reduction, efficiency of the workflow, and reduced cost of repeat films and printing, and the list goes on, there emerges a picture that over time the cost effectiveness is there.

As we look a little closer at the cost of digital imaging and PACS we see that staff efficiency is the main reason for the increased productivity across the board. The "soft copy" images are moved with greater efficiency than radiographic film so the radiologist time is not spent in flipping through image jackets and rolodex frames to read studies. The throughput is excellent and any physician who becomes comfortable with soft copy images will reluctantly go back to hard film imaging. One hospital's emergency room boast that before a patient can be transported back to the emergency room from the radiology department they have already seen the radiographs and begin planning treatment. According to the same ER department spokesperson, patients rave about the efficiency of the emergency room which has largely been because of the efficiency of radiographic imaging through PACS.

According to several technologists at Regions Hospital, Saint Paul, Minnesota (A Level 1 Trauma Center), they most appreciate not having to leave the patient to process or hang films. Images are available on the CRT seconds after being placed in the CR reader. Even the most difficult trauma patient is completed in less than 15 minutes. The ER trauma team is more than willing to travel to the radiology department with even a vented patient since imaging time is always less than 15 minutes. The ER Trauma physicians see these images immediately; so treatment planning begins before the patient's return to the ER.

As you can see cost effectiveness is intimately tied to effectiveness and productivity. In the world of cost effectiveness this is a factor that has little weight in capital purchase validation; however, any PACS radiologist can attest that their job in terms of volume would be impossible with the current radiologist shortage without digital imaging and PACS.

• The shape of a network as nodes are added to it is called its topology.
• There are 4 types of network topologies: Ring, Star, Bus, Tree; of these the Bus is the most common type for PACS.
• Before the PACS server performs any action on acquired images it must receive text information from the RIS/HIS server and create or add to an existing patient file.
• Data documents are sent from the server to workstations, NAS, DLT, and to the Matrix server if one is attached to the network.
• Advantages of a Matrix server is that image data stored on the main server or in NAS is also stored on the Matrix server so it can be accessed by the diagnostic workstations without competing with the main server for speed.
• Data is pre-fetched to the main server and to the matrix server so that there is no delay in reviewing prior studies stored in PACS.
• If PACS is global, that is distributed through the internet; a separate internet web server is required. This is because the web uses a different protocol called TCP/IP. This protocol allows data packets to be opened in the order it was sent since different routes for data may be taken along the internet.
• The term network media describes the cables used to connect the various nodes. The media is the networks backbone.
• Data transmission is measured in megabits per second (Mbps).
• Data transmission is related to the frequency and noise in the system. Frequency is expressed in Hertz.
• Analog signals are pulsed continuously having varying amplitude. These signals are degraded by system noise and amplified so that the final signal reflects the noise.
• Digital signals are pulsed discontinuously as discrete values of "0" or "1." Digital signals are affected by noise; however, a repeater is a device placed in the path of the signal that reproduces the "0" or "1" exactly eliminating noise artifacts.
• There are two important considerations in selecting the type of media for PACS transmission. The diameter and length of the media, which will determine the speed at which information is transferred. The type of cable such as coaxial vs. fiber optic lines and fiber optic channel that will determine the distance over which the data can travel without degradation.
• Crosstalk is the phenomenon of electrical or magnetic interference from other lines in a cable disturbing adjacent cables. Alien crosstalk is a term for different cables interfering with each other through magnetic force.
• Twisting and shielding of cable lines help reduce crosstalk.
• The speed at which a system responds to an unexpected drive failure is fault tolerance.
• Mirroring is a storage drive set-up in which exact copies of data are stored on adjacent mirror drives. If one drive goes out it can be replaced and the mirrored disk copied onto the new drive.
• Striping is a process of spreading data across multiple drives.
• Parity is a process of adding “bits” so that data can be checked for accuracy during the storage and retrieval process.
• RAID is a term for redundant array of independent disks. It is a term that specifies how data is stored on disks to protect it from loss. Raid level 5 is the most protection currently available to PACS. It consists of striping and parity design. There is duplex data transfer and added input/output speed inherent in RAID Level 5 design.

References

Copyright

2006 Nicholas Joseph Jr.