The term archive can be defined as a place where records or documents are preserved (Figure 10-1). In a picture archiving and communication system (PACS), the electronic archive serves as the new file room and warehouse for all digital imaging and communications in medicine (DICOM) imaging modalities (Figure 10-2). The PACS archive stores all patient and image data, often on magnetic tape or optical disk. Also, the PACS archive controls the receipt, storage, and distribution of new and historic images. Because of the explosive growth in the use of digital imaging in radiology, the archive is one of the fastest growing components in the PACS. Archive technology continues to make drastic improvements each year; the storage capacity is said to double every 18 to 24 months, and the price per gigabyte also continues to decrease.

FIGURE 10-1 The archives at the National Archives in Washington, DC.

FIGURE 10-2 Examples of PACS archive storage: Magneto-optical disk (MOD) and jukebox. (Courtesy Sun Microsystems.)

The archive is a complex arrangement of computers and storage space. As a whole, it consists of several components, both hardware and software. These can be divided into two major categories: image manager or controller and image storage or archive server. The next two sections discuss image management and image storage. Various types of image storage hardware are described. The chapter ends with a discussion of things to consider when choosing an archiving system.

Image Manager

The image manager contains the master database of everything that is in the archive. It controls the receipt, retrieval, and distribution of the images it stores and also controls all the DICOM processes running within the archive.

The image manager generally runs a reliable commercial database such as Sybase (Sybase Inc., Dublin, CA) or Oracle (Oracle Corp., Redwood Shores, CA) with structured query language (SQL). This database contains only the image header information, not the image data. The image data are stored on the archive server, which is discussed in the next section. The database is mirrored, meaning that there are two identical databases running simultaneously so that if one goes down, the system can call on the mirror and continue to run as normal, a very important feature.

The image manager is also the PACS component that interfaces with the radiology information system (RIS) and the hospital information system (HIS). This allows the PACS database to collect additional patient information that is necessary for its effective operation. Information extracted from these databases will be used in the prefetching and routing of images to various locations throughout the PACS. The image manager can also play a key role in populating image information into the hospital electronic medical record (EMR).

As mentioned earlier, the image manager database contains the DICOM header information, such as the patient name, identification information (ID), examination date, ordering physician, and location. These fields are organized within the database so that when someone queries for a study on a workstation, the image manager can quickly move through these data fields and locate the images that are being queried (Figure 10-3). The database has pointers associated with each image on the archive server that point back to the data fields within the database. The following list summarizes the process:

• An order is placed in the RIS for a radiology study.

• The images are acquired and sent to the archive.

• The image manager strips the image header from each image and assigns a pointer to each image or series of images.

• The database files the information in various fields and communicates back to the RIS to verify certain information.

• The study is then queried, and the pointers locate the images on the archive server and send the images to the workstation.

FIGURE 10-3 The PACS database can be queried using various data points.

Image Storage

The image storage or archive server consists of the physical storage device of the archive system. It commonly consists of two or three tiers of storage. A tier is a level, layer, or division of something. In an archive server, a tier represents a specific level of archive: short-term, mid-term, or long-term. Most PACS archive systems are set up with a short-term tier and a long-term tier. Short-term means being online or available very quickly, usually within 3 to 5 seconds. Long-term means near line, or images that must be retrieved from a tape or disk storage device and brought to a redundant array of independent (inexpensive) disks (RAID). This could take 1 to 5 minutes.

Short-Term Storage.

The short-term tier is commonly a RAID (Figure 10-4). A RAID is composed of several magnetic disks or hard drives that are linked together in an array (Figure 10-5). The size of the RAID ranges from several hundred gigabytes to several terabytes. As the individual disk sizes continue to increase, so does the potential size of the RAID.

FIGURE 10-4 A redundant array of independent disks (RAID). (Courtesy Sun Microsystems.)

FIGURE 10-5 A RAID array. (Courtesy Sun Microsystems.)

In 1988 David Patterson, Garth Gibson, and Randy Katz coined the term RAID in an article entitled “A Case for Redundant Arrays of Inexpensive Disks (RAID).” Their presentation introduced 5 levels of RAID; now there are 11 levels (Figure 10-6), most of which are combinations of the first 5. The following discussion, which includes diagrams and a short synopsis for each of the RAID levels 0, 1, 3, and 5, illustrates how the RAID levels differ from one another:

• RAID 0: Data are “striped” across all of the connected disks. Striping means that the data are broken up into pieces, and each disk will have one piece of the data (Figure 10-7). When the data are called up from the RAID, all of the data are put together from the disks and presented to the user as a whole.

• RAID 1: All of the data sent to the RAID are mirrored onto two disks (Figure 10-8). Mirroring means that all of the data are duplicated and placed onto two separate disks. This RAID level has full redundancy, meaning that if one disk goes down, the other one takes over and operation of the system continues. This is a very expensive system because only half of the total storage is used.

• RAID 3: The data are striped across all of the disks just like in RAID 0, but there is one disk that is set aside for error correction. This disk is known as the parity disk (Figure 10-9). This RAID level is rarely used.

• RAID 5: This RAID level is similar to RAID 3 but instead of having the parity written to one disk, it is striped along all of the disks within the RAID (Figure 10-10). RAID 5 is the most common level used for a PACS archive because it provides adequate redundancy and fault tolerance.

The striping of data increases the reliability and performance of the system. With certain levels of RAID, if one disk fails, the data from that disk can be regenerated using the redundancy of data on the other disks. The error correction detects any transmission errors, and the data will also be regenerated based on the information from the other disks. Striping also enhances performance because if all of the data were on one disk, data added to the disk first would be accessed first, requiring longer wait times for data added to the disk later. Spreading data over several disks allows all data to be accessed at the same time.