PACS archive evolves to meet enterprise needs

By: Edward M. Smith, SC.D.

A PACS was originally considered to be a radiology resource that stored only radiologic studies. Today it is recognized as an asset to the entire healthcare enterprise (HE), and the archive (archive manager and storage system) must eventually become a repository for all images, waveforms, structured reports, and data from all departments. A PACS must be seamlessly integrated, with bidirectional communication to each department's information system and the ability to access images and other data from within the electronic medical record. These data must be simultaneously available anywhere within the HE at any time.

The archive has evolved in the last three years from a complex multitier storage system using relatively unsophisticated rules to fetch prior studies to the current, less complicated architecture. On-demand study distribution that provides image retrieval times of three seconds or less is now possible, using Web technology.1 Uptime guarantees have risen from 99% or less to near or better than 99.99% fault-tolerant systems. The older systems required substantial staffing resources to manage the archive and provided primitive disaster recovery capabilities compared with today's automated disaster recovery systems.

An institution should ensure that it has met the following prerequisites prior to conducting detailed negotiations with PACS vendors:

• establish a PACS committee;
• implement an education program for the staff;
• develop purchasing standards for the infrastructure components and all future clinical databases and modalities to ensure that they satisfy integration and interoperability requirements;
• develop operational hospital and radiology information systems or plan to purchase an integrated RIS/PACS;
• become familiar with the PACS offering of the vendors; and
• prepare a functional request for proposal tailored to the institution's current and future needs.

Before signing a contract to purchase PACS, the institution, with the assistance of the vendor, should evaluate the infrastructure (local and wide area networks), DICOM compliance of modalities,2,3 and HIS/RIS/PACS integration.

The HE infrastructure must have adequate and scalable bandwidth based on current and projected requirements and sufficient robustness to ensure 99.99% to 99.999% uptime. PACS, with its larger data packets, requires greater bandwidth than a HIS or RIS. The LAN for PACS must be 100 Mbps switched with a 1-Gbps backbone. The bandwidth required for the WAN will depend on the HE environment and archive configuration. Documentation must be in place for all electronic closets, cable runs, and network drops.

To minimize the cost of PACS and to maximize its productivity and economic gains, the HE should establish purchasing ground rules for any device that will interoperate with the archive and PACS, including modalities, clinical databases, storage, printing and

display, and transcription or speech recognition. Where appropriate, these systems must support HL7,4 the necessary DICOM service classes, and other standards. The vendor should ideally use off-the-shelf (nonproprietary) components that are independent of the operating system.

In addition, the vendor should participate in Integrating the Healthcare Enterprise (IHE),5,6 support IHE integration profiles, and agree to work with the HE on satisfying Health Insurance Portability and Accountability Act (HIPAA) requirements. PACS and database vendors must specify other vendors' products with which they have integrated and advise whether they can provide bidirectional communications.

ARCHIVE IMPLEMENTATION TIMELINE

As soon as possible after all prerequisites for PACS implementation have been met and a purchase order signed, acquisition and storage of all studies from DICOM-compliant modalities, especially CT and MR, should begin. This can be achieved by either installing the archive first or outsourcing the studies until the archive is installed. The latter approach can reduce the initial capital outlay if the PACS is not purchased on a fee-per-use basis. A six to 12-month accumulation of studies is suggested. When the archive is installed, the outsourced studies can be downloaded to the permanent archive or they can remain outsourced if there is adequate short-term storage onsite and sufficient bandwidth to retrieve them from the offsite storage facility.

Multiple reasons exist for initiating study acquisition first. It is very difficult for a radiologist to read current studies in soft-copy format and prior studies from film. Most CT and MR studies are read in stack mode, while filmed studies are displayed in array mode. Archiving studies prior to routine soft-copy reading facilitates the transition to filmless for the following reasons:

• radiologists will find it easier to compare current and prior studies because the studies will be in the same display format;
• studies can be available at any time throughout the HE;
• film can be eliminated sooner; and
• productivity and cost savings available as a result of PACS can be achieved earlier.

STORAGE MEDIA

Media available for data storage include magnetic disk (hard drives) magnetic optical disks (MOD), magnetic tape, and various formats of CDs and DVDs.7-12

Enterprise storage systems typically use magnetic disk, MOD, or tape for data storage. The accompanying table compares hard drives with tape storage. Whenever you purchase media, buy only what is required for 12 to 18 months because the storage capacity will increase and data acquisition times and cost per GB tend to decrease every 18 months or so.

The cost of magnetic disk storage has decreased substantially and will continue to do so. Today, disk capacities range from 18 GB with rotational speeds of 15,000 rpm to 181 GB at 7200 rpm. By the end of this year or early next year, disks that can store around 400 GB will be available from many vendors. The cost of the media must be combined with the cost of the storage system in which it is contained and managed. Much of this cost can be related to the storage management software, the hardware system that is used to write and read data to the media, and associated infrastructure, which may be direct-attached storage, network-attached storage, or storage-attached network.13

To improve reliability, disk storage is usually configured as a redundant array of inexpensive disks for most PACS implementations.14 The two most frequently used RAID formats are RAID 1 and RAID 5. In RAID 1, data are mirrored on a second set of disks, providing 100% redundancy. If one disk fails, the other set can take over. This is an inefficient use of disk space, but it provides high availability for applications that require it, such as a demographic database. In RAID 5, if one disk fails, data on the failed disk can be rebuilt on the "hot" spare that provides the redundancy and fault tolerance.

The actual storage capacity in all RAID formats is less than the native disk storage capacity. When purchasing RAID storage, require the vendor to specify the uncompressed native storage capacity and uncompressed storage capacity, taking into account the overhead required by RAID and the capacity loss resulting from the required high water mark, which is typically 90% of the available storage. When the disk system is filled to 90% capacity, studies will be deleted from the disk system.

Among the many tape storage systems available today are Quantum DLT (this is usually used for backups and is not recommended for use in retrieval from long-term storage), StorageTek 9840, Sony AIT-3, and IBM LTO.

The characteristics of the tape drive, media, and storage management software determine the time to access, write, and read data to tape. Data management software, called hierarchical storage management, is required if different media are used in the HE storage system. The rate-limiting factor for most tape systems is the required GB of data to be read and written to tape for a specified period of time. The vendor must provide these data to meet the HE procedure volume and productivity requirements. The system has to be configured so that ad hoc fetches or fetches for emergent studies can be performed within a minute or two of the request.

STORAGE REQUIREMENTS

The PACS archive is used to store radiology studies, but institutions are increasingly using it to store cardiology studies, waveforms, and structured reports as well. In the near future, images from other medical specialties such as ophthalmology, dermatology, and gastroenterology will be stored as DICOM Part 10 images. The vendor must list all the DICOM storage service classes supported, provide new storage service classes as they become available, and work with the modality vendor to ensure both connectivity and interoperability.

The archive server and storage system should be sized for both number of studies performed and the associated storage requirements for these studies.1 Estimating typical storage requirements for a modality, let alone the radiology department or the HE, is becoming increasingly difficult. That these requirements will increase with time is a certainty, making it mandatory that the storage system be scalable and that the database supporting storage be expandable and flexible.

Multislice CT is an example of increasing storage requirements; studies comprise hundreds of slices and some protocols use 1024 x 1024 resolution rather than 512 x 512. In MRI, 3T magnets make it feasible to increase the image matrix from 256 x 256 to 512 x 512. Both of these advances increase the storage requirements four-fold. Additional MR applications such as functional MR and MR angiography and an increased number of sequences per study add considerably to the amount of storage required. Computed radiography, direct radiography, digital mammography, and angiography, with studies that can exceed 400 MB, will continue to expand the storage requirements of the archive.

ARCHITECTURE AND ATTRIBUTES

The archive (the storage component) and the archive manager or equivalent (the control center) interact with all components of the PACS. As the most critical component of a PACS, the archive must be highly available, or fault-tolerant. All studies will transit through and eventually reside on it. The archive has evolved from a complex, multitiered, multimedia system using study servers for study distribution to a distributed, simplified, highly available (99.99% uptime) storage and study distribution system that incorporates redundancy and disaster recovery (see figure). Many vendors use a single database and storage system to support diagnostic, intranet, teleradiology, and Web distribution of studies. At least one vendor has associated URLs with the DICOM objects to Web-enable the archive.

The required attributes and functionality of the archive and archive manager are evolving as PACS matures from being a desirable capability in radiology to an absolute necessity in both radiology and the HE. Many of the traditional attributes are still necessary, including the following:

• connectivity and interoperability with all other PACS components;
• DICOM functionality with other PACS components and modalities;
• seamless integration with the RIS;
• virtual image database with global query and retrieve (patient data or study can be retrieved from any workstation without knowledge of where that information is stored);
• alterable archive manager workflow to meet the HE's changing needs;
• scalable archive manager and storage system with respect to the number of studies performed (storage system must be media-independent);
• administrative functionality (monitoring system resources, audit trails with reporting capability, and proactive system monitoring, etc.);
• data migration from one medium to another when the media or read/write mechanisms are no longer available or actively supported; and
• redundancy and fault tolerance to provide 99.99% uptime with a maximum downtime of 15 minutes or less.

The archive manager has traditionally managed data flow, databases, and storage of data, as well as administrative functions. Data or study distribution has been "rules-based," using various algorithms for routing, fetching, and migrating studies from one storage system to another. The trend is toward on-demand distribution whereby the user queries current and prior studies as needed and the first image of a study is delivered in less than three seconds. Some vendors are introducing the concept of "intelligent" fetching of studies based on data in the DICOM header of the current study being requested.

The archive manager previously handled the demographic, image, and hierarchical storage management (HSM) databases. The current trend is for the HSM to be part of the storage management software. The demographic database contains the DICOM header information and is used to rapidly locate patient and study data. Each file in this database is typically 10 KB or less and is based on a commercial application such as Oracle or Sybase. The image database contains the pixel data and all or a portion of the DICOM header for an image; each file can exceed 10 MB for direct radiography images.

HSM or the equivalent application is required for all multimedia storage systems and for some all-spinning-media storage solutions. The HSM database used to be managed by the archive manager, but it is now part of the storage management software in many implementations.

A disaster recovery solution will be mandatory once HIPAA is in effect. The two approaches to disaster recovery are restoration of service and continuation of business. Restoration of service can be as simple as having backup copies of all data located in a secure offsite location, which allows the system to be rebuilt, even if it takes a month or more. The continuation of business model is far more expensive, as a concurrent copy of all data is maintained on a system that can take over the functionality of the primary data center if it is compromised. The advantage is that the system can be operational in a matter of seconds to minutes once a failure is detected at the primary data center.

FUTURE OF PACS AND HE ARCHIVE

PACS is evolving into a merged RIS/PACS for radiology. This trend will probably continue as cardiology and other specialties implement electronic information management.

As the HE archive evolves, it will store DICOM objects and data from all the enterprise's information systems, including radiology, laboratory, cardiology, billing, and other administrative applications. Rather than each information system operating on a separate server with its own storage system, a single storage system will consolidate all of them. The applications and operating systems will be maintained on the storage system. The servers for each application will process incoming requests and store or retrieve the required data from the storage system. Storage management software will manage the stored data, make backups as required, secure each application's data for the intended users, and facilitate disaster recovery and other functionality.

The cost for this type of storage system is decreasing and the benefits to the HE are numerous: reduced cost of information management, increased availability of data, reduced personnel cost, data mining, and integration of the HE information systems.

DR. SMITH is a professor of radiology and technical director of nuclear medicine at the University of Rochester in Rochester, NY.

References