Adding Intelligence to Archiving of Data, Images

From the May 2005 Issue

New enterprisewide management platforms support multisite, multisource storage.

By Ken Rosenfeld

Ken Rosenfeld is the worldwide general manager, PACS and Enterprise Archiving Software Business Div., Kodak’s Health Group, Rochester, N.Y.

Digital imaging offers impressive benefits for both patients and healthcare providers, but it also creates the need for intelligent tools to manage the industry’s exponential growth in data and image storage. Management issues are complicated by the decentralization of information and the use of departmental information management systems from multiple vendors.

Healthcare providers need an efficient way to manage these disparate storage systems while simultaneously providing efficient lifecycle management and meeting disaster recovery requirements specified by HIPAA regulations.

Standards Form a Sound Foundation
Newer generation enterprise archives, configured as network-attached systems and supporting a variety of standard interfaces and protocols, assist enterprisewide storage management. DICOM-enabled environments such as radiology modalities and PACS can interface to an archive through DICOM protocols, while cardiology servers are better served by more efficient NFS/CIFS protocols. FTP protocols provide an efficient interface for large data sets from PACS and other systems, and HTTP supplies Web access and retrieval. See Figure 1.

Database consistency on the enterprise archive can be maintained by using HL7 interfaces to communicate with departmental and hospital information systems and PACS. SQL synchronization is also an option. In an enterprisewide model, information can be exchanged using several different interfaces. For example, a cardiology server may use a mounted file system interface for archiving data and a DICOM interface to query and retrieve the same data. An image can be ingested via FTP and retrieved through an HTTP interface.

Object-based management is ideal, since this enables medical data coming from various sources to be consolidated into a patient record—and managed as a single object. This enables global functions to be applied to a patient record, in addition to data access, such as keeping a patient record on one piece of media.

Proper indexing is also necessary to support accurate search and retrieval. DICOM information can normally be accessed through a DICOM interface or via NFS. However, when information does not carry metadata, more classical indexing must be used and specific functions must be implemented to support reconciliation. File naming conventions such as XML descriptors can be used to partially solve this problem.

Managing Multiple Sites
Most healthcare facilities involve multiple sites, comprised of hospitals, outpatient clinics and surgical centers, and other remote locations. An enterprisewide approach is needed to assure timely access and distribution of patient information to each site. This requires leveraging existing imaging and information systems while preserving some operational independence at each site.

An optimal approach involves communication among archives rather than a centralized archive. Each archive stores its own objects but provides a way to share information with other archives in the network. See Figure 2.

In multisite archiving, a patient record must be visible and accessible to all or selected sites—even though the information that comprises it may be stored on multiple sites. The transparent retrieval and transfer between sites is a key feature of an enterprise archive. This involves synchronization of each site’s archived data to simulate a global archive. This synchronization must include proper fail-safe mechanisms that make it possible to preserve operational independence of individual sites when connectivity is lost.

Efficient Information Lifecycle Management
Healthcare facilities must retain medical records to satisfy legal obligations, but they can choose to move records to storage formats based on frequency of retrieval and storage costs. Typically, data is initially stored in online RAID archives and then moved to slower devices such as content-addressable storage systems, digital tape or DVD.

Optimal information lifecycle management is achieved when healthcare providers can apply unique “storage plans” for each type of imaging study. For example, one chain of imaging centers stores MR and CT imaging studies for 30 days on RAID, but keeps general radiography studies captured with CR systems for six months on RAID. This allows the imaging center to tailor its storage plan to its frequency of retrieval. Since the company has a much higher rate of retrieval for general imaging studies, it stores them online for a longer period of time. The imaging center chain’s storage plan specifies that all general exams are stored for seven years. After this period, the exams are removed from the storage system (after appropriate administrative logging). All pediatric cases are stored until the patient is 28 years old (21 plus seven years), and mammography cases are stored forever.

The effective use of storage plans enables an institution to optimize its archive infrastructure by enabling the appropriate performance/storage cost trade-offs to be made on a per-data type basis. Achieving this level of specificity requires intelligent management software—that is, the software must be able to read the header information, determine what kind of data it is and take the requested action as predefined in the storage plan that is applicable for the data type.

Information usage, legal requirements or technology availability may change over time; requiring the ability to change the way this information is handled in the global infrastructure. To support this, changes to the lifecycle of an information type should be easy and possible at all times through a storage plan modification. Once the storage plan is modified and applied, all matching data should automatically conform. This will significantly reduce the total cost of ownership by reducing the amount of manual management required by IT staffs.

Data Recovery, Remote Disaster Recovery
Legal obligations also require data recovery, which can be expedited with storage plans that allow copies of data that are linked to the same subject to be stored on various archives and retrieved independently. This implementation requires that software management does not prioritize one storage media over another, but allows flexible recovery mechanisms.

This new breed of storage plan solves problems inherent in previous hierarchical storage management (HSM) systems, which inhibited recovery from a secondary storage device. HSM systems also require complex IT manipulation and disk reconstruction to recover data.

Conformance to HIPAA also dictates that medical information must be available even during catastrophic events. Disaster recovery can be handled by creating backup copies of data on some type of removable media, but healthcare facilities must also be able to reconstruct the data in a timely fashion in the event of a disaster.

Classical IT backup approaches are not equipped to handle the terabytes of data now stored by most healthcare facilities. However, an enterprise archive platform can provide disaster recovery by creating a duplicate database and records at a remote site. In this model, a complete copy of the archive can be incrementally constructed on the remote site, including archive indexes. Recovery is simple and timely, since the remote archive can be queried by network users and used to retrieve data. See Figure 3.

Protection Against Obsolescence
No enterprise archiving management solution is complete unless it can facilitate data migration. The media on which the information is written will eventually become obsolete. We can estimate that storage technologies become obsolete every three years. Other associated IT technologies have an even faster refresh rate. This means that hardware obsolescence will be faced several times in the information’s lifetime. As storage systems become outdated, healthcare providers will want to move data to newer generation platforms that reduce costs and assure reliable access. In addition to the media on which the data is stored, the format in which the information is written may become obsolete, making it unreadable unless it is converted.

Previous generations of information systems often relied upon DICOM-level migration or other slow, manually intensive processes, which have a limited maximum daily throughput. This migration process has been a very expensive and error-prone process. In contrast, an intelligent lifecycle management solution will migrate data to new storage slots, which represent new devices, without the need for DICOM-level or other slow, manual transfers. The migration is performed by making it part of an updated storage plan that indicates the need for a copy of data on the new storage device. This migration can then happen efficiently and quickly in the background. If format changes are needed due to obsolete storage formats, the storage plan can be enhanced by an “execution slot,” which is a piece of software that can automatically perform the appropriate conversion of data.

The current growth in data and storage systems can only be expected to increase. Healthcare providers need to act now to implement intelligent, enterprisewide archiving software that enables efficient multisite and multidepartmental archive management. All current and future storage purchases should support industry-standard protocols and interfaces along with object-based management. A smart enterprise strategy that utilizes individual storage plans will also provide automated lifecycle management and disaster recovery. Together, these capabilities will create a more automated and flexible storage management environment that equips healthcare providers to satisfy a continued growth in data and images.

For more information about PACS and enterprise archiving software from Kodak’s Health Group,
www.rsleads.com/505ht-200