Long-term Data Management and Migration for PACS
PACS data migration is so important that every PACS acquisition should include a plan for outbound migration at the end of that system’s life, according to Frederick M. Behlen, PhD. Behlen is an officer, director, and shareholder of LAITEK Inc, also known as Migratek™ Data Migration Services, Homewood, Ill. He serves as cochair of the DICOM–HL7 joint working groups. On May 15, 2008, Behlen presented Policies Surrounding Migration during the PACS Operational Policies and Procedures session of SIIM U, held in Seattle during the Society for Imaging Informatics in Medicine’s annual meeting. The following day, Behlen covered the practical aspects of his topic by presenting Data Migration: Plan Early and Sleep Better. Together, his presentations provide a framework for devising and implementing a solid, cost-effective plan for data migration. Behlen points out that while the retention requirements for patient images and other information vary by state, they have become more serious over time, particularly since the 2005–2006 two-stage implementation of the HIPAA Security Rule. In addition to the need for confidentiality and accountability in handling patient data, standards of practice for data management have risen in recent years. Because electronic records became vital to business operations, uninterrupted access to them became vital, in turn, to business continuity. For this reason, disaster recovery has turned into an indispensable function and long-term data management is a part of strategic planning. Operations planning is needed as well, and should focus on the long-term management of records, including migration of data from one system to another, as required, to address both hardware/software obsolescence and disaster preparedness. Even when information has been moved to a stable storage medium, there must be a working machine at hand to read it. The longevity of the storage medium means nothing if the hardware and software required to extract the data are long gone, so the organization must define effective data-storage life as the shortest time that media, software, and hardware can be expected to survive, no matter which fails first. Although PACS data migration has become somewhat simpler as data standardization via DICOM and HL7 has been adopted more widely and connectivity has improved, Behlen says that migration is never a trivial matter. Ensuring that the PACS stores all DICOM information objects in Part 10 files is helpful, but Part 10 data must still be scanned and then imported into the databases of future systems. Because there may have been changes in examination numbers and patient demographics since the files were stored, the PACS database will need to update them. Being Prepared Future system transitions must be prepared for now, so that the organization will have some options available when the time comes to recover or migrate data from systems that no longer function well (or at all). Behlen emphasizes that for disaster recovery to be feasible, data must be secured now and a long-term plan for data management must be in place. In choosing system architectures, the ease of future transition should always be kept in mind. Migration policies should also be set well before they are needed as part of a data-maintenance protocol. Understanding the nature of the data is the first step in maintaining the archive, which Behlen says is best configured as a long-term archive that is not part of the PACS that handles everyday workflow (a two-tiered system). Letting a service contract for existing systems expire may not be advisable, he adds, and all migration projects should clearly specify where the responsibility lies for maintaining the old archive. Even if radiologists claim that they do not need certain studies that have been lost, it is worthwhile to make the effort to recover them and add them to the archive, since it is impossible to predict what will be useful in the future. Proprietary methods used for image annotation should be avoided because this information could be lost in migration, since it cannot be exported successfully. Instead of using proprietary annotation, Behlen recommends using DICOM Gray Scale Presentation State objects or DICOM overlay planes embedded in images. Many vendors also use proprietary methods to refer to images from within reports, and this can also cause migration problems. Web Access to DICOM Persistent Objects (WADO) reference is preferred, Behlen says, because WADO image references in reports will not be lost in transferring to a later system. The best reporting practice is to ensure that there is no clinical need for any image annotation or any image reference in the report. Behlen acknowledges, however, that this ideal standard cannot always be met if referring physicians expect such references and annotations to be present. Preventing Problems It is also advisable to discard all unnecessary images prior to migration; these could be demonstration images, test images acquired during maintenance or equipment-service procedures, or images that were initially unacceptable due to some type of modality malfunction. In addition, careful maintenance of the demographics and linkages between examinations and studies in the RIS and PACS is needed. For example, a head CT scan is a simple case that involves one examination and one study, but a CT scan of the chest, abdomen, and pelvis is treated as a group case (one study with multiple examinations). What Behlen calls a messy case, such as a CT scan of the chest, abdomen, and pelvis that is stopped for any reason and then restarted, will consist of multiple examinations and multiple studies. The migration process must ensure that cases of all three types are associated with the correct studies and examinations. Behlen says that the best data-management planning for the long term can involve radiology plus other services that use imaging, such as pathology, dermatology, and cardiology, creating an archive as an enterprise asset that will improve the disaster-recovery capabilities of today’s PACS. At the same time, this asset (perhaps in the form of a two-tiered PACS) will be ready for data migration to a future PACS because it provides storage that is independent of vendor considerations. Today, two-tiered PACS are used by many organizations to perform both everyday and archival data management. Nonetheless, Behlen says, the advantages of such an arrangement become fully available only if the interface that connects the long-term archive and the server conforms completely to the DICOM Query/Retrieve and Storage class standards.