|Katherine P. Andrile, PhD|
Early cost justification analyses for transitioning a traditional film-based medical imaging department to the picture and archiving communications system (PACS) environment identified the high cost and substantial inefficiencies of image archives as a major impediment. Recent trends in archival technology have shown the cost of digital storage media decreasing steadily and the capacity increasing, while analog devices, such as paper and film, continue to increase in cost. Improvements in storage devices, along with the use of intelligent software, have removed digital archives as a major stumbling block to implementing PACS. The use of compression can further the argument that it is now imperative for health care enterprises to function without film.
Clearly, digital image archival can be more efficient than the manual data storage of the traditional film file room. Studies of image examination retrieval from a PACS, as compared with film-based systems, have shown statistically significant reduction in retrieval times for the digital method, in many cases down from several hours to mere minutes or seconds.
Medical image data files are large compared to other text-based clinical data, and are growing in size as new digital applications, such as spiral CT and magnetic resonance angiography (MRA), prove clinically useful. A single-view chest radiograph, for example, requires approximately 10 megabytes (MB) of storage space. A full-body MRA can consist of more than 3,000 slices per study, requiring approximately 400 MB of storage space. Imaging activity continues to increase significantly as it becomes a key diagnostic triage event, with most diagnostic imaging departments showing an increase in overall volume of cases. A busy 250-bed hospital performing approximately 90,000 examinations, for example, can generate several terabytes (TB) of data per year.
Compression can be used to reduce both image transmission time and storage requirements. Note that compression also occurs clinically, for example, when not all images of a study are filmed. Lossless (or bit preserving) compression at 2:1 is done by most PACS archive systems already. Lossy or non-bit preserving compression by definition does not provide an exact bit-for-bit replica of the original image data upon decompression. However, studies have shown that numerically lossy compression can produce visually lossless images at compression ratios of 5:1 to 30:1 depending on modality (Radiology 1998;206:599-607 and J Digit Imaging 2001;14:18-23). Compression at these levels can achieve much greater space savings and appear to be of adequate image quality for image comparison and review of prior studies. Without compression, only 50 two-view digital projection x-ray examinations at approximately 10 MB per image can be stored on a single gigabyte (GB) optical disk. With compression at 25:1, approximately 1,250 examinations can be stored on a single GB disk.
Note that medical images have a life cycle in which, early on, immediate access to the data is critical and images are often needed by several people in many different places simultaneously. After a patient has been treated and discharged, however, that same imaging study may rarely need to be accessed again, and if it is, taking minutes or even hours to retrieve it may be acceptable. Note that most patients who do not return for continuing care within 18 months of the acute visit are unlikely to return at all. In other words, a large percentage of imaging examinations performed will never be re-reviewed after the original clinical episode, so slower access may be acceptable. This pattern of use suggests that hierarchical or staged archival strategies can be implemented for the most cost-effective use of storage technologies.
The requirements for fast retrieval of images initially followed by slower retrieval later, if at all, suggest that different types of storage devices could be used over time to archive images with cost savings. As fast retrieval times grow less important, images could be migrated to less costly, higher capacity, slower storage devices. Software must be used to handle the movement of data from one medium to another, and the strategy should make the actual physical storage device transparent to the end user. Such a strategy is known as a hierarchical storage management (HSM) scheme.
Three basic types of digital storage media (MD-RAID, OD-MOD, and tape) are currently being used in PACS, although new devices such as DVD appear promising. These media vary in their capacity, performance, and cost, and may be cost-effectively used in combination. RAID has traditionally been used for “near line” or intermediate short-term storage to minimize the number of transactions reaching the deep or long-term archive. It is becoming cheap enough per capacity to consider using RAID in larger configurations for high-performance, longer-term storage. This way a higher percentage of studies, perhaps accounting for a year or more, can remain available online for immediate access.
But this is at today’s data acquisition rates-what about tomorrow’s rates? Developments in medical imaging technologies are occurring at such a rapid pace that it is difficult to specify PACS designs today that will still satisfy requirements in the future. Imaging modalities are generating more data per image and more images per study. Many argue that this necessitates transitioning to PACS because these very large imaging examinations can not be viewed on film.? The fact that medical imaging is moving to the digital environment in turn feeds the cycle of development of imaging technologies that create even larger data sets. And so on.
Most PACS vendors today offer hardware-only solutions to unanticipated growth in storage requirements. Buy more archives, add more boxes. That may palliate the symptoms today, but what happens when clinical imaging routinely includes 3-D information, functional imaging, color, true cine, or some other wonderful new technology just below the horizon? Why not consider a software solution such as hierarchical storage management and compression.
A HSM scheme such as that described in the Journal of Digital Imaging 2001 mentioned above may be the smarter solution. In this scheme, the full resolution image data are viewed originally for primary diagnosis, while a losslessly compressed version is sent off-site to an inexpensive tape backup archive. The original data are wavelet lossy compressed and stored on a large RAID device for maximum cost-effective online storage of images for review and comparison. This scheme, which utilizes short-term archival of uncompressed DICOM data for primary diagnosis, in an on-site RAID, coupled with a very deep long-term archive of diagnostic-quality wavelet compressed data in an on-site optical jukebox, cost effectively maximizes online storage for immediate image retrieval. Diagnostically lossy compressed data (at ratios of 25:1 for computed radiography, 10:1 for CT, and 5:1 for magnetic resonance images) grow the on-site jukebox by 10 times, depending on the mix of cases, making 20 or more years available online. This effectively maintains the entire legal record of original plus two relevant prior examinations all online, without any additional hardware boxes.
As you assess the need to purchase a bigger, better archive device to meet your medical imaging needs, consider whether you are getting the most out of your current hardware for today’s storage needs as well as for tomorrow’s. A software solution may just be the storage strategy you need.
Katherine P. Andriole, PhD, is PACS clinical coordinator, and assistant professor of radiology, University of California, San Francisco.