The handling of large datasets will continue to be a fundamental issue for the medical imaging community, and online and offline data are going to cease being separate entities, forecasted Richard Morin, PhD, Brooks Hollern professor, department of radiology, Mayo Clinic-Jacksonville, at the annual meeting of the Society for Imaging Informatics in Medicine (SIIM, formerly SCAR) in Austin, Texas, last week.
“Medical image storage growth continues at a rapid speed,” said Morin during a SIIM TRIP (Transforming the Radiologic Interpretation Process) Initiative presentation. “We must think about the merging of online with offline data. It is critical we know how to do that efficiently.”
When it comes to healthcare storage infrastructures, Morin told audience members that PACS imaging archival has never been well served by the “one-size-fits-all” approach. While the disk-only archival solution is appealing to all parties, there are some inherent flaws in disk-only solutions.
“Another major issue of disk[-only] is that every three years, disk technology goes obsolete,” said Morin, adding that service quotes are difficult to draft and add-on price per megabyte is extremely difficult. In addition to disk, tape is another magnetic technology used in the healthcare realm to store medical images.
The optimal, efficient PACS archive should balance physician needs for data access, budget requirements, long-term data retention, technology non-obsolescence and scalability. Therefore, Morin recommends a “multivariate” solution that combines a mix of disk and tape magnetic technologies managed by appropriate hierarchal storage management software.
To understand better why a multivariate approach is necessary, it is key to understand current data storage trends. Four years ago, the standard CT scan was a 2-second single slice exam. Today, Morin said that the standard is a sub-second 64-slice exam. The average data acquired has gone from 20 MB to over 500 MB. The 64-slice detector CT may be replaced with 1,000-row flat panel amorphous silicon detectors. Cardiac CT will drive sub-second scanning even faster. Three-dimensional ultrasound will cause storage capacity needs to grow from the current 20 MB to 50 MB per study to over one GB. Also, digital mammography will more than double the radiology storage requirements over all other modalities combined.
It is, but should not be, assumed that medical imaging archived data are rarely accessed, Morin said, which is another reason a multivariate approach is appropriate for medical imaging PACS archives. “Ask yourself: what is the probability of accessing data that are two to three years old, not two years old and three days?” Morin said. Classic examples are pediatric, cardiac and radiation oncology cases where old data remains very critical for a long-period of time.
What has been learned about image overload and PACS archival strategies, Morin questioned audience members? “The disk we need is not getting cheaper,” he answered. “Your archival retrieval speed may not have to do with slow media, medical imaging exceeds disk capacity growth and vendor assumptions make us prisoners.
“What do we need,” Morin posed? “Cost-effective, managed storage, consistent data access times, large file system support, efficient database design, many years data migration upgrade and or transition plans and a balanced read/write archival bandwidth.”