From the first appearance of multislice CT, radiology pundits knew that data volume would be the issue of the future. Six years of quad, eight-, and 16-slice scanners have proven them correct, generating up to-and in some cases beyond-a thousand slices
From the first appearance of multislice CT, radiology pundits knew that data volume would be the issue of the future. Six years of quad, eight-, and 16-slice scanners have proven them correct, generating up to-and in some cases beyond-a thousand slices per study, way more than can be studied efficiently. One possible solution to this data overload may have been found in India.
Engineers there have developed a 3D multiresolution scheme designed specifically to speed the interpretation and management of large data sets. The technique provides multiresolution-based processing of volumetric data sets, processing that solves the problems encountered when looking at these data as discrete 2D slices, according to Sudipta Mukhopadhyay, Ph.D., lead engineer at Global Imaging Technology Laboratories at the GE India Technology Center in Bangalore.
Underlying the technique is the use of algorithms that allow any individual image to be viewed at different (lower) resolutions. In its simplest form, an image comprising 5122 pixels might be viewed at 2562, 1282, or 642 resolution.
When viewing volumetric images, the technique offers differing resolutions by scaling back data in the axial plane. If the user starts with a block of eight frames, using multiresolution in the axial direction provides eight-, four-, two-, and one-frame views that represent the more expansive block of data.
"With this, we can view all frames in full axial resolution or lower axial resolution, a facility absent in 2D schemes," Mukhopadhyay said.
The key to this technique is its ease of use and flexibility. All resolutions in GE's 3D multiresolution scheme operate off the same bitstream of data, essentially offering the option of choosing different levels of resolution to match different purposes. The volume of data might be examined quickly at lower resolution for pathologies, then reassessed at higher resolution at points representing specific anatomical locations. Alternatively, the data volume might be set for reconstruction at different resolutions, each one appropriate for a different procedure.
This is accomplished by setting the scanner to image the patient at its highest possible resolution, thereby providing a data set with the optimal flexibility. Rather than overloading the radiologist with slices, however, resolution can be set to meet the specific needs of the moment.
Without this technique, radiologists seeking to handle data overload might just set the scanner to obtain thicker slices, which afford lower resolution. This works fine, so long as no questions arise during the interpretation.
"For some procedures, radiologists may pick a 5-mm slice thickness," Mukhopadhyay said. "But then, after the reading, if they want a higher resolution image, they have to rescan, meaning added delay and increased cost."
In the method described by Mukhopadhyay, scanners acquire data at the highest possible resolution, relying on the 3D scheme to provide images at the same or lower axial resolution. This provides the best of both worlds.
"Radiologists can choose (to read at) a lower resolution, and then if they find something interesting, they can dynamically switch to higher resolution," he said. "This allows them to choose the slice thickness they feel is appropriate, thereby saving time."
The significance of being able to do this will increase in coming years. The looming generation of CT scanners offering 32, 40, and 64 slices has raised alarms among radiologists, who fear that an already difficult situation will be made unmanageable by the thousands of slices per exam these new systems could generate. Only by managing slices more efficiently can the radiology community reap the true benefits of this more powerful CT technology.