Iterative reconstruction techniques cut CT dose

May 22, 2009

Software that improves image quality is on track to be a champion of CT dose reduction now and in the future.

Software that improves image quality is on track to be a champion of CT dose reduction now and in the future.

Adaptive statistical iterative reconstruction (ASIR), a reconstruction engine that sweeps noise from CT images, allows radiologists to cut dose by up to 66% in abdominal scans with no change in spatial or temporal resolution, according to Dr. Amy K. Hara, associate professor of abdominal imaging at the Mayo Clinic in Arizona.

And this software is only the beginning. A more complex iterative technique, called model-based iterative reconstruction (MBIR), may lead to further reductions or better image quality in the future, while dual-energy CT is showing promise in cutting dos, when imaging patient with renal stones.

In presentations at the International Symposium of Multidetector CT in San Francisco, Hara described ASIR as an especially important tool, when imaging patients such as those with Crohn’s Disease whose management requires frequent CT scans over a long time.

Using ASIR to process routine body CT scans, the Mayo Clinic has achieved dose reductions between 22% and 66%, depending on patient body type and application.

This reconstruction engine, developed by GE Healthcare and is available exclusively on the company’s CT scanner, takes no special training and requires no more time to do.

“It’s completely integrated with workflow,” Hara said. “The only thing that changes in the protocol is to reconstruct with ASIR rather than filtered back projection (FBP).”

ASIR-generated images are comparable to ones reconstructed at higher dose with FBP, the conventional means for improving signal to noise, she said.

There are some possible drawbacks from its use, however. One happens when doing liver scans.  ASIR can produce a jagged edge around the reconstructed liver, when data are acquired at too low a dose, Hara noted. Another problem occurs in thin patients when dose is cut too much. This reduces image quality. Bumping the dosage up a bit, while remaining well below the typical exposure, solves the problem, she said.

The third drawback from ASIR happens because the reconstruction engine actually does too good a job of cleaning the image. The algorithms create images with a seemingly artificial texture.

“Your gut reaction is that it doesn’t look like CT examinations are expected to look,” she said.

Solving this problem may be as simple as getting used to the look of the images. “After doing ASIR for a year, you get used to a new normal,” Hara said.

As good as ASIR is today, even greater dose reductions may be in the offing with a processing package that Hara calls “a kind of super ASIR.” This iterative reconstruction technique, MBIR, has produced encouraging results in preliminary tests. It will probably replace ASIR someday as the reconstruction engine of choice, she said, but that day is still a ways off. Because the algorithm is computationally intensive, computers must get faster or its operation must get simpler before entering mainstream practice.

Meanwhile, dual-energy CT is showing the potential for cutting dose by eliminating the need for pre-contrast images.  Research on dual-source CTs built by Siemens, which use two x-ray imaging chains, have suggested that such gains might be possible. The work reported by Hara, however, rely on a single source that switches between two energies, a capability being built into GE scanners.

Energy subtraction methods can produce “virtual precontrast”, she said, when imaging patients for suspected renal stones.   These patients are scanned only after the intravenous injection of contrast. Affording lesser radiation exposure, yet having no impact on the ability to detect stones.  Hara reported being able to see stones between 1mm and 2mm in size.