Low-dose CT Techniques & Applications

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
ReadingRoom-HI-MayJune2014.jpg - Chest CT
Chest CT of a 77-year-old man (BMI: 28 kg/m2) with acceptable demonstration of lungs on SpS-SubmSv IRT image. No streak artifacts in the image datasets at 92% dose reduction.
Source: Ranish Deedar Ali Khawaja, MD


CT is the workhorse of imaging, but as utilization surged in the new millennium, so did concerns about radiation dose. Work on refining reconstruction and acquisition techniques and driving doses down started a number of years ago but recent refinements are pushing dose lower than ever. Meanwhile, low-dose CT’s role in screening has taken center stage.

The rapid expansion of CT was propelled by the fact it had become a critical component of quality healthcare, says William P. Shuman, MD, vice chair and medical director of the Department of Radiology at University of Washington Medical Center in Seattle. “CT is very valuable in the process of getting accurate diagnosis and without good diagnosis, there is no good therapy.”

Still, concerns about dose both within medical literature and in the popular media have propelled efforts to advance low-dose CT technique. While the risks of medical radiation are debatable—and miniscule compared with the risks of not receiving a scan for most indications—patient safety is always paramount. There used to be a trade-off between image noise and dose, Shuman notes. Now, the noise index can be increased to lower dose, and advanced reconstruction can help clear the noise.

“We end up with the best of both worlds—extremely low dose and good images,” he says.

For decades, CT images were reconstructed from raw data using filtered back projection (FBP). FBP algorithms were limited by assumptions that simplified CT geometry to balance reconstruction speed and image noise, but more complex techniques, such as adaptive statistical iterative reconstruction (ASIR) were used to reduce noise while preserving resolution. There is a minimal trade-off in terms of speed, with raw datasets reconstructed with FBP at 15 images per second vs. 10 images per second with ASIR. This is a minor concern as it still takes longer to move the images to a workstation for viewing than it does to convert raw data to images using simple iterative reconstruction.

The next generation of reconstruction is a still more complex algorithm, model-based iterative reconstruction (MBIR), which uses detailed models of radiation and CT equipment. MBIR is much more demanding, requiring multiple computers working for 30 minutes or more to create a complete scan. But the result is quality images at radiation doses as much as 80 percent lower than standard reconstruction. Shuman says he thought physicians in areas like the emergency department, where urgency is king, might not be interested because of the delay. Not so, they are happy to wait in some cases, such as with pediatric patients who are more sensitive to radiation.

Shuman and colleagues compared FBP, ASIR and MBIR in routine-dose clinical CTs of the liver in a study published in the May 2013 issue of the American Journal of Roentgenology. A total of 36 patients with known liver lesions were scanned, and randomized images were presented to a pair of independent blinded reviewers who categorized the appearance of lesions and scored lesion conspicuity.

For the 51 focal lesions in the scans, the reviewers found no statistical difference in lesion detection among the reconstruction algorithms. There was, however, significant difference in background noise. With MBIR, background noise in air was 76 percent and 80 percent lower compared with ASIR and FBP, respectively. Likewise, background noise in subcutaneous fat was 42 percent lower with MBIR compared with ASIR and 54 percent lower with MBIR compared with FBP.

This kind of promising research with MBIR opens the door to still further reduce dose.

Dose falling through the gaps

In addition to complex reconstruction algorithms, other techniques can be used to reduce CT. One technique, called sparse sampling, was on display at the 2013 annual meeting of the Radiological Society of North America. There, Ranish Deedar Ali Khawaja, MD, a radiology research fellow at Massachusetts General Hospital (MGH) and Harvard Medical School, presented work by himself and colleagues, which was conducted as a collaboration between MGH and Philips Healthcare. Sparse sampling is a data sampling technique that can generate quality CT lung screening scans at doses as low as 0.25 mSv. For comparison, the average effective dose of a typical standard-dose chest CT is around 7 mSv, while the low-dose scans used in the National Lung Screening Trial (NLST) were close to 2 mSv.

The basic idea behind sparse sampling is that noncontinuous data could be acquired