Knowledge-based iterative reconstruction enables ultra-low current CT colonography

In a phantom study, reducing tube voltage for CT colonography to 80 kVp cut median effective radiation dose to a level lower than a typical abdominal radiograph, but also resulted in severe deterioration in image noise and per-polyp sensitivity unless paired with a full knowledge-based iterative reconstruction (IR) algorithm, according to findings published July 11 in Radiology.

“Despite the intrinsic high contrast between colonic air and adjacent soft tissue on CT colonographic images, radiologists’ performance on CT colonography at 80 kVp was not acceptable with FBP (filtered back projection) and hybrid IR algorithms,” wrote Cheong-Il Shin, MD, of Seoul National University Hospital, and colleagues.

Since dose is a concern in CT colonography, and reducing tube current is the simplest method to effectively reduce radiation dose, Shin and colleagues sought to examine the effect of ultra-low tube current on image noise.

Full knowledge-based IR, the next generation of IR algorithms, has been developed to reduce noise in a model-based process. “Unlike previous IR algorithms, it does not involve blending with FBP images, and it is mathematically more complex but also more accurate,” wrote the authors. “The knowledge-based approach accurately determines data, image statistics, and system models, which depict the geometry and physical characteristics of the CT scanner.”

Shin and colleagues evaluated low-tube current scanning in a study featured seven colon phantoms and 68 simulated polyps. The phantoms were prepared from pig colons and the simulated polyps were all 6 mm or larger. Phantoms were scanned at three peak voltage settings—80,100 and 120 kVp—and 10 mAs. To compare reconstruction techniques, images were reconstructed with FBP, hybrid statistic-based IR and knowledge-based IR algorithms.

Median effective radiation dose of CT colonography at 80 kVp and 10 mAs was only 0.167 mSv, which is lower than doses from plain abdominal radiography. However, this low dose technique is not feasible with standard FBP because of heightened image noise and a per-polyp sensitivity that dipped to 14.7 percent and 7.4 percent in the two reviewers, respectively.

While hybrid statistic-based IR helped decrease noise, full knowledge-based IR had the greatest effect. With the application of full knowledge-based IR, the two radiologists’ per-polyp sensitivity for 80 kVp CT colonography was 98.5 and 94.1, respectively, which is in line with sensitivities seen at 100 and 120 kVp, and acceptable in practice.

“Although the results of a phantom study cannot be directly applicable to those of a human study, our results suggest that reduced-dose CT colonography at 80 kVp and 10 mAs may be feasible with the aid of knowledge-based IR on CT colonographic examinations,” summed Shin and colleagues.