Study lends weight to 3D as primary CTC reading strategy

A new study suggests that in CT colonography scanning, 3D interpretation with 2D confirmation is better at finding cancerous polyps in an asymptomatic population than other approaches. The study also validated earlier large and widely publicized CT colonography studies that relied on 2D interpretations and produced less than stellar results.

The study conducted at the University of Wisconsin and presented at the RSNA meeting Wednesday compared different reading results and used data from a Department of Defense study of 730 consecutive colonscopy-proven CT cases using 2D for primary interpretation. Those results were compared with data from 1223 asymptomatic patients and interpreted using 3D as the primary method.

"We would read with 3D for detection, as it is more sensitive, and 2D for confirmation, as it is more specific," said presenter Dr. Andrew D. Lee, a fellow in abdominal imaging.

In a patient-by-patient analysis, sensitivity for polyps 6 mm or greater was 43% using the primary 2D interpretation and 84% using the 3D interpretation. For polyps 10 mm or greater, it was 63% with 2D and 86% with 3D. This compared with 39% and 55%, respectively, in a 2004 study by Cotton, and 56% and 59% in a 2005 study by Rockey.

Lee noted that the sensitivity of the 2D method was similar to that found in the previous large 2D screening trials.

Others who observed the presentation, however, said trial methods may have biased the results. Lee said the 10 radiologists who participated in the study were experienced CT colonographers, something that might have boosted their 2D results but didn't. But a member of the audience wondered if they were well trained in 3D interpretation but not in 2D reads.

"Most tend to start with 2D but gravitate to 3D," Lee said. "In general, they started out in 2D but became more comfortable with 3D. The software we use for 2D capabilities in my assessment was very good, but you do have a good point."

A session cochair, Dr. Judy Yee of the University of California, San Francisco said the reading times found in the study - six to seven minutes per scan - may have been too fast and reflected a rush to get through a large volume of studies. She said 10 minutes would have been a more reasonable figure.