2D versus 3D primary read debate sputters at RSNA

December 1, 2004

Radiologists eagerly expecting a clear winner to emerge from the RSNA session on CT colonography interpretation technique walked away disappointed Tuesday afternoon. A series of four presentations provided no clear consensus on which method would increase sensitivity for polyp detection.

Radiologists eagerly expecting a clear winner to emerge from the RSNA session on CT colonography interpretation technique walked away disappointed Tuesday afternoon. A series of four presentations provided no clear consensus on which method would increase sensitivity for polyp detection.

Ever since the landmark Pickhardt trial indicated that CT colonography had sensitivity numbers as good as those of optical colonoscopy, researchers have been trying to replicate and to explain the good results (NEJM 2003;349(23):2191-2200). Many had postulated that increased sensitivity may have been achieved through the use of 3D primary reads with 2D problem solving by the Pickhardt group.

Two studies from the Tuesday RSNA session disagreed with that interpretation. Both indicated that while there was no significant difference in sensitivity between the two techniques, 3D primary reads took significantly longer to perform.

In the first study, Julie Lee from the New York University Medical Center presented a retrospective analysis of 30 patient data sets - 15 normal and 15 presenting with polyps. Investigators randomly assigned three radiologists with at least 25 CT colonography cases under their belt to read exams using either primary 2D or 3D evaluation. After one week, the readers switched to the opposite evaluation method.

Readers using 3D interpretations for primary reads took significantly longer, clocking in at an average of 16.4 minutes compared with 10.9 minutes for 2D primary reads. Differences in overall sensitivity were not statistically significantly for polyps larger than 6 mm.

Three-D primary interpretation has additional limitations, possibly missing hidden polyps, increasing the number of false positives, and increasing radiation dose due to multiple navigations, Lee said.

She did note that the study had an extremely small number of subjects and that reader experience was not uniform.

Another study done in Italy using the same software as the Pickhardt study evaluated 50 patients. Three radiologists read the exams. Mean sensitivity for 3D (76.6%) and for 2D (73.3%) were not statistically significant. Readers took an average of 18.8 minutes to read 3D exams versus 11.2 minutes for 2D exams. Fecal tagging was used in the study.

A study from researchers in Amsterdam came to a different conclusion. Investigators examined 77 patients. Three reviewers with varying levels of experience all received training on 30 cases using both 2D and 3D techniques. They first read the 77 exams using 3D, then 2D reads.

For large polyps (greater than 10 mm), 3D primary reads had an average 83% sensitivity compared with 72% for 2D. Thrre-D led to only one perception error (defined as a polyp or patient identified by at least one but not all reviewers), while 2D led to six errors.

A study from the University of Wisconsin, Madison tackled the topic from a slightly different perspective and took away a positive interpretation of increased reading time for 3D. It included 20 proven colorectal polyps from 14 patients. Results were presented in terms of the time and distance of 3D polyp visualization.

Investigators used the default fastest speed (1.54 cm/sec) for the 3D fly-through and a conservative 2D axial reading rate of 0.77 cm/sec for 2D.

Mean time and distance for polyp visualizations were 0.5 seconds and 1.1 cm for 2D reads. Distances and times for 3D increased to:

  • retrograde flight: 3.8 sec and 5.8 cm

  • antegrade flight: 3.0 sec and 4.6 cm

  • total (antegrade plus retrograde) flight: 6.8 sec and 10.4 cm

Researchers hypothesized that the longer time and distance used for 3D reads indicated that there was greater opportunity for polyp detection.

Dr. Michael Barish, an assistant professor of radiology at Brigham and Women's Hospital, questioned this assumption during a question and answer period. He asked if the Wisconsin researchers had quantified whether increased dwell time really did lead to increased sensitivity.

Study presenter Dr. Andrew Lee reiterated that they were not saying that radiologists shouldn't look at 2D for screening, but that 3D endoluminal viewing may have better sensitivity.

Moderator Dr. Judy Yee, chief of radiology at the San Francisco VA Medical Center, called the session a draw. She noted that many other issues could be involved in the increased sensitivity numbers seen by Pickhardt's group last year, including fecal tagging and reader training and experience. For the future, CAD may also play a role in improving sensitivity numbers.

"In the long run, I don't think there's going to be one right way to do the exam," Yee said.

For more information from the online Diagnostic Imaging archives:

CT colonography proponents scope easier target

Uncommon CT colonography perforations still cause concern