Thierry Nicolas Boellaard MD, PhD, Presenter: Nothing to Disclose

Vincent F. van Ravesteijn, Abstract Co-Author: Nothing to Disclose

Marije Petronella van der Paardt MD, Abstract Co-Author: Nothing to Disclose

Frans M. Vos PhD, Abstract Co-Author: Nothing to Disclose

Lucas J. van Vliet PhD, Abstract Co-Author: Nothing to Disclose

Jaap Stoker MD, PhD, Abstract Co-Author: Research Consultant, Robarts Clinical Trials

Bowel preparation is a burdensome aspect of CT colonography and preferably the least burdensome (i.e. minimal) preparation is used. The use of 3D and especially a primary 3D read are hindered by cleansing artifacts (e.g. incomplete cleansing causing floating debris and ridges at air-tagging boundaries), allowing a 2D reading only. We hypothesized that we could develop an algorithm that allows a primary 3D read. The purpose was to develop and test a new cleansing algorithm, to enable a primary 3D read in minimally prepped CT colonography.

Sixty CT colonography cases with 4 different minimal bowel preparations (N=4x15; all 24-hour and with 22.5, 30, 45 and 60 grams of iodine, respectively) were used. After developing a new minimal principal curvature algorithm, it was optimized using 20 cases. Subsequently, two observers read the remaining 40 cases twice, primary 2D (uncleansed) and primary 3D (cleansed), with a 6 week interval and at random 2D or 3D first. Reader confidence (3-point scale), effort (4-point scale), and time were scored and colonic lesions were annotated. To evaluate the algorithm’s effects on polyp shape, the conspicuity was scored (5-point scale) for air-surrounded polyps and for tagging-surrounded polyps after cleansing. For diagnostic performance a McNemar, for reading time a paired t-test and for 3-5 point scales a Wilcoxon signed-rank test was performed to compare two reading strategies. A P value < 0.05 indicated a significant difference.

Sensitivity for >=6mm lesions was higher for primary 3D read than primary 2D (84% (57/68) vs 68% (46/68); P=0.013). Specificity for >=6mm cases was similar (3D 96% (48/50) vs 2D 94% (47/50); P=1.00). Reader time (5 min 39 sec vs 7 min 9 sec; P=0.005) and confidence (2.3 vs 2.1; P=0.013) were superior for primary 2D read. Reader effort was similar (2D 2.8 vs 3D 2.6; P=0.06). Conspicuity was similar for cleansed submerged lesions (P=0.06).

A minimal principal curvature flow cleansing algorithm enables a primary 3D read in minimally prepped CT colonography. Although confidence and reading time are somewhat better in a primary 2D read, most importantly the sensitivity was higher for >=6mm lesions in a primary 3D read.

Minimally prepped CT colonography can be read 2D only, while optimal reading requires both 2D and 3D reading. Our new cleansing algorithm enables a 3D reading in minimally prepped CT colonography

Boellaard, T, van Ravesteijn, V, van der Paardt, M, Vos, F, van Vliet, L, Stoker, J, Electronic Cleansing for Limited Bowel Preparation CT Colonography Using Minimal Principal Curvature Flow.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12031891.html