RSNA 2012 

Abstract Archives of the RSNA, 2012


SSJ08-06

Computer-aided Detection for Minimal-Preparation Multi-Energy CT Colonography

Scientific Formal (Paper) Presentations

Presented on November 27, 2012
Presented as part of SSJ08: Gastrointestinal (CT Colonography Technique)

Participants

Janne Johannes Nappi PhD, Presenter: Royalties, Hologic, Inc Royalties, MEDIAN Technologies
Hiroyuki Yoshida PhD, Abstract Co-Author: Patent holder, Hologic, Inc Patent holder, MEDIAN Technologies
Se Hyung Kim, Abstract Co-Author: Consultant, INFINITT Healthcare Co, Ltd

PURPOSE

To assess the detection accuracy of CAD for colorectal lesions at minimal-preparation multi-energy CT colonography (ME-CTC). ME-CTC can improve detection performance under challenging reading conditions, because it provides more information about the chemical composition of colonic materials than does conventional CT colonography.

METHOD AND MATERIALS

Thirty-seven patients (23 males, 37-76 yrs (mean age 58 yrs)) with suspected lesions underwent one-day minimal bowel preparation (18 g of magnesium citrate mixed with water). For fecal tagging, Group A (12 patients) ingested iodine and barium, and Group B (25 patients) ingested 50 ml of non-ionic iodine only. No intravenous contrast was used. CTC examinations were performed in dual positions at 140 kVp and 80 kVp by use of a dual-energy CT scanner with 1-mm slice thickness. Experienced radiologists correlated the ME-CTC data with colonoscopy-confirmed findings. A fully automated multi-energy CAD (ME-CAD) scheme was developed and trained to detect colorectal lesions with Group A based on dual-energy information. The trained ME-CAD was then tested with Group B. In addition, conventional CAD was trained and tested similarly by use of virtual single-energy CT colonography images of the studies. The detection accuracies of ME-CAD and conventional CAD were assessed and compared by use of the non-parametric method recommended for CAD evaluations.

RESULTS

Group B had 39 lesions (20 ≥10 mm, 19 6 – 9 mm) in 15 patients. The per-lesion (per-patient) sensitivity of the ME-CAD was 100% (100%) for lesions ≥10 mm at a median of 8 (4) false positives (FPs) per study and 89% (100%) for lesions 6 – 9 mm at 7 FPs per study. For flat lesions (7 ≥10 mm), the per-lesion (per-patient) sensitivity of ME-CAD was 100% (100%) at 8 (1) FPs per study. For cancers (4 ≥10 mm), the detection sensitivity of ME-CAD was 100% at 4 FPs per study. The detection accuracy of ME-CAD for lesions ≥6 mm was significantly higher than that of conventional CAD (p<0.01).

CONCLUSION

The ME-CAD scheme that was developed in this study yields clinically acceptable detection accuracy for significant colorectal lesions and has potential to outperform conventional CAD in minimal-preparation ME-CTC.

CLINICAL RELEVANCE/APPLICATION

Minimal colon preparation can improve patient compliance for colorectal examinations. Multi-energy information can improve the detection accuracy of computer-aided detection (CAD) and CT colonography.

Cite This Abstract

Nappi, J, Yoshida, H, Kim, S, Computer-aided Detection for Minimal-Preparation Multi-Energy CT Colonography.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12037382.html