Computer-aided detection (CAD) using 360 degree virtual dissection: can CAD in a first reviewer paradigm be a reliable substitute for primary 2D or 3D search?

Kristina T. Johnson, Joel G. Fletcher, C. Daniel Johnson

Research output: Contribution to journalArticle

7 Scopus citations

Abstract

OBJECTIVE: The purpose of this study was to evaluate the feasibility of a new computer-aided detection (CAD) software program as a first reviewer for detecting colorectal polyps when applied to 360 degrees virtual dissection image display. MATERIALS AND METHODS: Forty-one consecutive patients who underwent imaging without oral contrast material for stool tagging from a teaching file database constituted the patient population for this feasibility study. Using CT colonography equipped with CAD software, reviewers evaluated each possible polyp detected by the software using virtual dissection images combined with axial and 3D endoluminal views and compared the results with optical colonoscopy, the reference standard. Two experienced radiologists blinded to the reference standard findings interpreted the CAD detections to be true or false. The false detections were reviewed and categorized. RESULTS: Sensitivities for polyps that were 6-9 mm were 78.3% (18/23) and 91.3% (21/23) for reviewers 1 and 2, respectively. For polyps > or = 1 cm, sensitivities were 94.9% (37/39) and 97.4% (38/39) for reviewers 1 and 2, respectively. Per-patient sensitivities for polyps > or = 6 and > or = 10 mm were 94.4% (34/36) and 95.1% (39/41) for reviewer 1 and 97.2% (35/36) and 97.6% (40/41) for reviewer 2, respectively. The average number of false detections per acquisition was 4.28. The average interpretation times were 4 minutes 26 seconds and 5 minutes 38 seconds for reviewers 1 and 2, respectively. CONCLUSION: Colorectal polyp detection using CT colonography equipped with CAD and virtual dissection as a first reviewer is feasible. Detection rates are similar to colonoscopy. Interobserver variability is low and interpretation times are short. False-positive detections per patient are few in number.

Original languageEnglish (US)
Pages (from-to)W172-176
JournalAJR. American journal of roentgenology
Volume189
Issue number4
DOIs
StatePublished - Oct 2007

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging

Fingerprint Dive into the research topics of 'Computer-aided detection (CAD) using 360 degree virtual dissection: can CAD in a first reviewer paradigm be a reliable substitute for primary 2D or 3D search?'. Together they form a unique fingerprint.

  • Cite this