Stuart Andrew Taylor MBBS, Presenter: Research Consultant, Medicsight, Inc, London, UK Grant, E-Z-EM, Inc, Westbury, NY

Gen Iinuma MD, PhD, Abstract Co-Author: Nothing to Disclose

Steve Halligan MD, Abstract Co-Author: Research Consultant, Medicsight, Inc, London, UK

Leslie Honeyfield, Abstract Co-Author: Employee, Medicsight, Inc

Mary Elizabeth Roddie MD, Abstract Co-Author: Consultant, Medicsight, Inc, London, UK

Aurora Costea, Abstract Co-Author: Employee, Medicsight, Inc

CT colonography is well-established for detection of adenomatous polyps in asymptomatic patients and the role of computer-assisted-detection (CAD) is increasingly advocated now that systems are available commercially. However, the potential of CAD for cancer detection in symptomatic patients has been relatively neglected. We investigated CAD for detection of cancer rather than polyps

Ethical permission was granted. Prone and supine CT colonography was performed in 59 patients with 59 proven cancers (established by colonoscopy and histology) following full-bowel preparation and using multi-detector row machines. Patient datasets were analysed using a commercially available CAD system (ColonCAD 4.0, Medicsight PLC), applied at four different filter settings (0, 50, 75, 100). Computer prompts for each patient were categorised by a single observer as either true positive if the detection boundary overlapped tumour boundary, or false-negative if located elsewhere. For the purposes of this study, detections on benign polyps were regarded as false-positive. Statistical advice was sought and sensitivity (true-positive) and specificity (false-positive) rates at each of the four different filter settings was determined.

Sensitivity: CAD detected 54 (91.5%), 53 (89.8%), 53 (89.8%), and 47 (88.7%) of the 59 cancers at filter settings of 0, 50, 75, and 100 respectively. Specificity: The median number of false-positive CAD marks per patient was 19, 17, 12, and 6 respectively, including polyps. Of the 54 cancers detected by CAD, 21 (38.9%) were only detected on either the prone or supine dataset

CAD is effective for detection of cancer. Similar to polyp detection, optimal results for CAD require both prone and supine CT acquisitions.

CAD is effective for detection of cancer, and may have a role in a symptomatic patient population.

Taylor, S, Iinuma, G, Halligan, S, Honeyfield, L, Roddie, M, Costea, A, CT Colonography: Computer-assisted Detection (CAD) of Colorectal Cancer.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5003496.html