Ayso de Vries MD, Presenter: Nothing to Disclose

Gert Antonius Schoonenberg, Abstract Co-Author: Nothing to Disclose

Henk Willem Venema PhD, Abstract Co-Author: Nothing to Disclose

Simona Grigorescu, Abstract Co-Author: Nothing to Disclose

Joost Peters, Abstract Co-Author: Nothing to Disclose

Jaap Stoker MD, PhD, Abstract Co-Author: Nothing to Disclose

Roel Truyen, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

Purpose of this study is to evaluate the feasibility of computer aided polyp detection (CAD) on ultra low dose CT-colonography data and compare it to normal dose data.

Twenty selected patients with at least one colonoscopically proven polyp larger than 5mm were scanned with a multislice CT scanner (Philips Mx8000) in 2 positions. Patients were extensively prepared. The scans were performed with the following parameters: tube voltage 120 kV; 25 to 100 mAs (average 70 mAs); collimation, 4 x 2.5 mm; rotation time, 0.75 second; pitch, 1.25; section thickness, 3.2 mm and reconstruction interval 1.6 mm. Ultra low dose scans (6.25 mAs) were generated using a validated simulation technique. As a reference we used polyps detected by an experienced reviewer and confirmed by colonoscopy. The reference contained 70 polyps in total. The CAD algorithm was the same for both dose levels, but the system was trained on data of the corresponding dose level only. The results were obtained using the leave-one-out method at patient level. The performance (sensitivity and number of false positives) of the CAD was measured and compared between normal and ultra low dose. Results are reported per polyp.

At a sensitivity of 90% we measure a median number of 6 false positives for normal dose and 9 false positives for the ultra low dose data. Moving the operating point of the algorithm to also reach 6 false positives at ultra low dose causes the sensitivity to drop to 87%.

CAD on (simulated) ultra low dose data (as low as 6.25 mAs) is feasible at the cost of only a slight increase of number of false positives or a small drop in sensitivity as compared to normal dose.

S.G.,J.P.,R.T.: Authors are employees of Philips Medical SystemsA.d.,G.A.S.: Authors receive a research grant of Philips Medical Systems

de Vries, A, Schoonenberg, G, Venema, H, Grigorescu, S, Peters, J, Stoker, J, Truyen, R, et al, , Feasibility of Automated Detection of Colon Polyps on Simulated Ultra Low Dose CTC Datasets.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4418790.html