Stuart Pomerantz MD, Presenter: Speaker, General Electric Company Advisory Board, General Electric Company Research support, General Electric Company

Grzegorz Babiarz, Abstract Co-Author: Nothing to Disclose

Shahmir Kamalian MD, Abstract Co-Author: Nothing to Disclose

Mukta Joshi, Abstract Co-Author: Employee, General Electric Company, Chalfont St. Giles, UK

Ramon G Gonzalez, Abstract Co-Author: Nothing to Disclose

Michael Howard Lev MD, Abstract Co-Author: Speaker, General Electric Company Advisory Board, General Electric Company Speaker, Bracco Group Advisory Board, Bracco Group Advisory Board, CoAxia, Inc

Purpose: Although CTA offers critical data in emergency neurovascular evaluation, widespread adoption is limited by need for manual 3D post-processing. We studied if a novel post-processing algorithm utilizing prototype automatic bone removal software - executed faster, without special training vs. manual techniques - can provide similar diagnostic accuracy for intracranial aneurysm detection. Methods: All 64-slice CTAs performed over 6-months for intracranial aneurysm detection, for which there was unequivocal proof of diagnosis, were post-processed on a 3D workstation using both the manual algorithm routine at our institution, and a semi-automated algorithm with automatic bone removal (prototype software, GE Medical Systems). Mean processing time was recorded in a representative subset. Cases were divided into separate sets by aneurysm location [Ant. Communicating (ACOM), Mid. Cerebral (MCA), Post. Communicating (PCOM), Int. Carotid (ICA), Ant. Cerebral (ACA), Basilar (BA) and Ophthalmic (OPH)] and randomized with normals. A single reader, blinded to the final diagnosis, rated each set for the presence of aneurysms. Results: Mean processing time was 12 minutes for the manual and 3 minutes for the semi-automated algorithm. There were 62 aneurysms in 56 cases (34% <5 mm, 43% 5-10mm, and 23% 10-25 mm); 41 cases were normal. Overall sensitivity for aneurysm detection was 79% for the manual and 82% for the semi-automated algorithm. Specificity was 99% for both algorithms. Pairwise percent sensitivities for the manual versus semi-automated algorithms by location were, respectively: ACOM(70, 100), MCA(86,57), PCOM(86,71), ICA(50,75), ACA(75,75), BA(87,87) and OPH(100,100). Pairwise specificities were: ACOM(100,100), MCA(100,100), PCOM(100,97), ICA(100,100), ACA(91,100), BA(100,100) and OPH(100,100). Conclusions: CTA post-processing utilizing Automated-Bone Removal provides at least as good accuracy as manual techniques for intracranial aneurysm detection, but takes only 25% of the time and does not require specialized skills. Automated 3D post-processing has the potential to make CTA a more practical technique for neurovascular screening. Disclosure: See attached

Pomerantz, S, Babiarz, G, Kamalian, S, Joshi, M, Gonzalez, R, Lev, M, Hot Topic: 3D Post-Processing of CT Angiography Datasets with Prototype Automated Bone Removal Software: Faster, Easier, and at Least as Accurate as Manual Techniques for Intracranial Aneurysm Detection.  Radiological Society of North America 2006 Scientific Assembly and Annual Meeting, November 26 - December 1, 2006 ,Chicago IL. http://archive.rsna.org/2006/8002499.html