Garen Boghosian MD, Presenter: Nothing to Disclose

Adam E. Flanders MD, Abstract Co-Author: Consultant, Geron Corporation, Denver, CO

Richard Joseph Thomas Gorniak MD, Abstract Co-Author: Nothing to Disclose

Khan Mohammad Siddiqui MD, Abstract Co-Author: Partner, iVirtuoso, Inc, Baltimore, MD Medical Advisory Board, General Electric Company, Barrington, IL Research Consultant, Mercury Computer Systems, Inc, Chelmsford, MA

Tabassum Ahmad MD, Abstract Co-Author: Nothing to Disclose

Katrina M. Read MS, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV

Mark Olszewski PhD, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV

CT perfusion (CTP) is an accepted technique for obtaining quantitative measures of cerebral perfusion by measuring the first pass kinetics of a intravenous contrast bolus through the cerebral circulation. All vendors of CTP post-processing software use similar convolutional algorithms to derive the cerebral blood flow characteristics from a time-resolved CTP dataset. The purpose of this study is to determine if different post-processing vendors produce similar CTP results from identical source datasets.

The CTP source data from ten patients were analyzed retrospectively on three discrete vendor platforms by three readers. A single slice location was chosen for analysis. Arterial input and venous outflow functions were generated by each vendors semiautomated method using the anterior cerebral artery and superior sagittal sinus. Regions of interest (ROI) were manually traced of the lentiform nuclei in both hemispheres by each reader. Each measurement was repeated twice by each reader. Arterial input function measures such as time-to-peak (TTP), change in Hounsfield units and calculated perfusion metrics [cerebral blood volume (CBV), cerebral blood flow (CBF) and mean transit time (MTT)] of the ROIs were compared across platforms. Inter-rater and intra-rater variations were also calculated.

Statistical comparison of CTP values derived from the vendor post-processing packages showed significant differences between platforms overall (p<0.001). This included significant differences in CBV (p<0.001), MTT (p<0.01) and TTP (p<0.01) values. Interestingly, there was no signficant difference in CBF values (p=0.193) between vendor platforms. This incongruous result may be a direct effect of substantial intra- and inter-observer variability. Variability between vendor platforms exceeded observer variation.

CT perfusion (CTP) calculations can vary considerably when processed on different vendor platforms.

Accuracy in determination of the absolute values of CBV and MTT are critically important in predicting size of ischemic core and penumbra (respectively) in stroke.

Boghosian, G, Flanders, A, Gorniak, R, Siddiqui, K, Ahmad, T, Read, K, Olszewski, M, Vendor Variability in Brain CT Perfusion (CTP) Calculations.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5006692.html