Luca Roccatagliata MD, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: An important advantage of using CT, rather than MR, perfusion imaging in acute stroke triage is speed and simplicity. This concept extends to image interpretation. In this abstract, we ask if a quick, visual, freehand estimate of the size of ischemic lesions on CT cerebral blood volume (CBV) and cerebral blood flow (CBF) maps is sufficient to obviate the need for a more careful, thresholding approach to delineating abnormal regions. Specifically, we compare inter-observer variability in the detection of hypoperfused CBF and CBV lesion areas, between using a "quick and dirty" approach, versus a more careful, methodical approach. Methods and Materials: 10 consecutive acute stroke patients received quantitative, first pass CT perfusion (CTP) imaging prior to intra-arterial thrombolysis. Using a commercially available software package, two independent observers segmented ischemic regions of interest (ROIs), by hand, on the resulting CBV and CBF maps. The first observer segmented the ROIs by gross visual inspection alone, without knowledge of the final infarct size. The second observer, who was unblinded to the final infarct size, segmented the ROIs using a threshold tool to better recognize the boundaries of the lesion, and was careful to eliminate CSF filled spaces, uncalculated pixels, and large vessels from the field-of-view. For each of the CBF and CBV maps, we also calculated a "normal gray / white matter (GM, WM) conspicuity ratio" {mean GM ROI - mean WM ROI} / {mean WM ROI}, an image "signal-to-noise ratio" {mean GM ROI - mean WM ROI} / Ö {(standard deviation GM ROI)2+ (standard deviation WM ROI)2}, and an "infarct conspicuity ratio" {mean GM ROI - mean maximally hypodense lesion ROI} / {mean GM ROI}. Results: The variability in ischemic area measurements was smaller for lesions on the CBF maps (±11%), than for lesions on the CBV maps (±33%). Mean gray matter conspicuity was 1.78 for CBF maps and 1.00 for CBV maps, mean signal to noise ratio was 0.98 for CBF maps and 0.61 for CBV maps, and mean infarct conspicuity was 0.85 for CBF maps and 0.64 for CBV maps (all differences were significant, p<0.05). Conclusion: Segmentation of ischemic regions on CBF maps by gross visual inspection is significantly more reliable than segmentation of CBV maps, and is likely sufficient for estimating "ischemic penumbra" size in routine clinical practice. Because CBV maps have inherently worse signal-to-noise ratio and infarct conspicuity than CBF maps, estimating the size of the "infarct core" may be facilitated by a more methodical thresholding approach to segmentation.       Questions about this event email: lroccatagliata@partners.org

Roccatagliata MD, L, Estimating the Size of Ischemic Regions on CT Perfusion Maps in Acute Stroke: Is Freehand Visual Segmentation Sufficient?.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3108218.html