Local cerebral blood flow by Xenon enhanced CT.

Local Cerebral Blood Flow by Xenon Enhanced CT To the Editor: We take exception to the remarks of Gur et al in their recent Progress in Cerebrovascular Disease article (Stroke 13: 750-758, 1982), which may mislead your readership as to the magnitude of the errors associated with xenon CT measurements of rCBF. In discussing our error analysis of the single-scan "autoradiographic" approach to xenon CT rCBF measurements' (an approach originally championed by Gur and his colleagues ), Gur et al fault us for using the fast-flow component, fg, rather than "total flow" as the reference flow in our definition of percent error. They go on to conclude that "if one corrects for this mistake the errors are significantly lower than reported." As is apparent from figures 1 and 2, which depict the effect of tissue (gray matter-white matter) heterogeneity in the absence of CT noise, the magnitude of measurement error depends upon (i) whether gray matter (GM) flow, fg, or "total flow," f, = w g f g + w w f w [where w are proportional weights, and the subscripts g and w refer to gray and white matter (WM), respectively], is used as the reference flow and (ii) wheth er X g or X, [ = Wg>.g + w w X w ] is used as the blood-brain partition coefficient for xenon in Kety's blood flow equation. Three definitions of percent error (E,, E 2 , E 3 ) are illustrated: E, = (fg kgXg)/fg, E 2 = (f( k,A 1)/f landE 3 = (f, kgX.g)/f,, where k g and k, are the solutions to Kety's equation when Xg and A,, respectively, are substituted for the partition coefficient. We originally proposed E, to emphasize that, in a onecompartmental model, an anatomic region of interest (ROI) is presumed to be homogeneous ; the quantity k%Xg = fg was therefore interpreted as estimated GM flow, and the "true" GM flow, fg, was taken as the reference flow in Equation Ej. If, however, one measures (or extrapo lates) the regional partition coefficient, this "measured X" is, in a heterogeneous ROI, the weighted average of GM and WM partition coefficients, and the quantity k,X, = f, could be interpreted as estimated "total flow," though the physical meaning of f, is problematical. If f, is taken as the reference flow, then E 2 is probably an appropriate definition of percent error. In contrast, the use of E 3 , proposed by Gur et al 3 as the most favorable definition of percent error, cannot be justified; k gX g should not be taken as an approximation of total flow, and the use offt as the reference flow is inappropriate. Using the parameter values in the legend to figure 1, errors predicted by E 3 are significantly lower than