qi The effects of CT drift on xenon/CT measurement of regional cerebral blood flowa) K. J. Kearfott, b) H. C. Lu, and D. A. Rottenberg Department of Neurology, Memorial Sloan—Kettering Cancer Center, New York New York 10021 M. ID. F. Deck Department of Radiology, Memorial Sloan—Kettering Cancer Center, New York, New York 10021 (‘Received 2’2 November 1983; accepted’for publication 24 April 1984) A systematic increase in computed tomography (CTi) number of approximately 0.13 Hounsfield unit per scan (‘HU/scan) was observed when serial øeltaScan 2020 CT scans of a uniform water phantom were equally spaced at-0.5, 1.0, or 2.0 mm and a shaped aluminum beam-hardening filter employed. Much smaller drifts (cO.O 6 HIJ/scan) were observed with flat aluminum or shaped beryllium oxide filters. T’his machine drift, which was not associated with a rise in water phantom temperature and did not consistently correlate with estimated x-ray tube heat, could result in a significant overestimation of regional cerebral blood flow (r€BF) for a xenon/€iT rCBF protocol involving 5—7 sequential scans obtained at 1-mm interscan intervals. INTRODUCTION Techniques have recently developed for clinical measurement of regional cerebral blood flow (rCBF) using stable xenon and computerized tomography. The accura cy of such measurements depends upon statistical (noise) limitations well the mathematical model and scanning protocol emp1oyed. This paper describes systematic drift in computed tomography (CT’) number observed with the Ie1taScan 2020, which, though insignificit for conven tional CT imaging, cannot neglected for quantitative Xe non/CT rCBF studies involving regional enhancements (in creases CT number) of 2.5—10 Hounsfield units (HU), which are typically observed when 40%—50% xenon is in haled. The explanation for this drift remains unknown, al though an empirical correction is possible. METHODS Standard head scans (-50 mA, s, l’20 kVp) were per formed using the IDeltaScan 2020 CT scantier. In order j examine the relationship between observed €T number and percent maximum x-ray tube heat (estimated and indicated on the operator’s control console), we repeatedly scanned a 22-cm-diam uniform water-filled phantom. Following each scan, time (0.7 to 2.0 mm) was allowed for the tube to “cool” to a specified console-indicated tube heat level before the next scan was initiated. Thirteen scans were also obtained with an interscan spacing of approximately 2 mm such that the console-indicated tube heat was always 43%—44% at the beginning of each scan. Systematic drift in CT number was further investigated by serially scanning uniform water phantom at equal inter- scan intervals of 0.5, 1.0, or 2.0 iifin using 100-mA 4-s and 50-mA 8-s techniques with an energy of 120 kVp. An alumi num “bow-tie” shaped filter was used for these experiments. To investigate the effects of the physical filter on CT drift, studies with 50-mA 8-s technique and 1-mm equal interscan spacing were performed on the water phantom with a 3-mm- flat aluminum filter and a shaped beryllium oxide ifiter. iW20.4-cm-diam Lucite phantom were also obtained using the flat aluminum and shaped beryllium oxide filters in order to determine the dependence of scanner drift on CT number (phantom material). Large (1’29 x 129 pixel) regions of interest (ROTs) centered in the images were used for data analysis to mitigate the dS of statistical noise on the results while avoiding the edges of (1x3 phantom. The average drift in HU/scan was determined from the slope of linear least-squares fit of average €‘T number a function of scan number. The possi bility that CT drift resulted from a progressive increase in phantom temperature was studied-by using a thermocouple thermometer immersed in the water phantom, the output of which was monitored continuously by means of a strip-chart recorder. significance of observed drift on (end-tidal) xenon/ CT rCBF measurements was investigated by simulating ar terial xenon and CT enhancement curves with and without systematic drift. FiveR seven scans equally spaced at 1-mm intervals were simulated with systematic drift of 0.01 to 0.20 H’I!J/scan. Protocols consisting of five scans equally spaced at@17-5 or 1.2-5 mm were also considered. A step-exponential arterial input function (4 = emf, where m = 1.4 min ‘)was assumed. Gray and white matter partition coefficients for xenon were taken as 0.8 and 1.5, respectively, and the corre sponding compartmental rate constants were assumed to be 0.8 and ®l5 min. Maximum enhancements of 5 and 10 HL!J for gray matter and 2.5 and 5.0 HI!J for white matter were assumed. RESULTS AND DISCUSSION Figure 1(a) illustrates the observed linear relati • nship between the change in CT number and percent maximum tube heat indicated on the operator’s console. Unfortu nately, the relationship between console-indicated tube heat and CT number may not be used to fully account for the observed CT drift- since a significant drift was observed when the apparent tube heat was maintained constant throughout 686 Med. Phys. 11(5), Sep/Oct 1984 0094-2405/84/050686-04$01 .20 1984 Am. Assoc. Phys. Med. 686