A New Method for Quantitative Regional Cerebral Blood Volume Measurements Using Computed Tomography
Background and Purpose Knowledge of cerebral blood volume (CBV) is invaluable in identifying the primary cause of brain swelling in patients with stroke or severe head injury, and it might also help in clinical decision making in patients thought to have hemodynamic transient ischemic attacks (TIAs). This investigation is concerned with the development and clinical application of a new method for quantitative regional CBV measurements.
Methods The technique is based on consecutive measurements of cerebral blood flow (CBF) by xenon/CT and tissue mean transit time (MTT) by dynamic CT after a rapid iodinated contrast bolus injection. CBV maps are produced by multiplication of the CBF and MTT maps in accordance with the Central Volume Principle: CBV=CBF×MTT. The method is rapid and easily implemented on CT scanners with the xenon/CBF capability. It yields CBV values expressed in milliliters of blood per 100 grams of tissue.
Results The method was validated under controlled physiological conditions causing changes that were determined both with our technique and from pressure-volume index (PVI) measurements. The two independent estimates of CBV changes were in agreement within 15%. CBV measurements using this method were carried out in normal volunteers to establish baseline values and to compare with values using the ratio-of-areas method for calculating both CBF and CBV from the dynamic study alone. Average CBV was 5.3 mL/100 g. The method was also applied in 71 patients with severe head injuries and in 1 patient with hemodynamic TIAs.
Conclusions The primary conclusions from this study were (1) the proposed method for measuring CBV accurately determines changes in CBV; (2) the MTT×CBF determinations are in agreement with the ratio-of-areas method for CBV measurements in normal volunteers and are consistent with other methods reported in the literature; (3) MTTs are significantly prolonged early after severe head injury, which when combined with the finding of decreased CBF and increased arteriovenous difference of oxygen indicates increased cerebrovascular resistance due to narrowing of the microcirculation consistent with the presence of early ischemia; and (4) CBV in the patient with TIAs was increased in the hemisphere with the occluded internal carotid artery, indicating compensatory vasodilation and probable hemodynamic cause.
- cerebral blood volume
- cerebral blood flow
- tomography, emission computed
- intracranial pressure
- diagnostic imaging
According to the Monro-Kellie doctrine, changes in ICP are determined by changes in brain tissue mass, CBV, and CSF. Treatment of raised ICP, usually after trauma or stroke, requires reduction of CBV, CSF, or both. If either of these compartments is increased, it would be best, in principle, to reduce that particular compartment to treat high ICP. However, it is not always known why ICP is increased, and the question of edema (vasogenic edema or cytotoxic edema) or increased CBV with “vasoparalysis” being responsible is difficult to answer in individual cases of stroke or traumatic brain injury. A number of authors have equated high CBF with increased CBV and ICP, but we have shown elsewhere that the relationship between CBF and CBV is not straightforward.1 Thus, the only way to resolve this problem is by measuring CBV directly.
Another reason to measure CBV is that in the case of low CBF, knowledge of CBV gives some insight into the reason for low CBF. For instance we have found global or regional ischemia early after head injury; if vasospasm of the conducting arteries were the cause, increased CBV in the area of ischemia could be expected,2 but if diminished cerebral metabolic rate of oxygen, increased local brain pressure due to edema or clogging of the resistance, or capillary vessels with leukocytes was the cause, diminished CBV could be expected.3 A third reason to measure CBV is that changes in CBV or in the CBV/CBF ratio should allow for the differentiation between irreversible infarction and reversible ischemia in cases of stroke, TIA, or subarachnoid hemorrhage.2 3 4 5 In all these cases, low CBF is present either locally, regionally, or globally. However, when irreversible infarction is present, low CBF is accompanied by low CBV; whereas in viable brain tissue, compensatory vasodilation in the microcirculation takes place, resulting in high CBV.3
Although a number of methods for measuring CBV have been described, ideally the method should allow regional CBV to be correlated with regional CBF. Thus far, only PET can accomplish this, but the availability of PET scanners is very limited, the running costs are extremely high, and it is practically impossible to do very acute (within 30 to 60 minutes of admission) measurements. To circumvent these problems, we have devised a method of measuring CBV using a CT scanner equipped with a commercial package for measuring CBF with stable xenon. CBV can be calculated with the simple formula where MTT is mean transit time.
Calculation of local, regional, or global CBF can easily be accomplished with the stable xenon-CT method, sometimes even within 1 hour of a stroke or postinjury in a severely head-injured patient. In this article we describe a method to derive MTT by rapid serial (“dynamic”) scanning while an intravenously injected bolus of radiographic contrast passes through the brain. Moreover, MTT and CBV global and regional data from normal volunteers are presented. Finally, changes in CBV measured with this new method are correlated with changes in CBV as calculated from the changes in ICP with known PVI6 before and after hyperventilation. All studies were approved by the Committee on Conduct of Human Research at the Medical College of Virginia.
Theory And Principles
Dynamic CT to Measure MTT
If a radiopaque medium is administered intravenously as a bolus, its passage through the brain can be monitored by means of rapid serial CT scanning. Modern CT scanners offer the option of continuous scanning for several tens of seconds, with scan times sufficient for mapping out the tissue concentration curve. The tissue enhancement expressed in Hounsfield units can be plotted against time for a selected brain slice following the rapid injection of a 50-mL bolus of iodinated contrast (Fig 1⇓). The points in the graph represent the CT enhancement in two consecutive measurements in the same ROI but with different Paco2. The continuous curves represent gamma-variate fits to the data (see discussion below and Fig 2⇓). The rise of the data after the first bolus pass represents contrast recirculation. The possibility of using a CT scanner to measure CBV was first discussed by Axel in 1980.7 There are some fundamental issues involved in applying a CT technique to the measurement of MTTs: (1) CT scanning measures the residual indicator in the tissue rather than the concentration in the draining veins as is required by the “central volume principle”; (2) the input bolus into the brain is not infinitesimally brief (ie, a “delta- like function”); and (3) there is incomplete mixing of the bolus with the capillary blood.
To address these issues, a well-defined artery and vein must be visible in the CT slice. However, no major arteries are available in the scan at a level above the circle of Willis. The arteries that can be identified are small and subject to partial-volume averaging with surrounding tissue. For these reasons, we have chosen to use the IW as a measurement of the mean transit time, <t>. The IW is computed from the fitted curve—a gamma-variate function—and represents the transit time of the densest part of the bolus. This function conveniently describes indicator dilution curves without recirculation: where t is the time after injection; ta is the indicator appearance time; k, α, and β are fit parameters; and C(t) is the indicator concentration that is proportional to the CT number. The arrival time ta as well as the choice of a cutoff point on the downslope side (to avoid recirculation) must be specified by the operator or computed automatically. The use of a gamma-variate function greatly simplifies the computation of the MTT. Specifically, the following results are applicable for functions of the form given above. IW=width across inflection points of C(t) It should be noted that at the inflection points the second derivative, C“(t), vanishes. These points on the tissue concentration curve correspond to the entrance of the densest bolus part (on the upslope) and its exit (on the downslope). We have chosen to use the IW as a measure of the tissue MTT in order to minimize several methodological issues inherent in the use of this technique with CT. A gamma-variate function is shown in Fig 2⇑ for typical values of k, α, β, and ta. Point A marks the appearance time of the bolus in a brain voxel after a peripheral injection at time zero. Point B marks the arrival of the densest part of the bolus, a point of inflection on the upslope side of the tissue enhancement curve. Indicator particles continue to arrive after point B until the entire bolus is essentially contained within the scanned tissue and the enhancement curve reaches its peak at point C. Beyond that point, the tissue enhancement decreases correspond to outflow of iodine. Point D marks the exit of the densest part of the bolus. The IW represents the MTT of the densest portion of the bolus.
Although the ratio-of-areas method is attractive for measuring CBV because it only requires a dynamic study,8 9 in practice a number of difficulties complicate its use for absolute measurements. Our approach to measurement of CBV is based on independent measurements of CBF by the stable xenon/CT method and of the cerebral mean transit (IW) of a nondiffusible indicator (iodine) using rapid sequential CT scanning.10 CBV is then calculated from the central volume principle by simple multiplication.
Measurement of CBF
After a diagnostic CT scan, a stable xenon/CT CBF study is performed on a Siemens CT/Plus or GE 9800 Scanner equipped with a xenon gas delivery system and a CBF software analysis package. Scans are performed at three axial planes with a thickness of 5 mm each, 20 mm apart. Two baseline scans are performed at each level, followed by multiple enhanced scans during inhalation of 30% xenon and 70% oxygen. From measurements of CT enhancement and the end-tidal curve, CBF maps are calculated by means of the Kety-Schmidt equation.
Measurement of MTT
Following the xenon/CT CBF study, dynamic CT scans are performed at the middle level for which CBF was previously determined. A bolus of 50 mL of iodinated nonionic contrast medium (Isovue 300, Squibb Diagnostics) is injected manually in 5 seconds or less through a central or peripheral line. Each cerebral hemisphere is chosen as a separate ROI, and the mean Hounsfield number in each region is plotted as a function of time and a global IW calculated from the gamma-variate fit. This is the technique used in the study of the head-injured patients (see Table 4⇓). The ROI excludes the major sulci and vessels, ventricles, and cisterns. Also, CBV maps are calculated on a pixel-by-pixel basis for tissue only as described below. The tissue concentration (or, CT enhancement)–vs-time curves for the large ROI or each pixel separately are fitted to a gamma-variate function as described above.
Calculation of the CBV map involves pixel-by-pixel multiplication of the CBF map and the IW map. The CBF map is obtained directly from the xenon/CT method. The IW map is derived by performing a gamma-variate fit on the dynamic bolus CT data on a pixel-by-pixel basis rather than on regions of pixels. Calculation of the IW map on a pixel-by-pixel basis is more difficult than IW calculations on regions of pixels. Pixel noise inherent in the CT imaging process is effectively filtered out when large numbers of pixels are included in a ROI. Thus, for these regions, the beginning and ending times of the bolus transit curve (ta and tf) for the purpose of fit are clearly defined. This is not the case for calculations on individual pixels. Accurate determinations of ta and tf are imperative if reliable values of IW are to be obtained for each pixel in the map. Since it is not practical for the operator to manually input ta and tf for the roughly 15 000 pixels in a typical map, an automated technique has been developed.11 The original 512×512 CT images are reduced to 256×256 by application of a 2×2 averaging kernel. This reduces the inherent image noise by a factor of 2. The data-fitting algorithm uses multiple passes through the data to determine ta and tf for each pixel and to produce optimal gamma-variate fits. Since we are primarily interested in CBF and CBV for brain tissue, at this stage vascular pixels are identified based on a weighted score for each pixel that includes peak times, peak height, and area under the curve. These vascular pixels are then eliminated from the data set, and the CT images are smoothed by applying three passes of a 3×3 gaussian filter. Extensive computer simulations have shown that this level of smoothing reduces inherent CT image noise to a standard deviation of less than 0.6 of a Hounsfield unit with an error in the computed IW and area of less than 3%. The data-fitting algorithm is then applied to the smoothed CT data. Again multiple passes through the data are used to determine accurate ta and tf values and produce optimal gamma-variate fits. The resulting parameter maps contain only tissue pixels. Finally, the IW and xenon/CT CBF maps are multiplied to produce the CBV map.
For the studies of validation of changes in CBV, patients were selected who had an intraventricular catheter in place, thus permitting ICP measurements and CSF withdrawal/injection for PVI measurements.6 These patients, whose ICPs were responsive to changes in Paco2, were still on respirators so that their CO2 levels could be manipulated, and for whom wide swings in ICP were judged to not be detrimental. Over a period of 3 years only two such patients were found, both middle-aged men who were in the subacute phase after severe head injury, with extremely poor prognoses, a ‘’Do-Not-Resuscitate“ status, and just before pulling the ventriculostomy tube. Informed consent had been obtained from the next of kin after explaining that these studies were not done for the benefit of that particular patient. For these two patients, CBV measurements were carried out before and after hyperventilation using the techniques described above. Simultaneous monitoring of ICP and measurements of the PVI allowed calculation of the resulting change in CBV from the pressure-volume curve. The results for these patients are shown in Table 1⇓. The measured enhancement-vs-time curves are shown in Fig 1⇑. In the hyperventilated state, the overall reduction in the area (height) indicates reduced blood volume. The broadening of the curves reflects increased vascular resistance and resultant reduced blood flow. Similar results were obtained for the other patient. Comparison of the computed CBV changes between the two techniques showed agreement within 15%.
An example of the calculated CBV map for a head-injured patient from the measured CBF and MTT maps is shown in Fig 3⇑. A ROI has been drawn over the frontal white matter and the caudate nucleus (gray matter). The measured values for these two regions are shown in Table 2⇓.
Since we are interested in CBF and CBV for brain tissue, vascular pixels need to be eliminated from the computed maps. This is done, as discussed earlier, by calculating a score for each pixel based on a combination of three values calculated from the gamma-variate parameters for each pixel. The calculated score represents the likelihood that a particular pixel is contained within a vessel or tissue. From this information a vessel map is generated depicting arteries and veins within the slice. Such a map is shown in Fig 4⇑ where the two shades of gray represent early arriving bolus (arteries, dark shade) and late arriving bolus (veins, bright pixels).
CBV in Normal Volunteers
Dynamic CT and xenon/CT CBF studies were conducted with a group of healthy normal volunteers on GE 9800 and Siemens CT Plus scanners that use the same CBF software and hardware (xenon delivery system). The studies performed on the GE scanner were analyzed using large ROIs and calculating global IWs. Due to the better time resolution available on our Siemens scanner, studies performed on this scanner were used to compute IW and CBV maps as described earlier. A total of 10 subjects were studied (7 males/3 females), with a mean age of 27±4 years. They were breathing spontaneously through a face mask, from which the end-expiratory CO2 (Peco2) was sampled. After a period of stabilization to get used to the face mask, consecutive CBF and dynamic studies were performed, and the measured values of CBF, IW, and CBV for the Siemens-based studies are shown in Table 3⇓. Also listed in Table 3⇓ are the CBF and CBV values calculated from the bolus dynamic CT alone. The ratio-of-areas technique was used to compute CBVdyn, with the hematocrit correction used by Gobbel et al.8 9 We assumed hematocrit values of 0.40 and 0.35, respectively, for males and females and a value of 0.76 for the ratio HT/HPA.12 The CBFdyn values were determined using the central volume principle: CBFdyn=CBVdyn/IW. The xenon-determined CBF values are plotted against the calculated CBFdyn in Fig 5⇑. The straight line representing the best fit through the data is described by the following equation, for which r=.94: The average absolute difference between CBFdyn and CBFXe is 10.4.
Seventy-one patients had early stable xenon/CT CBF and CBV measurements performed on a GE 9800 scanner following a diagnostic CT scan. The age of the patients varied from 15 to 82 years (average, 34 years). Patients with a Glasgow Coma Scale of 8 or less were studied immediately after stabilization in the Emergency Room or in follow-up studies. The time after injury at which these studies were performed, ranged from 45 minutes to 4 days. All patients were intubated, mechanically ventilated, and hemodynamically stabilized prior to examination. The results are shown in Table 4⇓. The data show that (1) CBF is reduced early after head injury; (2) cerebral transit times are significantly prolonged early after injury, indicating increased cerebrovascular resistance consistent with the presence of early ischemia; and (3) CBV, obtained from the product of CBF and MTT, is reduced from normal values, indicating narrowing of the microcirculation.
Patient With Hemodynamic TIAs
We have used these measurements once in making the decision to perform an extracranial-intracranial anastomosis in a rare patient with multiple hemodynamic TIAs. This 29-year-old man suffered a mild stroke after a minor trauma to the neck. Angiography showed total occlusion of the left internal carotid artery in the neck and only faint filling of the left hemisphere through the anterior communicating artery but not through the posterior communicating artery or the ophthalmic artery. For almost a year the patient suffered at least weekly TIAs consisting of aphasia and sometimes right hemiparesis lasting 1 to 60 minutes. When first seen by us he was neurologically normal except for slight anomia. CBF was 48 mL/100 g per minute and 42 mL/100 g per minute in the right and left hemispheres, respectively, with corresponding global CBV of 3.6 mL/100 g (right) and 4.0 mg/100 g (left). On the basis of these measurements it was felt that the left hemisphere indeed had low perfusion pressure for which it compensated by vasodilation, and so a left superficial temporal artery to middle cerebral artery branch anastomosis was performed. The patient moved away from our region, which precludes a follow-up exam, but in a telephone interview the patient stated that he has not had a repeat TIA in the year since the operation. We have not used this method in the more common patients with embolic TIAs.
Comparison with other CBV Methods
CBV measurements are presently carried out with PET and SPECT techniques that image the equilibrium distribution of an intravascular radioactive tracer. Kuhl et al13 reported quantitative tomographic measurements of regional CBV in humans using 99mTc-labeled RBC as a single-photon emitting tracer. They found CBV values ranging from 2 to 4 mL/100 g depending on anatomic location; higher values corresponded to regions of gray matter. More detailed determinations of regional CBV using 99mTc-RBC were published by the same group14 for a series of normal subjects and patients after head injury. The mean±SD for normal control subjects was 4.34±0.50 mL/100 g. A recent SPECT 99mTc-RBC study reports a CBV value of 3.2±0.3 mL/100 g for a series of 9 normal subjects.15 CBV measured by PET using 11C-RBC tracers yielded normal values of 4.2±0.4 mL/100 g16 and 4.3±0.41 mL/100 g.17 Recently, regional PET studies by Leenders et al in a group of 34 normal volunteer subjects produced CBV values of 5.2±1.4 mL/100 g for gray matter, 2.7±0.6 mL/100 g for white matter, and 4.7±2.0 mL/100 g for cerebellum.18 All of these studies used a cerebral-to–large vessel hematocrit ratio of 0.85 for correcting CBV. Sakai et al made independent determinations of cerebral hematocrit and CBV in a series of 10 normal volunteer subjects using sequential SPECT measurements of RBC and plasma volumes.12 They found a mean cerebral hematocrit of 31.3±1.8%, which amounted to 75.9±2.1% of the large-vessel hematocrit, resulting in a mean CBV of 4.81±0.37 mL/100 g.
The use of dynamic CT scanning after the administration of a bolus of iodinated contrast has been advocated over the years by many investigators as a rapid means of evaluating cerebral hemodynamics and possibly obtaining measures of CBF and CBV. The availability of faster CT scanners has renewed interest in this area, and preliminary results show promise for clinical application.8 9 19 20 21 Most of the reported work remains qualitative in nature, and few attempts have been made to systematically develop and validate this technique for routine application in the clinical arena. Gobbel et al have used an ultrafast CT scanner to measure regional CBF in dogs under varying physiological conditions.8 9 Their results showed significant correlation with simultaneous CBF measurements by the radiolabeled microsphere method. Steiger et al used dynamic scanning with a GE 9800 CT scanner to determine CBV and CBF in 13 normal human volunteers.20 Normal CBF and CBV were found to be 50±13 mL/100 g per minute and 5.8±1.2 mL/100 g per minute, respectively. These authors used a 50-mL bolus injection with a 4.5 second scanning resolution. No details are provided regarding the duration of the bolus injection, and their calculation of CBV is based on the ratio of peak CT values as opposed to the (correct) ratio of areas.8 9 Also, no correction for hematocrit is included.
Our method, depending on independent measurements of CBF and transit time, yields a normal CBV value of 5.3±0.4 mL/100 g, as deduced from calculated CBV maps for a standardized anatomic slice. This value is in general agreement with the values from the radionuclide approaches. Differences might arise from methodological problems associated with our method as described earlier or with problems inherent in the other techniques. It should be noted that our xenon CBF values obtained with a 5-minute xenon inhalation protocol may overestimate CBF by about 20%,22 leading to a corrected normal CBV value of 4.5 mL/100 g. Also, the selection of different anatomic slices with varying amounts of gray/white matter ratios further complicates direct comparison. Our studies with the normal volunteer subjects used a standardized anatomical slice that included the frontal, parietal, and occipital lobes, internal capsule tracts, the genu and splenium of corpus callosum, and basal ganglia structures such as caudate head and thalamus. Our global results with the normal volunteers in which we used the ratio-of-areas method yield a CBVdyn value of 4.4±0.8 mL/100 g in agreement with that derived from the independent measurements of CBF and MTT and corrected for the 5-minute CBF overestimation. In particular, the derived CBF values from the dynamic study are strongly correlated with our gold standard of xenon/CBF (r=.94) strengthening the possibility of deriving both CBV and CBF from a single dynamic study. However, as was emphasized earlier, this entails certain assumptions on the hematocrit values that might not be valid in the ischemic or traumatized brain. Also, the absence of a well-defined artery in our chosen slice and the possibility of partial volume averaging may lead to overestimations of CBVdyn. The regional results presented in Table 3⇑ demonstrate a ratio of approximately 2 to 1 for the CBV values of gray-to-white matter when the IW×CBF method is used in approximate agreement with the results from the ratio-of-areas method.
To assess the effect of the duration of the bolus injection, we also determined IW using MRI scanning, where 10 mL of gadolinium, which can be injected in 2 seconds, already provides adequate enhancement (authors’ unpublished results, 1997). With the shorter, 2-second, boluses in the MRI measurements, we found an IW of 6.0 seconds compared with an almost identical value of 6.1 seconds determined with a 5-second bolus in the CT scanner. This indicates that the IW substantially compensates for the finite-duration bolus effects. Clearly, more work is necessary to ascertain (1) the accuracy of the transit times measured by our proposed 5 s/50 mL bolus protocol and (2) the use of the ratio-of-areas approach to patients with severe head injury or cerebrovascular disease.
The theory and techniques of PVI measurements have been extensively described elsewhere.6 This method is primarily used to gain insight into intracranial compliance after severe head injury and in hydrocephalus and to predict which patients are prone to develop ICP problems. Yoshihara et al have also used it to calculate changes in blood volume brought about by changes in Paco2.23 With the PVI method itself no assumptions are necessary, but to derive the CBV changes per 100 g of brain tissue the total milliliter change must be divided by brain weight, which cannot be measured in vivo. We assumed an average brain weight of 1200 g. With this assumption our method yielded 11% higher values that the PVI-derived values, which in our opinion is adequate for physiological data of this nature. We were able to do these comparison studies in only 2 patients due to stringent criteria for patient safety. Most patients in whom these studies could safely be done do not have an intraventricular catheter any longer.
The application of this new method in studies of severely head injured patients is described more extensively elsewhere.24 The results for the head-injury patients presented in Table 4⇑ indicate significant prolongation of the transit times early after injury when compared with measurements at 89 hours postinjury. There is a clear trend of IW toward normal levels as time increases. This initial prolongation in IW indicates increased cerebrovascular resistance consistent with the presence of early ischemia.25 26 As seen in Table 4⇑, CBV in head-injured patients is reduced from normal values, reflecting narrowing of the microcirculation.
For patients with cerebrovascular disease, particularly TIAs, hemodynamic parameters have been investigated most extensively with PET scanning.2 3 4 5 Although the question as to which patients may benefit from extracranial-intracranial anastomosis is still not answered, it appears that certain features are favorable: lower CBF but with increased CBV and increased oxygen extraction fraction.4 5 The first two can now be assessed using the technique described in this paper, while oxygen extraction fraction can be measured using new MRI techniques.27 Our patient fit the criteria for a favorable response and indeed has had no more TIAs since the extracranial-intracranial anastomosis. Thus, measurement of CBV in the CT scanner should be explored further as an aide in clinical decision making in patients with TIAs.
In conclusion, the method described in this paper yields global, regional, and local CBV values that can be correlated with CBF in the same region. The values of normal CBV obtained with our method are in general agreement with other methods. Furthermore, any acute changes in CBV correlate well with the PVI-determined changes, indicating that the technique is well suited for application in individual cases in which we are mostly interested in whether CBV is below, at, or above normal. We have found the method to be easily applied after severe head injury, while preliminary data support its usefulness in patients with hemodynamic TIAs.
Selected Abbreviations and Acronyms
|CBF||=||cerebral blood flow|
|CBFdyn||=||dynamic cerebral blood flow|
|CBV||=||cerebral blood volume|
|CBVdyn||=||dynamic cerebral blood volume|
|MTT||=||mean transit time|
|PET||=||positron emission tomography|
|RBC||=||red blood cell|
|ROI||=||regions of interest|
|SPECT||=||single-photon emission computed tomography|
|ta/tf||=||beginning and ending times of the bolus transit curve, respectively|
|TIAs||=||transient ischemic attacks|
This work was supported by grant #5R01NS29412-03 from the National Institutes of Health and the Lind Lawrence Fund. A. John Kuta, MD, and S.J. Cochran, MS, have provided invaluable help in obtaining the CT scans in patients and volunteers.
Reviews of this manuscript were directed by Mark L. Dyken, MD.
- Received March 5, 1997.
- Revision received June 10, 1997.
- Accepted June 10, 1997.
- Copyright © 1997 by American Heart Association
Powers WJ, Raichle ME. Positron emission tomography and its application to the study of cerebrovascular disease in man. Stroke. 1985;16:361-376.
Gobbel GT, Cann CE, Fike JR. Measurement of regional cerebral blood flow using ultrafast computed tomography. Stroke. 1991;22:768-777.
Gobbel GT, Cann CE, Iwamoto HS, Fike JR. Measurement of regional cerebral blood flow in the dog using ultrafast computed tomography: experimental validation. Stroke. 1991;22:772-779.
Bouma GJ, Fatouros PP, Muizelaar JP, Stringer WA. Determination of cerebral blood volume by simultaneous xenon-enhanced computed tomography and dynamic CT enhanced by intravenous contrast. In: Yonas H, ed. Cerebral Blood Flow Measurement with Stable Xenon-Enhanced Computed Tomography. New York, NY: Raven Press; 1992:100-104.
Fatouros PP, Schröder ML, Muizelaar JP, Kuta AJ, Cothran SJ. Dynamic CT and cerebral blood volume. In: Tomonaga M, Yonas H, Tanaka A, eds. Quantitative Cerebral Blood Flow Measurements Using Stable Xenon/CT. Mount Kisko, NY: Futura Publishing Co; 1995:95-110.
Kuhl DE, Reivich M, Alavi A, Nyari I, Straum MM. Local cerebral blood volume determined by three-dimensional reconstruction of radionuclide scan data. Circ Res. 1975;36:610-619.
Sabatini U, Celsis P, Viallard G, Rascol A, Marc-Vergnes JP. Quantitative assessment of cerebral blood volume by single-photon emission computed tomography. Stroke. 1991;22:324-330.
Phelps ME, Huang SC, Hoffman EJ, Kuhl DE. Validation of tomographic measurement of cerebral blood volume with C-11-labeled carboxyhemoglobin. J Nucl Med. 1979;20:328-334.
Leenders KL, Perani D, Lammertsma AA, Heather JD, Buckingham P, Healy MJ, Gibbs JM, Wise RJ, Hatazawa J. Cerebral blood flow, blood volume and oxygen utilization: normal values and effect of age. Brain. 1990;113:27-47.
Nagata K, Asano T. Functional image of dynamic computed tomography for the evaluation of cerebral hemodynamics. Stroke. 1990;21:882-889.
Steiger HJ, Aaslid R, Stoos R. Dynamic computed tomographic imaging of regional cerebral blood flow and blood volume: a clinical pilot study. Stroke. 1993;24:591-597.
Toriyama T, Tanizaki Y, Hongo K, Osawa M, Kobayashi S. Functional image of dynamic computed tomography in diagnostic and prognostic evaluation of ischemic stroke within the first 6 hours. Stroke. 1993;24:1933-1944.
Fatouros PP, Schröder ML, Kuta AJ, Cothran SJ, Wist AO, Dalby PS, Muizelaar JP. Accuracy of low CBFs using the xenon/CT method: effect of xenon inhalation time. In: Tomonaga M, Yonas H, Tanaka A, eds. Quantitative Cerebral Blood Flow Measurements Using Stable Xenon/CT. Mount Kisko, NY: Futura Publishing Co; 1995:275-281.
Schröder ML, Muizelaar JP, Fatouros PP, Kuta AJ, Choi SC. Regional cerebral blood volume after severe head injury in patients with regional cerebral ischemia. Neurosurgery. (Submitted for publication)