To evaluate reproducibility of total cerebral blood flow (CBF) measurements with phase contrast magnetic resonance imaging (pcMRI).
To evaluate reproducibility of total cerebral blood flow (CBF) measurements with phase contrast magnetic resonance imaging (pcMRI).
We repeated total CBF measurements in 15 healthy volunteers with and without cardiac triggering, and with and without repositioning. In eight volunteers measurements were performed at two different occasions. In addition, measurement of flow in a phantom was performed to validate MR measurements.
A difference of 40.4 ml/minute was found between CBF measurements performed with and without triggering (P < 0.05). For repeated triggered measurements, the coefficient of variation (CV) was 7.1%, and for nontriggered measurements 10.3%. For repeated measurements with repositioning, the CV was 7.1% with and 11.2% without triggering. Repeated measurements at different occasions showed a CV of 8.8%. Comparing measured with real flow in the phantom, the triggered differed 4.9% and the nontriggered 8.3%.
The findings of this study demonstrate that pcMRI is a reliable method to measure total CBF in terms of both accuracy and reproducibility. J. Magn. Reson. Imaging 2002;16:1–5. © 2002 Wiley-Liss, Inc.
METHODS THAT CURRENTLY ARE USED to assess total cerebral blood flow (CBF) can be divided into two groups based on the underlying concepts. On the one hand, total CBF can be estimated based on information generated by flow in the capillaries; on the other hand, total CBF can be assessed by measuring flow in the supplying vessels of the brain. Methods that are based on measurements of flow in the capillaries are single positron emission computed tomography (SPECT), xenon-computed tomography (Xe-CT), and perfusion magnetic resonance imaging (MRI). Blood flow in the supplying vessels of the brain can be measured using Doppler ultrasound and phase contrast MRI (pcMRI). Advantages of both Doppler ultrasound and pcMRI are no need for using ionizing irradiation or administration of intravenous agents, and the possibility of repeated measurements on a short-term basis. However, limitations of Doppler ultrasound are its operator dependency and overestimation of total blood flow in a given vessel due to the fact that only the highest flow in the center of the vessel is assessed (1). Using pcMRI, total CBF can be assessed by simultaneously measuring flow in the internal carotid arteries and the basilar artery. As compared to Doppler ultrasound, pcMRI has the advantage of being operator independent and involving straightforward flow quantification. In addition, pcMRI can be added to morphologic MRI sequences, offering the option to correlate flow to morphology based on data generated during one examination.
In vitro and in vivo studies have demonstrated that pcMRI provides reliable flow data (2–4). However, data on short-term and long-term reproducibility are scarce. Furthermore, there is an ongoing discussion whether total CBF should be measured using a cardiac-triggered or a nontriggered pcMRI technique (4–6). Finally, using pcMRI, considerable differences in total CBF have been found between healthy volunteers (7, 8). These differences could be due to a lack of accuracy and reproducibility of the method, or they could be due to biologic variation.
The aim of this study was to assess the long-term and short-term reproducibility of total CBF measurements using triggered and nontriggered pcMRI techniques.
In the current study total CBF was defined as the sum of the flow in the left and right internal carotid arteries and in the basilar artery. To assess repeatability and short-term reproducibility of total CBF measurements, repeated total CBF measurements were performed in 15 healthy volunteers. To determine the effect of triggering on total CBF measurements, we performed measurements both with and without triggering. Triggered and nontriggered scans were made immediately after each other. To assess the repeatability of the measurements, a second acquisition was performed without changing the position of the subject and using the same scan plane. To assess short-term reproducibility, an effect that is bound to influence scans performed at different occasions, a third measurement was performed. Before the third measurement was performed, volunteers were moved out of the gantry, asked to sit upright and lie down again, and moved into the gantry again. Subsequently, a scan plane was chosen based on a new plan scan. To study the long-term reproducibility, which may be influenced by physiologic changes in patients as well as machine drift, total CBF measurements were performed two times at different occasions in eight other volunteers with a mean interval of 72 days (range, 7–101). The measurements were all performed by one researcher (A.S.) on the same MR system. In addition, we performed measurements on a phantom to externally validate our method of measuring the total CBF. To determine the influence of the postprocessing technique on the calculated total CBF, one observer (A.S.) performed postprocessing of all the acquisitions twice with a period of 8 months in between.
The volunteers were all healthy and had no previous history of disease. After explanation of the study protocol, each volunteer gave informed consent. The group of volunteers involved in the reproducibility study comprised eight women and seven men, with a mean age of 29 years (range, 21–48 years). The long-term reproducibility of scanning at different occasions was assessed in a second group of eight male volunteers with a mean age of 21 years (range, 18–26 years).
Flow curves were simulated with the same pump used by Frayne et al (9) and consisted of a single piston and cylinder driven by a stepping motor (UHDC flow simulator; University Hospital Development Corporation, London, Canada). This motor was controlled by a computer, which allows the generation of both constant-flow and arbitrary pulsatile-flow waveforms. The phantom consisted of two tubes that were connected via a branching system of three tubes of glass. The diameters of the three glass tubes were 9.0, 8.0, and 5.6 mm, respectively. The wall thickness of the glass tubes was 1 mm. These three glass tubes were each positioned in a polyvinyl chloride (PVC) tube. The space between the inner glass tube and the outer PVC tube was filled with water to obtain good contrast between the wall of the glass tube and its surroundings. We used two different waveforms provided by the manufacturer of the pump and resembling the flow in the carotid artery: one directly mimicking the physiological waveform, the other taking into account the compliance of the tubing. The maximum flow was set at 30 ml/second with average flows of 552 and 550 for the physiologic waveform and the waveform taking into account the tube compliance, respectively. This average was derived by the area under the curve of the flow profile and provided by the flow pump. A frequency of 75 per minute was used throughout the phantom study. We used a mixture of 60% H2O with 40% glycerin for fluid in the phantom (10).
All imaging was performed on a 1.5-T ACS-NT15 Philips MR system equipped with a Powertrak 6000 gradient system (Philips Medical Systems, Best, The Netherlands). A radio frequency spoiled gradient echo phase contrast technique (TR/TE = 16/9 msec, flip angle = 7.5°, slice thickness = 5 mm, field of view (FOV) = 250 × 188 mm, matrix = 256 × 154) with a velocity encoding of 100 cm/second was used for flow measurements (2). Both in the phantom and with the volunteers a birdcage coil was used. The phase-encoding was in the slice select direction. In the other directions, flow compensation was used. The scan plane was chosen perpendicular to the basilar artery and both internal carotid arteries. The number of signal averages (NSA) was one for triggered scans and eight for nontriggered scans. In the nontriggered acquisition, flow was averaged during several heart cycles. Gating was performed retrospectively with a peripheral pulse unit. Retrospectively, 16 heart phases were reconstructed. Total scan duration was 3 minutes and 5 seconds for the triggered sequence and 36 seconds for the nontriggered sequence.
Flow measurements were analyzed on a Sun UltraSparc 10 workstation with the internally developed software package FLOW® (Division of Image Processing, Department of Radiology, Leiden University Medical Center) (11). For this analysis a region of interest (ROI) was manually drawn around the vessel in the magnitude image by one observer (A.S.) and copied into the phase images. For triggered measurements a vessel contour was drawn in one heart phase and the software copied this contour to the other phases. Visually, all phases were screened for correct positioning of the ROI. If required, the ROI was adjusted. The flow volume is calculated by integrating the flow velocity values within the contour multiplied by the area. Flows in the basilar artery and the left and right internal carotid arteries were added and considered to represent the total CBF (ml/minute).
A Bland and Altman analysis (12) was performed to assess the agreement between the triggered and nontriggered measurements, the degree of reproducibility, and the influence of repositioning. A paired samples t-test was used to determine the statistical significance of the differences. The measured flow was compared with the flow provided by the flow simulator. Also, we compared the difference between the triggered and nontriggered measurements found in the volunteers with the differences of the triggered and nontriggered measurements found with the phantom. An intraclass correlation coefficient was calculated to assess reproducibility of repeated postprocessing.
A P value below 0.05 was considered significant. As a measure of relative variability, the coefficient of variation (CV) was calculated. This was defined as the SD divided by the mean.
Irrespective of the waveform, the flow measured in the phantom with triggered and nontriggered pcMRI was always lower than the true flow (552 or 550 ml/minute). The average difference between measured and real flow was 4.9% for the triggered scans and 8.3% for the nontriggered scans. The difference between triggered and nontriggered measurement was 3.4% ± 7.8%. The difference between the triggered and nontriggered measurement was not significant when assessed with a paired samples t-test (P = 0.24). The coefficient of variation for the repeatability of the measurement in the flow phantom was 3.4% and 9.0% for the triggered and nontriggered sequences, respectively.
In all volunteers, one scan plane could be found that permitted acquisition perpendicular to both carotid and basilar arteries. On T2-weighted images no abnormalities were found in any volunteer's brain.
Table 1 shows the statistics of the triggered and nontriggered total CBF measurements; differences between the first triggered and first nontriggered measurements were statistically significant (P < 0.05). The data are presented in a Bland and Altman plot in Figure 1. On average, total CBF values generated by triggered measurements were smaller than those generated by nontriggered measurements (40.4 ml/minute, 5.7%, P = 0.037). The SD of the differences was 67.9 ml/minute.
|Subject||Gender||Age||Triggered measurements||Non-triggered measurements|
|I (ml/min)||II (ml/min)||III (ml/min)||I (ml/min)||II (ml/min)||III (ml/min)|
The data in Table 1 also show the repeatability and reproducibility of the triggered and nontriggered total CBF measurements. There was no significant difference between the first and second triggered measurements (rescanning without repositioning, repeatability) and the first and third triggered measurements (rescanning with repositioning, short-term reproducibility) (P = 0.97 and P = 0.87, respectively). The mean difference and SD between the first and second triggered measurement was 0.76 ml/minute (±73 ml/minute). The mean difference between the first and third triggered measurement was 3.12 ml/minute, and the SD of the differences 73 ml/minute. Figure 2 shows the plots of the Bland and Altman analysis of the first and second triggered measurements. This figure also illustrates that there is no systematic difference between the first and second triggered measurements and between the first and third triggered measurements. CVs were 7.1% between the first and second triggered measurements, and 7.1% between the first and third measurements. The nontriggered measurements also showed no significant difference between the first and second and first and third measurements (P = 0.88 and P = 0.86, respectively). The mean difference and SD between the first and second nontriggered measurements were 4.6 and 112 ml/minute, respectively (CV = 10.3%). The mean difference between the first and third nontriggered measurement was 5.6 ml/minute, and the SD of the differences 121 ml/minute (CV = 11.2%).
In Figure 3 the Bland and Altman plot of the measurements of the eight volunteers at two different points in time is plotted. The average time between the two occasions was 72 days, with a range of 7–101 days. The two measurements did not differ significantly (P = 0.21). The mean difference was 43.4 ml/minute, with an SD of the differences of 88.7 ml/minute; the CV was 8.8%. The data for the long-term reproducibility are shown in Table 2. Repeated postprocessing of the first total CBF measurements revealed a very good agreement, with an intraclass correlation coefficient of 0.97.
|No. subject||First measurement (ml/min)||Second measurement (ml/min)||Gender||Age|
The findings of this study demonstrate that pcMRI is a reliable method to measure total CBF in terms of both accuracy and reproducibility. Flow measured with pcMRI did not differ more than 11% from real flow in the phantom. Using pcMRI, the CV of repeated total CBF measurements was 11% or less, irrespective of using cardiac triggering or repositioning of the subject, or whether the repeat scan was performed during the same session or on a different occasion.
Although both methods proved reliable, we found better reproducibility in triggered pcMRI measurements than in nontriggered measurements. In our phantom experiments, both methods proved equally accurate. However, in vivo measurements yielded a lower total CBF when triggered pcMRI was used. This discrepancy between in vitro and in vivo observations is probably related to biologic variations in heart frequency. Furthermore, our in vivo findings contrast with observations in two other studies showing lower CBF values using nontriggered pcMRI (4). The artificially high CBF values that were found in these studies using triggered pcMRI were explained by Tarnawski et al by selective gating during systole (4). Possibly, this selective gating did not occur in our experiments because we used retrospective gating. Whatever the cause, the change between triggered and nontriggered measurements in our hands originated from pixels in the periphery of the vessel. When only the flow of the central pixels was analyzed, the difference between triggered and nontriggered measurement almost completely disappeared. Apparently, averaging flow information over the heart cycle gives rise to a loss of information in the periphery of the vessels, resulting in artificially high flow values. The cause of this phenomenon remains to be elucidated.
Under physiological conditions, total CBF is independent of perfusion pressure and is kept relatively constant within individuals by local vasomotor adjustments in cerebral vascular resistance (13). This cerebral autoregulation mechanism guarantees a tight coupling between oxygen supply and demand of the brain. In this study, total CBF was found to be constant within the 95% confidence intervals (651–813 ml/minute for the long-term measurements) within individuals, which is in line with the knowledge that total CBF is kept relatively constant (13). However, as in previous studies, a considerable variety of total CBF was found between individuals, and the variance in total CBF we found in this study is similar to that observed by others (8). Probably, interindividual variation in basal total CBF reflects differences in ways (anatomy) and effectivity of oxygen transport between subjects (14).
In conclusion, pcMRI is an accurate and reproducible method to measure total CBF noninvasively in humans. Therefore, the variation of total CBF that is found between healthy individuals does not result from a lack of reproducibility of pcMRI. The amount of variation in measured total CBF values we encountered in this study in the same healthy individuals, small as it was, should be taken into account when designing and interpreting studies using pcMRI-based total CBF estimates.