| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the Departments of Radiology and Medical Physics, University of Wisconsin, 600 Highland Ave, Madison WI 53792. Received May 12, 2003; revision requested July 2; revision received October 10; accepted November 17. Address correspondence to V.H. (e-mail: vmhaughton@wisc.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: A cardiac-gated phase-contrast MR technique was used to acquire images at 14 time points evenly spaced through the cardiac cycle in 10 volunteers and eight patients with Chiari I malformation. Instantaneous CSF velocities were displayed as temporal and spatial plots and examined for homogeneity, differences between flow anterior and flow posterior to the spinal cord, synchronous bidirectional flow, and equivalence of the caudad flow and craniad flow in each voxel. Indexes for flow homogeneity, synchronous bidirectional flow, and preferential flow direction were calculated, and differences between the patient and volunteer groups were tested for significance with a t test of the means.
RESULTS: In volunteers, diastolic velocity peaked in two regions in the anterior paramedial subarachnoid space. Patients had greater inhomogeneity of flow than volunteers. They had substantially increased flow (jets) in the anterior paramedial locations. Synchronous bidirectional flow was seen in six of the patients and in none of the volunteers. Cephalad flow in the jets or nodes (P = .05), proportion of cephalad and caudad flow in the anterior compartment (P < .005 for both), and the fraction of voxels with flow directionality (P = .03) differed significantly between patients and volunteers.
CONCLUSION: CSF flow in symptomatic patients with Chiari I malformation, unlike that in volunteer subjects, is characterized by flow jets, regions with a preponderance of flow in one direction, and synchronous bidirectional flow.
© RSNA, 2004
Index terms: Brain, abnormalities, 152.1473, 1538.1473 • Brain, MR, 127.12144, 1532.12144 • Cerebrospinal fluid, flow dynamics, 167.12144 • Cerebrospinal fluid, MR, 167.12144
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Theoretically, the presence of tonsils in the foramen magnum alters cerebrospinal fluid (CSF) flow patterns, especially during those portions of the cardiac cycle with maximal flow into (cephalad flow) or from (caudad flow) the cranial vault. Numerous investigators have documented abnormal CSF flow patterns in the foramen magnum in patients with Chiari I malformation (1–9). Still to be determined, however, are the parameters that differentiate normal from abnormal CSF flow as well as the possible relationships between these abnormal flow patterns and signs and symptoms in patients. Most published studies have involved recording CSF flow in the foramen magnum over relatively large regions of interest or the entire subarachnoid space with the implicit assumption that CSF flow is spatially uniform, but CSF flow or velocity averaged over a large region of interest may not reveal local regions within the subarachnoid space with large spatial gradients in velocity (jets). Also, the measurement of a peak velocity may underestimate the velocity in such a flow jet and fail to reveal the temporal duration or the spatial extent of the jet.
Given the lack of detailed understanding regarding spatial variations in CSF flow in the foramen magnum, the purpose of our study was to measure spatial and temporal variations in CSF flow through the cardiac cycle and throughout the subarachnoid space at magnetic resonance (MR) imaging in volunteers with no known neurologic or spinal problems and in patients with Chiari I malformation.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MR Imaging Data Collection
Each patient and volunteer underwent MR imaging (with a 1.5-T MR imaging unit [Advantage; GE Medical Systems, Milwaukee, Wis]), at which T1- and T2-weighted sagittal images of the cervical spine were obtained to evaluate for syringomyelia and tonsillar descent into the foramen magnum. The anatomic images in the volunteers and patients were evaluated by a neuroradiologist (V.H.) for evidence of a posterior fossa malformation. All images were considered technically adequate, without evidence of motion artifact. Tonsillar herniation exceeding 5 mm was seen in each patient. The foramen magnum, cervical spine, and posterior fossa were considered normal in each volunteer.
Phase-contrast MR images of the foramen magnum were acquired in the transverse plane when the subject had achieved a steady heart rate. A commercially available phase-contrast flow sequence was used (flip angle, 20°; repetition time msec/echo time msec, 20/5; section thickness, 5 mm; field of view, 180 mm; matrix, 256 x 256; encoding velocity, 10 cm/sec). The acquisition was gated prospectively by means of triggering from the R wave of the chest electrocardiogram. In volunteers the section was chosen at the foramen magnum and in patients it was chosen at the tip of the tonsils as follows: In the volunteers, the section location for phase-contrast MR imaging was chosen by the technologist to be perpendicular to the spinal axis and coincident with the undersurface of the occipital bone at the foramen magnum. In the patients, the section location for phase-contrast MR imaging was chosen by the technologist to be at the inferior edge of the tonsil and perpendicular to the spinal canal. Fourteen image frames were acquired at regular time intervals to fill the R-R interval. The offset velocity, estimated from the velocity of stationary tissue, was used to correct for phase shifts introduced by eddy currents. The spatial resolution within each image was 0.7 mm per pixel. For an annular CSF space between the dura and the spinal cord of approximately 5 mm at the foramen magnum, this resolution is adequate to study spatial patterns in transverse CSF flow. Phase-contrast MR data from the 18 subjects were considered technically adequate. No data sets were discarded for motion or other technical deficiencies.
Flow Analysis
The analysis of CSF flow patterns (M.F.Q.) focused on two basic parameters: velocity and volume flow rate. In terms of velocity, we studied the component in the superoinferior directions as a function of space within the subarachnoid space and time within the cardiac cycle. To study spatial variations in superoinferior velocity, we considered the flow structure known as a jet in fluid mechanics. A jet consists of a spatially concentrated region of high velocity that is embedded in a surrounding region of lower velocity. Physically, the volume flow rate (Q) through a surface refers to the volume of fluid flowing through the surface per unit time. Volume flow rate for each voxel is the velocity times the area of the voxel. For any region in the subarachnoid space, volume flow rate (Qreg) is the sum of the flow rates for each voxel in the region. The volume flow rate for the entire subarachnoid space (Qt) is the sum of the volume flow rates for each voxel.
Initially, voxels representing subarachnoid space were identified with a computer program (developed in-house and based on Interactive Display Language, Research Systems, Boulder, Colo) that chose voxels having a cephalad flow velocity of at least 0.2 cm/sec during one part of the cardiac cycle and craniad flow in another part of the cardiac cycle. The spinal cord was excluded by disregarding the region within the canal having velocities less than 0.2 cm/sec throughout the cardiac cycle. A neuroradiologist (V.H.) verified the anatomic correctness of the computer-chosen voxels.
Further analysis was performed to identify those voxels exhibiting aliasing or confounding effects of arterial or venous flow. Arterial or venous flow was assumed to have monotonically positive or negative cumulative flow. Aliasing was assumed to be characterized by abrupt changes in apparent direction (ie, a change in velocity sign from positive to negative or vice versa). Voxels exhibiting arterial or venous contamination were removed from the voxels assumed to represent CSF flow. Voxels exhibiting aliased velocities were corrected with the algorithms of Lee et al (2). Velocities and flow volumes were corrected for the effects of aliasing before the analysis of flow patterns.
For the temporal and spatial analysis of flow, programs were developed (by M.F.Q.) in Interactive Display Language (Research Systems), a scientific data language oriented toward manipulation and analysis of graphics. The spatial distribution of CSF velocities was displayed for each point in the cardiac cycle with a color plot and with a surface contour plot in which the z axis encoded the velocity for each voxel. The temporal pattern of velocity and cumulative flow were displayed with a time-course plot for each voxel.
To check for systematic errors due to incomplete sampling of the cardiac cycle, volume flow rates in the entire subarachnoid space were calculated and compared for systole and diastole. Since uneven sampling of the R-R interval would result in less complete sampling during diastole than during systole, the difference between systolic and diastolic volume flow rates was measured. Differences less than 20% were considered evidence that the cycle was adequately sampled. In all volunteers and patients, the differences were less than 20%.
Analysis of Flow Jets
Investigators (V.H., M.F.Q., and B.I.) inspected the color maps and the surface contour maps in the volunteers and patients to determine the relative homogeneity of flow and the presence of bidirectional flow. In this study, jets were identified by reference to the color and contour plots of CSF velocities. When velocities in one region of the subarachnoid space exceeded those in the adjacent region by 100%–200%, the region was characterized as a node, and if the disparity in velocity was even greater, the region was classified as a jet. Using Interactive Display Language (Research Systems), one investigator (M.F.Q.) calculated the fraction of the entire subarachnoid space occupied by the nodes or jets and the proportion of total flow that was present in high-velocity jets or nodes.
Comparison of Flow in Anterior versus Posterior Subarachnoid Space
So that we could compare the proportion of flow in anterior versus posterior compartments, a line was placed horizontally on the transverse image at the widest diameter of the spinal cord. Voxels were classified as being in either the anterior or the posterior subarachnoid space in reference to the line. The volume flow rate in each region was computed. The ratios of anterior to posterior flow in the cephalad direction and anterior to posterior flow in the caudad direction were calculated. Finally, the flow rate through regions identified as nodes or jets was computed and compared with the total flow rate in the subarachnoid space.
Analysis of Synchronous Bidirectional Flow
Each color image was inspected to detect evidence of simultaneous cephalad and caudad flow. When two or more consecutive images demonstrated evidence of both caudad flow and cephalad flow with velocities greater than 0.5 cm/sec, they were interpreted as showing simultaneous bidirectional flow in the subarachnoid space. The relative location of these voxels was noted, and the difference in velocity (shear) between adjacent or nearby voxels with opposite flow directions was calculated. The number of voxels with evidence of synchronous bidirectional flow was also noted.
Analysis of Preferential Flow Direction
To study variations in flow rate over the cardiac cycle, we defined the cumulative flow through a voxel as the sum of the flow volumes in the voxel at each of the preceding time points. A cumulative flow volume near zero at the end of an R-R interval was assumed to indicate nearly equal flow in both caudad and cephalad directions, while a positive or negative value at the end of the cardiac cycle was assumed to indicate a net caudad or cephalad flow in the voxel. An index for preferential flow direction was calculated as the proportion of all CSF voxels that had a cumulative flow in one direction of at least three times that in the other direction at the end of the cardiac cycle.
Statistical Testing
To test the significance of differences in indexes between the patients and the volunteers, a t test of the means was performed, with P 
.05 considered to indicate a statistically significant difference.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Analysis of Flow Jets
In the 14 images acquired during the cardiac cycle in each volunteer, color plots showed spatial variations in the magnitude of CSF velocity in the subarachnoid space (Fig 1). The surface contour plots effectively demonstrated the variations in velocity in different regions of the subarachnoid space (Fig 2). The spatial variations in flow were displayed more effectively in color images than in black and white images. In each volunteer, two regions, or nodes, of increased velocity and flow were evident in the anterior subarachnoid space near the midline. These regions were regular in shape, symmetrical around the midline, and roughly equal from side to side.
|
|
In patients, spatial and temporal variation in CSF flow velocities was much more evident than in the volunteers (Figs 3, 4). Increased flow was evident in the anterior nodes, which occupied, on average, 14% of the subarachnoid space. In four patients, diastolic velocities in the anterior nodes exceeded 4.0 cm/sec, and these regions were referred to as jets. Unlike nodes in the volunteers, the nodes and jets in patients tended to be asymmetric and irregular in shape. Whereas in volunteers, maximal systolic and diastolic velocities occurred in roughly congruent regions of the subarachnoid space, in patients, maximal systolic and diastolic velocities occurred in distinctly different areas. In patients, maximal diastolic velocities occurred in the anterior nodes, as in volunteers, but maximal systolic velocities occurred elsewhere.
|
|
Comparison of Flow in Anterior and Posterior Subarachnoid Spaces
In volunteers, differences in flow volume between the anterior and posterior subarachnoid spaces were small. Overall, the most cephalad and caudad flow in volunteers occurred in the anterior compartment (57% and 55%, respectively). Five of seven patients had large posterior areas in which there was little or no flow (Figs 3, 4). In patients, the average fractions of cephalad and caudad flow in the anterior compartment were 72% and 81%, respectively. The proportion of cephalad and caudad flow in the anterior compartment differed significantly between volunteers and patients (P = .002 and P = .003, respectively).
Synchronous Bidirectional Flow
Simultaneous cephalad and caudad flow was evident in six patients but in no volunteers (Fig 5). Such bidirectional flow was recognized in three or more consecutive images (20%–25% of the cardiac cycle), mainly at the transition from cephalad to caudad or from caudad to cephalad flow. In all cases, bidirectional flow was characterized by large cephalad velocities in the anterior nodes or jets and lower caudad velocities in the adjacent regions. Regions exhibiting such caudad flow were termed counterjets. In pronounced cases of bidirectional flow, large velocity differences in adjacent subregions (shear) reached 9.0 cm/sec (7.5 cm/sec in the jet and –1.5 cm/sec in the adjacent counterjet). Shear velocities of more than 4.0 cm/sec were present in all patients but in none of the volunteers.
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
CSF movement in the caudad or cephalad direction through the foramen magnum cannot be characterized as plug flow. CSF flow in the foramen magnum is highly complex both spatially and temporally, even in healthy subjects. Velocities of and inhomogeneities in flow were greater during the phase of cephalad (diastolic) flow than during the phase of caudad (systolic) flow in both volunteers and patients. This higher cephalad velocity suggests relatively greater pressure differentials during the cephalad flow of CSF than during caudad flow and greater fluctuations in CSF pressures in the spinal canal than in the cranial vault. This observation may help explain the greater incidence of spinal cord cysts compared with brainstem cysts in Chiari I malformation.
Previous investigators have reported increased CSF velocities in the subarachnoid space of patients with Chiari I malformation (1–3), as we observed in our study. Heiss et al (4), for example, using data acquired with a sagittal plane at phase-contrast MR, found peak velocities averaging 11 cm/sec. Previous investigators (1–9) have described temporal variation in CSF flow, but they have not emphasized the marked variations in flow with location in the subarachnoid space.
In this study, indexes of CSF flow were defined to test the statistical significance of differences between patients and control subjects. The indexes were not optimized for diagnosing flow abnormalities. Additional study is needed to determine the most effective criteria to distinguish normal from abnormal flow. Such criteria may in fact require more sophisticated image acquisitions. Evaluation of CSF flow in the foramen magnum may require multisection or volume acquisitions to measure flow at multiple levels and multiple phase-encoding directions to measure the direction of flow accurately and completely. Smaller voxels may be needed to achieve highly accurate velocity measurements.
Synchronous bidirectional CSF flow in the foramen magnum has been noted previously in patients with Chiari I malformation (10). A spatial variation in the time at which flow reverses would not explain the presence of bidirectional flow over as much as 25% of the cardiac cycle. Confounding effects of arterial or venous flow have been excluded. Since bidirectional flow appears in close proximity to regions with large jets, it seems to be related to countercurrents resulting from jets.
The limitations of this study were due mainly to the restricted amount of flow data acquired in each subject and patient. CSF velocities may not have been maximal in the section selected for measuring velocity. The sections in the volunteers were located at the foramen magnum, where maximal velocities were anticipated. Those in patients were located below the tip of the tonsil, where a larger subarachnoid space was present for sampling flow. Additional studies are needed to determine if CSF velocities may have been higher at the foramen magnum in these patients. The differences in section location may have diminished the differences in CSF velocities between the control and patient groups.
Since flow was measured only in the superoinferior direction and since some CSF flow though the foramen magnum undoubtedly occurs oblique to the superoinferior direction, velocities within the selected transverse section were likely underestimated. Effects of blood flowing in arteries and veins, which may confound the measurement of CSF velocity, were probably minimized in this study.
Another confounding effect may be the incomplete sampling of velocities throughout the cardiac cycle. When the 14 images do not effectively fill the entire cardiac cycle, data are lost. Incomplete sampling of the cardiac cycle would result in the loss of CSF velocity data during diastole. Since total caudad and cephalad flows differed in most cases by a few percentage points and in no cases by more than 30%, the portion of the cardiac cycle not sampled was assumed to be small. Finally, the number of patients and control subjects was limited. No pediatric cases were included because normative data were not acquired in pediatric control subjects. The patients were not a homogeneous group with respect to age, sex, or symptoms. Multiple other factors affect the accuracy of CSF velocity calculations (5).
The important finding of this study is that CSF flow abnormalities in Chiari I malformations can be severely underestimated or completely undetected when flow is averaged temporally or spatially. A minority of voxels in the subarachnoid space in patients display elevated velocities. When elevated velocities are found in jets, countercurrent or modest velocities are found in other regions. Measurements of global velocity or flow (ie, flow averaged over the entire subarachnoid space) are unlikely to reflect the severity of flow abnormalities in patients. CSF flow data that do not account for flow inhomogeneity may therefore represent an underestimation of the complexity and velocity of CSF flow in Chiari I malformation.
In conclusion, the findings of this study show markedly nonlaminar CSF flow in the foramen magnum in volunteers and, to a greater extent, in patients with symptomatic Chiari I malformation. The complexity of flow may affect the development of signs and symptoms in patients and skew the measurement of flow with standard phase-contrast MR imaging methods. This complexity may explain some discrepancies between the results of previously reported studies. More work is needed to characterize the flow of CSF in the foramen magnum. To characterize flow accurately, flow data must be acquired in multiple sections or as a volume, in multiple directions, and with the smallest possible voxels.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, M.F.Q.; study concepts and design, V.H., B.I., M.F.Q.; literature research, V.H.; clinical studies, B.I.; data acquisition, V.H.; data analysis/interpretation, M.F.Q., M.N.; statistical analysis, M.F.Q.; manuscript preparation, all authors; manuscript definition of intellectual content, V.H., B.I., M.F.Q.; manuscript editing, M.A.Q., M.N.; manuscript revision/review, V.H., M.A.Q., M.F.Q.; manuscript final version approval, V.H., M.F.Q.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. K. Hofkes, B. J. Iskandar, P. A. Turski, L. R. Gentry, J. B. McCue, and V. M. Haughton Differentiation between Symptomatic Chiari I Malformation and Asymptomatic Tonsilar Ectopia by Using Cerebrospinal Fluid Flow Imaging: Initial Estimate of Imaging Accuracy Radiology, November 1, 2007; 245(2): 532 - 540. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Ley, C. Fink, M. Puderbach, J. Zaporozhan, C. Plathow, M. Eichinger, W. Hosch, K.-F. Kreitner, and H.-U. Kauczor MRI Measurement of the hemodynamics of the pulmonary and systemic arterial circulation: influence of breathing maneuvers. Am. J. Roentgenol., August 1, 2006; 187(2): 439 - 444. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
