Patients with multiple sclerosis with structural venous abnormalities on MR imaging exhibit an abnormal flow distribution of the internal jugular veins.
ABSTRACT To evaluate extracranial venous structural and flow characteristics in patients with multiple sclerosis (MS).
Two hundred subjects with MS from two sites (n = 100 each) were evaluated with magnetic resonance (MR) imaging at 3 T. Contrast-enhanced time-resolved MR angiography and time-of-flight MR venography were used to assess vascular anatomy. Two-dimensional phase-contrast MR imaging was used to quantify blood flow. The MS population was divided into two groups: those with evident internal jugular vein (IJV) stenoses (stenotic group) and those without (nonstenotic group).
Of the 200 patients, 136 (68%) showed IJV structural abnormalities, including unilateral or bilateral stenoses at different levels in the neck (n = 101; 50.5%) and atresia (n = 35; 17.5%). The total IJV flow normalized to the total arterial flow of the stenotic group (56% ± 22) was significantly lower than that of the nonstenotic group (77% ± 14; P < .001). The arterial/venous flow mismatch in the stenotic group (12% ± 15) was significantly greater than that in the nonstenotic group (6% ± 12; P < .001). The ratio of subdominant venous flow rate (Fsd) to dominant venous flow rate (Fd) for the stenotic group (0.38 ± 0.27) was significantly lower than for the nonstenotic group (0.59 ± 0.23; P < .001). The majority of the stenotic group (67%) also had an Fsd of less than 3 mL/s, a Fd/Fsd ratio greater than 3:1, and/or a total IJV flow rate of less than 8 mL/s.
MR imaging provides a noninvasive means to separate stenotic from nonstenotic MS cases. The former group was more prevalent in the present MS population and carried significantly less flow in the IJVs than the latter.
Patients with Multiple Sclerosis with Structural
Venous Abnormalities on MR Imaging Exhibit an
Abnormal Flow Distribution of the Internal
E. Mark Haacke, PhD, Wei Feng, PhD, David Utriainen, BS,
Gabriela Trifan, MD, Zhen Wu, MD, Zahid Latif, RT,
Yashwanth Katkuri, MS, Joseph Hewett, MD, and David Hubbard, MD
Purpose: To evaluate extracranial venous structural and flow characteristics in patients with multiple sclerosis (MS).
Materials and Methods: Two hundred subjects with MS from two sites (n ? 100 each) were evaluated with magnetic resonance
(MR) imaging at 3 T. Contrast-enhanced time-resolved MR angiography and time-of-flight MR venography were used to assess
vascular anatomy. Two-dimensional phase-contrast MR imaging was used to quantify blood flow. The MS population was divided into
two groups: those with evident internal jugular vein (IJV) stenoses (stenotic group) and those without (nonstenotic group).
Results: Of the 200 patients, 136 (68%) showed IJV structural abnormalities, including unilateral or bilateral stenoses at different
levels in the neck (n ? 101; 50.5%) and atresia (n ? 35; 17.5%). The total IJV flow normalized to the total arterial flow of the stenotic
group (56% ? 22) was significantly lower than that of the nonstenotic group (77% ? 14; P ? .001). The arterial/venous flow
mismatch in the stenotic group (12% ? 15) was significantly greater than that in the nonstenotic group (6% ? 12; P ? .001). The
ratio of subdominant venous flow rate (Fsd) to dominant venous flow rate (Fd) for the stenotic group (0.38 ? 0.27) was significantly
lower than for the nonstenotic group (0.59 ? 0.23; P ? .001). The majority of the stenotic group (67%) also had an Fsd of less than
3 mL/s, a Fd/Fsd ratio greater than 3:1, and/or a total IJV flow rate of less than 8 mL/s.
Conclusions: MR imaging provides a noninvasive means to separate stenotic from nonstenotic MS cases. The former group was
more prevalent in the present MS population and carried significantly less flow in the IJVs than the latter.
CCA ? common carotid artery, CCSVI ? chronic cerebrospinal venous insufficiency, CE ? contrast-enhanced, EJV ?
external jugular vein, Fd ? dominant venous flow rate, Fsd ? subdominant venous flow rate, FIJV ? IJV flow, Fta ? total
arterial flow, LIJV ? left internal jugular vein, IJV ? internal jugular vein, MIP ? maximum-intensity projection, MS ?
multiple sclerosis, PC ? phase-contrast, RIJV ? right internal jugular vein, 3D ? three-dimensional, TOF ? time of flight,
2D ? two-dimensional, Venc ? encoding velocity
Recent advances in multiple sclerosis (MS) suggest that
there may be an association between impaired venous flow
and the disease (1–5). Ultrasound (US) imaging (1,6–8)
and magnetic resonance (MR) imaging (9–12) have been
used to study these venous abnormalities. Each method has
its strengths and weaknesses, but the three-dimensional
(3D) information available to MR imaging might provide a
more complete, or at least complementary, picture versus
From the Departments of Biomedical Engineering (E.M.H., W.F.) and Radiol-
ogy (E.M.H., Z.L., Y.K.), Wayne State University; Magnetic Resonance Inno-
vations Inc. (E.M.H., D.U., G.T., Z.W.) and Magnetic Resonance Imaging
Institute for Biomedical Research (E.M.H.), 440 E. Ferry St., Unit 2, Detroit, MI
48202; Synergy Health (J.H.), Costa Mesa; Applied Functional Magnetic
Resonance Imaging Institute (D.H.), San Diego, California; and Department of
Electrical and Computer Engineering (E.M.H.), McMaster University, Hamil-
ton, Ontario, Canada. Received April 28, 2011; final revision received Sep-
tember 15, 2011; accepted September 17, 2011. Address correspondence
to E.M.H.; E-mail: firstname.lastname@example.org
E.M.H. is the president of Magnetic Resonance Innovations Inc. (Detroit, Michigan).
None of the other authors have identified a conflict of interest.
Appendices I–VII are available online at www.jvir.org.
© SIR, 2012
J Vasc Interv Radiol 2012; 23:60–68
that presented with US, and it could also serve as an
excellent means to judge anatomy and flow simultaneously.
In the present study, data are presented for 200 patients with
MS recruited from two sites (100 subjects from each site)
who underwent conventional MR imaging as well as what
is referred to as the chronic cerebrospinal venous insuffi-
ciency (CCSVI) protocol. This protocol collects data for
vascular anatomy and flow in the major veins in different
parts of the neck and brain. Combining these two pieces of
information provides a powerful means to study vascular
anomalies in MS. The goal of this study is to highlight the
major venous anomalies as seen with MR imaging and
report the flow characteristics, especially flow through the
internal jugular veins (IJVs), in this population. This infor-
mation might prove useful in better stratifying patients with
abnormal venous flow and perhaps prove useful in treat-
MATERIALS AND METHODS
A conventional clinical protocol for MS was combined
with a specially designed CCSVI protocol. The conven-
tional protocol included a T1-weighted scan before and
after contrast agent administration, a two-dimensional
(2D) fluid-attenuated inversion recovery scan, and a T2
scan when cerebrospinal fluid had to be visualized. The
CCSVI protocol consisted of the following vascular se-
quences: a coronal 3D time-resolved contrast-enhanced
(CE) MR arteriovenography scan, a transverse 2D time-of-
flight (TOF) MR venography scan, and a 2D phase-contrast
(PC) MR scan for flow quantification (sequence parameters
detailed in Appendix I [available online at www.jvir.org]).
Deidentified data from 200 patients with MS (100 from
each of two sites) who underwent the CCSVI MR imaging
protocol were evaluated. Of the 200 patients, the MS type
was available for 133 patients: primary progressive MS (n ?
34), secondary progressive MS (n ? 31), and relapsing/
remitting MS (n ? 68). Institutional review board approval
was obtained to perform the quantitative evaluation pre-
sented in this article.
Data were collected at site 1 on a 3-T Tim Trio scanner
(Siemens, Erlangen, Germany) with a 12-channel head/
neck coil arrangement and at site 2 on a 3-T Signa HDxt
scanner (GE Healthcare, Milwaukee, Wisconsin) with an
eight-channel head/neck coil arrangement. The subject was
centered at the orbital ridge for the brain scans. T1, fluid-
attenuated inversion recovery, and, when needed, T2 data
were acquired. After these scans, the subject was moved to
the center at the chin in preparation for contrast medium
injection. Just before injection, 2D TOF MR venographic
data were collected. Contrast medium was then injected
(OptiMARK [Covidien, Hazelwood, Missouri] at site 1 and
Magnevist [Bayer, Wayne, New Jersey] at site 2, both with
0.2 mL/kg [0.1 mmol/kg]), and time-resolved MR angio-
graphic data were collected for 20 time points. Then, 2D PC
MR scans were obtained at three different levels in the neck
(C2/C3, C5/C6, and T1/T2). Finally, the subject was recen-
tered at the orbital ridge and a postcontrast T1 image
matching the precontrast image was collected. Parallel im-
aging was used to speed up data acquisition by a factor of
two when possible.
Data were processed by using Signal Processing in Nuclear
MR software (SPIN, Detroit, Michigan) to evaluate vessel
morphology. The time-resolved 3D CE MR arteriovenog-
raphy data were viewed as original and subtracted data (ie,
venous phase minus arterial phase) in 3D. Poor flow indi-
cated by persistent contrast enhancement was cross-exam-
ined with 2D TOF MR venography and PC MR imaging.
Anatomic assessment of data involved the identification of
stenosis, aplasia, and atresia of the IJVs.
The 2D TOF MR venographic images were used to
quantify any narrowings perceived in the coronal views of
the 3D CE MR arteriovenography images because it pro-
vided higher in-plane resolution for cross-sectional mea-
surements. IJVs that showed a cross-sectional area of less
than 25 mm2(ie, one third of the cross-sectional area for an
average IJV diameter of 1 cm [13,14] assuming a circular
shape) in the lower-level C5/C6 or T1/T2 levels were
considered stenotic. For upper-level narrowing (C2/C3), a
cross section of less than 12.5 mm2was considered to
represent a stenotic IJV. Observed stenoses were then cat-
egorized into groups according to their location: upper and
lower neck levels, unilateral on the right IJV (RIJV) or left
IJV (LIJV), or bilateral stenosis. At either neck level, any
measured cross-sectional area greater than the stenosis
threshold was considered to represent a nonstenotic result.
Atresia of an IJV was identified when the vessel came
to a clear and abrupt ending at some point in the IJV body
between the sigmoid sinus and its confluence with the
subclavian vein. Time-resolved 3D CE MR arteriovenog-
raphy data and 2D TOF MR venography were used to
confirm the lack of structural patency of a vessel. This
could appear in the form of a terminating sigmoid sinus
with the IJV being reconstituted from the vertebral system
or as a lower-level truncation with a thin string down one
side. Aplasia, a congenital condition in which an IJV fails
to develop entirely, could potentially be identified by the
complete lack of ability to image a complete IJV in any of
the data collected.
The vessel lumen for veins and arteries seen in PC MR
images were delineated by using the magnitude and phase
images as input. A baseline correction was applied to re-
move any phase shift caused by eddy currents and gradient
imperfections (Appendix II, available online at www.jvir.
org). In the case that phase aliasing was present, phase
unwrapping was performed (Appendix III, available online
at www.jvir.org). The corrected and unwrapped phase val-
ues in the vessels were then mapped to velocity measure-
ments by using the encoding velocity (Venc) value. Subse-
Volume 23 ? Number 1 ? January ? 2012 61
quently, quantitative measurements such as cross-sectional
area, flow velocity (average, peak positive, and peak neg-
ative), volume flow rate (average, positive, and negative),
total flow volume per cardiac cycle, and reflux were com-
puted (as defined in Appendix IV, available online at
www.jvir.org). Blood flow velocities through all the dis-
cernable veins and arteries were measured. These typically
included the following, on the left and right sides: common
carotid arteries (CCAs), vertebral arteries, IJVs, external
jugular veins (EJVs), vertebral veins, deep cervical veins,
anterior jugular veins, and posterior EJVs. Because the
venous system is known to exhibit significant variations,
other veins were also included when encountered.
To evaluate the flow differences between different
groups, paired and unpaired t tests were performed as
appropriate. A significance level of P ? .05 was used to
determine whether the measurements between groups were
The data from both sites were processed by trained
processors by using in-house software. Inter- and intrapro-
cessor variability were assessed by using a multiple-way
analysis of variance with a randomized complete block
design (Appendix V, available online at www.jvir.org). A
significance level of P ? .05 was used to determine whether
the inter- and intraprocessor variability was significant by
using the null hypotheses that the interprocessor factor
effect and the intraprocessor factor effect equal 0 (15).
Subjects were categorized into two groups: stenotic and
nonstenotic. Those subjects who showed one or more of the
structural abnormalities (stenosis, atresia, and aplasia) were
assigned to the stenotic group, whereas those who did not
were assigned to the nonstenotic group. Anatomic assess-
ment for all 200 subjects is presented in the Table. Note
that there were more patients from site 2 with stenotic IJVs
than from site 1 (80 vs 56). Of the 200 patients with MS,
68% were determined to be in the stenotic group. Examples
of these abnormalities are presented in Figures 1–4.
Atresia was identified in 25 cases at the upper neck
level and in 10 cases at the lower neck level inferior to
where the common facial vein typically joined the IJV,
making a total of 35 cases of atresia. Examples of superior
and inferior atresia of the IJVs are shown in Figure 1. A
thin strip connected the truncated ends of the atresia in 14
of the 35 cases. Atresia in a vessel was sometimes coupled
with stenosis above the atresia or after the vessel had
become reconstituted through collateral flow from vessels
such as connections from the vertebral plexus, facial veins,
EJVs, and thyroid veins. Stenosis and atresia were present
in the same IJV in 14 cases. Aplasia was not observed in
Stenoses were observed as narrowing in caliber in the
upper and lower neck levels. They could be caused by
ectatic carotid arteries or compression by other tissue as
shown in Figure 2a and 2b. Although some stenoses ap-
peared as region-specific narrowings, others were diffuse
with stenotic caliber observed through the entire length of
the IJV (Fig 2c, 2d). Figure 3 shows a 3D CE MR arte-
riovenography coronal maximum intensity projection with
an apparent stenosis and the transverse 2D TOF MR venog-
raphy images from which the cross-sectional areas of the
vessel lumen were determined. Significant collateral flow
pathways were observed in some cases with structural ab-
normalities in one or both of the IJVs. Examples of these
highly developed collateral vessels are shown in Figure 4.
The high degree of variability in the veins of the neck was
evidenced with vessels such as the deep cervical vein,
anterior jugular vein, EJV, and posterior EJV carrying
Examples of abnormal flow patterns, including circulatory
flow in an IJV, reflux in an IJV, low flow rates in all veins,
and only one dominant vein are shown in Figure 5.
The key quantitative flow measurements are detailed in
Appendixes VI and VII (available online at www.jvir.org).
Major findings were as follows. The blood flow through the
left CCA was found to be similar to that through the right
CCA for the population of site 1 (P ? .44) and greater than
that through the right CCA for the population of site 2 (P ?
.04). The flow through the RIJV was significantly greater
than that through the LIJV (P ? .001 for both sites), which
is consistent with the findings of others (8). The average
percentages of total IJV blood flow normalized to total
arterial flow (Fta) for the two sites were ?67% ? 21 at site
1 and ?58% ? 23 at site 2. In comparison, the average
percentages of IJV blood flow normalized to total venous
flow were 70% ? 20 at site 1 and 66% ? 22 at site 2. The
Table. Anatomic Assessment of Patients with MS
(N ? 200)
1 (n ? 100)
2 (n ? 100)
Unilateral lower neck
Unilateral upper neck
Bilateral lower neck
Bilateral upper neck
Note.—Values in parentheses are percentages. Patients may
have multiple forms of stenosis, and therefore the sum of
individually counted numbers for all subcategories is greater
than the total number of patients with stenosis.
62 ? Abnormal IJV Flow Distribution in MS with Venous AbnormalitiesHaacke et al ? JVIR
arterial/venous mismatches, which reflect how much ve-
nous blood was flowing out of the brain through small veins
and collateral veins, were measured at 6% ? 12 and 13% ?
15, respectively. The ratios of subdominant venous flow
rate (Fsd) to dominant venous flow rate (Fd) were 0.52 ?
0.24 and 0.52 ? 0.26, respectively. Reflux flow percentages
were found to be 4% ? 5 and 2% ? 3, respectively, for the
LIJV and 2% ? 2 and 1% ? 1, respectively, for the RIJV.
Of interest in the study of stenosis is the fact that the vessel
cross-sectional area (measured from vessel lumen contours
drawn on PC MR imaging data) was found to have more
variability than the vessel flow rate: 59 ? 38 mm2and 47 ?
39 mm2, respectively, for the LIJV and 78 mm2? 47 and
57 mm2? 52, respectively, for the RIJV.
To further assess the flow characteristics for the MS
populations, all subjects from both sites were combined and
divided into stenotic and nonstenotic groups. Several flow
measures suggested that there were significant differences
various forms of atresia for IJVs. Before the MIPs were created, images of the early arterial phase were subtracted from images of the
strongest venous phase to remove the arterial signal for better observation of the veins. (a) In a patient from site 1, a 96-slice MIP
shows that the RIJV is malformed at the midneck level (long arrow) with the right sigmoid sinus draining into the vertebral
plexuses (short arrow). (b) In a patient from site 1, 96-slice MIP shows that both IJVs are truncated at the upper neck (long arrows)
level with the sigmoid sinuses draining into the vertebral plexuses (short arrows). The IJVs received branches from the vertebral
plexuses and the common facial veins near the midneck level, which reconstituted caliber size through the lower neck level. (c) In
a patient from site 2, 64-slice MIP shows that the LIJV is truncated at the midneck level (long arrow) with a thin connection
between the midneck level and the inferior jugular bulb near the confluence with the subclavian vein (short arrow). (d) In a patient
from site 2, 84-slice MIP shows that both IJVs show continuous enhancement from the sigmoid sinus through the upper neck
level but are truncated near the midneck level (long arrows). A thin connection is seen in the LIJV between the truncated end at
the midneck level and the inferior jugular bulb (short arrow). The RIJV shows no connection between the midneck level and the
Coronal maximum-intensity projections (MIPs) of time-resolved 3D CE MR arteriovenography data from four patients show
observed in MS. (b–d) Arterial signals were removed by subtracting the arterial phase. (a,b) In a patient from site 1, 96-slice MIP shows
an example of carotid ectasia in which the left internal carotid artery is compressing the LIJV. The subtracted data (b) show the
compression more clearly. (c) In a patient from site 1, 96-slice MIP indicates that the LIJV shows diffuse stenosis at all neck levels (long
arrow) whereas the RIJV shows stenosis at the lower neck level (short arrow). (d) In a patient from site 2, 76-slice MIP shows string-like
stenosis in the LIJV (long arrow) with asymmetrical enhancement of the vertebral plexuses on the affected side (short arrow). The
latter enhancement might be considered as a representation of the development of collateral flow caused by restricted flow in the
LIJV. (In this case the LIJV carried a flow of only 1.31 mL/s while the RIJV carried a flow of 7.94 mL/s.)
Coronal MIPs of time-resolved 3D CE MR arteriovenography data show various forms of stenoses of the IJVs (arrows)
Volume 23 ? Number 1 ? January ? 201263
between the two groups. Although the Fta rates were not
significantly different (15.00 mL/s ? 2.95 for the stenotic
group and 14.84 mL/s ? 2.73 for the non-stenotic group;
P ? .71), the total IJV flow (FIJV) rates normalized by Fta
rate for the stenotic group were significantly lower than
those for the nonstenotic group (56% ? 22 and 77% ? 14,
respectively; P ? .001). Similarly, the total cross-sectional
areas of all the arteries were not significantly different
between the two groups (96 mm2? 21 for the stenotic
group and 91 mm2? 18 for the non-stenotic group; P ?
localized to show the appearance of tight stenosis on different imaging methods. (a) MIP (96 slices) of 3D MR arteriovenography data
with the arterial signals removed by subtracting the arterial phase in a patient from site 1. (b–e) Axial 2D TOF MR venography data
reveals the cross-sectional area information in the transverse view for all major veins in the neck corresponding to the lines shown
in a. Arrows indicate the LIJV in each transverse image. The LIJV shows a stenosis at the lower neck level (b) with a cross-sectional
area of 22 mm2and a more severe stenosis near the midneck level (c) with a cross-sectional area of 6 mm2. The left common facial
vein drains into the LIJV superior to the severe stenosis (d). This case demonstrates that a tight stenosis that showed little or no
enhancement on 3D CE MR arteriovenography can show a clear signal and caliber on the high-resolution axial 2D TOF MR venography
data, in part because of the inherent sensitivity of 2D TOF MR venography to very slow flow.
Coronal MIP of time-resolved 3D CE MR arteriovenography data with corresponding axial 2D TOF MR venography slices
other than the IJVs showed large calibers (short arrows). Arterial signals were removed by subtracting the arterial phase to better view
venous structures. (a) In a patient from site 1, 96-slice MIP indicates that the LIJV shows a narrow caliber near the sigmoid sinus and
a narrow caliber near the confluence with the subclavian vein. The right and left EJVs (short arrows) show large calibers compared
with the LIJV. (b) In a patient from site 1, 96-slice MIP indicates that the LIJV shows atresia at the upper neck level with the sigmoid
sinus draining into the vertebral plexus. The RIJV shows stenosis at the upper neck level. The left EJV and the right deep cervical vein
(short arrows) show large caliber and strong enhancement. (c) In a patient from site 2, 72-slice MIP indicates that the LIJV shows a
narrow caliber near the confluence with the subclavian vein. The left anterior jugular vein shows a large caliber (short arrow). (d) In
a patient from site 2, 68-slice MIP indicates that the RIJV shows atresia at the upper neck level (long arrow) with the sigmoid sinus
draining into the vertebral plexus. The right and left EJVs (short arrows) show large caliber and strong enhancement, with the left EJV
having a dilated caliber.
Coronal MIPs of time-resolved 3D CE MR arteriovenography data show collateral venous development whereby vessels
64 ? Abnormal IJV Flow Distribution in MS with Venous AbnormalitiesHaacke et al ? JVIR
.10), but the total IJV cross-sectional area normalized by
the total arterial cross-sectional area was significantly
smaller for the stenotic group than for the nonstenotic group
(107% ? 181 and 186% ? 78, respectively; P ? .001). The
arterial/venous mismatch for the stenotic group was signif-
icantly larger than for the nonstenotic group (12% ? 15 and
6% ? 12, respectively; P ? .0059), meaning that more
venous blood drained through small collateral veins. The
Fsd/Fd ratio was significantly lower for the stenotic group
as well (0.38 ? 0.27 for the stenotic group and 0.59 ? 0.23
for the nonstenotic group; P ? .001). For patients with IJV
reflux flow, reflux through the LIJV and RIJV were not
significantly different between the two groups (LIJV, 4% ?
6 for the stenotic group and 2% ? 3 for the nonstenotic
group [P ? .23]; RIJV, 2% ? 2 and 2% ? 2, respectively
[P ? .72]).
Figure 6 shows a scatterplot of flow rates of the two
veins with the highest flow. It is seen that many of the
patients with stenotic IJVs had an Fsd rate of less than 3
mL/s, a combined flow rate for the two dominant veins of
less than 8 mL/s, and/or an Fsd/Fd ratio of less than 1/3
(which had been shown to have implications for cerebral
vascular problems in craniotomies ). Integrating all
these conditions captures 67% of patients with structural
abnormalities, whereas the remaining region captures 70%
of those with no structural abnormalities.
from two sites. Blue and red correspond to site 1 and site 2,
respectively. Based on the FIJV normalized by the correspond-
ing Fta, diamonds, circles, and squares correspond to type I
(FIJV/Fta ratio ? 2/3), type II (2/3 ? FIJV/Fta ratio ? 1/3), and type
III flow (FIJV/Fta ratio ? 1/3) (6). Solid symbols are cases of MS
with stenosis (upper- and lower-level stenosis, bilateral stenosis,
diffuse stenosis, and atresia). The light gray shade under the
negative diagonal dashed line represents cases with a sum of Fd
and Fsd values of less than 8 mL/s. The horizontal medium gray
shade covers cases with an Fsd value of less than 3 mL/s. The dark
gray shade under the solid line covers cases with an Fsd/Fd ratio
of less than 1/3. The dashed line along the positive diagonal line is
the equivalence line for the Fsd and Fd values. As regions overlap,
the shades change accordingly.
Scatterplot of Fsd and Fd for 100 MS patients each
with solid and dotted lines, respectively), (b) reflux during part of cardiac cycle through the LIJV, (c) missing RIJV with all other
veins (not shown to avoid clutter) carrying low flow (? 2 mL/s; also note low pulsatility in LIJV), and (d) RIJV representing the
only dominant vein with all other veins carrying low flow (? 2 mL/s). All veins had low pulsatility. Note that, in b–d, the average
flow rates are plotted.
Examples of abnormal flow patterns: (a) circulatory flow through both IJVs (positive and negative flow components plotted
Volume 23 ? Number 1 ? January ? 201265
Inter- and Intraprocessor Variability
The three-way analysis of variance showed that neither
inter- nor intraprocessor variability was significant for the
FIJV test (P ? .53 for interprocessor variability and P ?
.18 for intraprocessor variability) and stenosis test (P ? .62
and P ? .39, respectively). The interclass correlations for
intraprocessor variability between the two days for the three
processors were 0.99, 0.97, and 0.91 for flow measurements
and 0.99, 0.99, and 0.99 for stenosis measurement.
The present work was initiated in an attempt to demonstrate
the presence of CCSVI in patients with MS, and, more
specifically, to evaluate what type of venous structural and
flow abnormalities occur. The use of MR angiography and
flow quantification for the study of the vascular system in
patients with MS is new. To date, the role of abnormal
extracranial venous flow as having a major association with
MS has been controversial (17,18). In the present work, we
have seen evidence of a variety of structural abnormalities,
including atresia, string-like stenoses, and local stenoses.
Some recent studies discussed the case of MR imaging for
the evaluation of jugular venous anomalies and comparing
patients with MS and normal volunteers (11,12,19). The
findings of these studies suggest that MR anatomic infor-
mation by itself may be insufficient to allow strong conclu-
sions to be drawn about how the presence of venous ab-
normalities relates to the role of CCSVI in MS. The present
work does not make such a comparison, but rather com-
bines the use of MR-assessed anatomy and flow in subcat-
egorizing the MS population.
By combining anatomic assessment with the flow
quantification measurements, we have shown that most
patients with low FIJV rates were patients who had stenotic
IJVs. In comparing sites 1 and 2, we found that the popu-
lation from site 2 had lower FIJV rates (normalized by Fta
rates), smaller IJV cross-sectional areas (normalized by
total arterial cross-sectional area), and greater arterial/ve-
nous mismatches compared with the population from site 1.
This could be because more of the subjects from site 2 (80
of 100) had stenotic IJVs compared with those from site 1
(56 of 100). We also found that the reflux flow percentage
for the RIJV for the site 2 subjects was significantly higher
than for site 1 subjects. The fact that the arterial measure-
ments between these two populations were not significantly
different demonstrates that there was a change in the ve-
nous flow pattern for the patients with MS with stenosis.
Finally, an arterial/venous mismatch increase indicated that
more venous blood drains out through smaller veins (ie,
vertebral plexus) that could not be measured with PC MR
Based on a previously described categorization method
(6), whereby subjects were divided into three categories
based on FIJV normalized by Fta, of the 100 patients each
from site 1 and site 2, 56.0% and 37.0%, respectively, were
classified as type I (ie, FIJV/Fta ratio ? 2/3), 37.0% and
50.0%, respectively, were classified as type II (2/3 ? FIJV/Fta
type III (FIJV/Fta ratio ? 1/3; Fig 7a). Figure 7b and 7c show
the same categorization in nonstenotic and stenotic cases
from both sites. The similarity of FIJV distribution found
between the nonstenotic group and the normal group pre-
sented by Doepp et al (6) may suggest that the MS popu-
lation showed venous insufficiency mainly in the form of
lower FIJV, and therefore these patients carry more flow in
their collateral veins. The fact that the FIJV measurements
on Doppler US and MR imaging were normalized by the
corresponding arterial measurements removed most, if not
all, of the intermodality bias between them.
Low flow in one jugular and reflux in the jugular
veins have negative implications. Low flow in one jug-
ular vein can be a risk factor for cerebral vascular prob-
lems in craniotomies (16), and reflux has been associated
with a number of other neurologic conditions such as
premature aging (8), transient global amnesia (20), and
optic neuritis (21).
There are a number of limitations to the present study.
No data from normal control subjects were collected with
MR imaging. Although a comparison between this study
and the findings from a US imaging study was described
earlier, a statistically sound comparison between the two
could not be made. Had we collected MR data on normal
control subjects, we may have found that they too could be
broken into two similar categories: those with apparent
structural abnormalities and those without (22). This will be
a subject of future research. Other research has shown that
the diameter of the RIJV is significantly larger than that of
the LIJV (13,14), which was validated in the present study.
Although we used a conservative estimate of IJV diameter
to determine stenosis, the thresholds we used did not
distinguish between the LIJV and RIJV. In addition, no
threshold was set to assess the dilation of the IJVs above
the confluence with the subclavian vein, which was ob-
served in some cases and may be considered a structural
abnormality. As the data were collected by two different
clinical sites with different MR scanners, the absolute
flow measurements might contain systematic bias. The
arterial venous mismatch measurement may not be per-
fect, as the decision of which vessels were too small for
flow quantification was subject to the processors’ discre-
tion. Finally, the accuracy of the baseline correction of
the PC MR phase values depends on the region of
interest the processor drew, as well as the signal-to-noise
ratio of the acquired images.
Compared with other MR studies that used PC MR
imaging (9,10), in which Venc values of 70 cm/s and 80
cm/s were used for flow quantification, a much smaller
Venc of 50 cm/s was used in the present study. A smaller
Venc value gives a better signal-to-noise ratio, but can lead
to phase aliasing for fast blood flow that exceeds the Venc
value. The phase-unwrapping algorithm was capable of
removing the phase aliasing for velocities of less than 85
66 ? Abnormal IJV Flow Distribution in MS with Venous AbnormalitiesHaacke et al ? JVIR
cm/s. In rare cases, the flow velocity in the CCAs could
exceed this value, and the phase aliasing was not fully
corrected and resulted in smaller arterial flow measure-
In conclusion, we have shown that there were venous
flow abnormalities in patients with MS. However, to un-
derstand the role MR imaging would play in the assessment
of patients with MS, much more data will be required from
normal control subjects by using the same MR imaging
techniques, and patients will need to be followed before and
after treatment to objectively assess changes in flow and
their effects in patient recovery after percutaneous translu-
Li, and Areen Al Bashir for their contributions in processing
the data, and Shuang Feng, Alexander Korostelev, and Ana
Daugherty for their valuable discussions of statistical
1. Zamboni P, Galeotti R, Menegatti E, et al.
nous insufficiency in patients with multiple sclerosis. J Neurol Neurosurg
Psychiatry 2009; 80:392–399.
Chronic cerebrospinal ve-
et al (6). Type I flow represents an FIJV/Fta ratio greater than 2/3; type II flow represents a ratio between 2/3 and 1/3, and type III
represents a ratio lower than 1/3. The numbers for the normal population are from Doepp et al (6). (b) Distribution of patients with MS
with no apparent stenosis (n ? 44 from site 1 and n ? 20 from site 2) compared with the 50 normal control subjects described by Doepp
et al (6). (c) Distribution of patients with MS with clear stenosis (n ? 56 from site 1 and n ? 80 from site 2). NST ? patients with no
stenosis, ST ? patients with stenosis.
(a) Distribution of 200 MS patients from two sites and a normal population according to the criteria presented by Doepp
Volume 23 ? Number 1 ? January ? 201267
2. Singh AV, Zamboni P.
tion in multiple sclerosis. J Cereb Blood Flow Metab 2009; 29:1867–
3. Zamboni P, Galeotti R, Menegatti E, et al.
study of endovascular treatment of chronic cerebrospinal venous insuf-
ficiency. J Vasc Surg 2009; 50:1348–1358.
4. Zamboni P, Menegatti E, Galeotti R, et al.
venous haemodynamics in the assessment of multiple sclerosis. J Neu-
rol Sci 2009; 282:21–27.
5. Zamboni P, Menegatti E, Weinstock-Guttman B, et al.
chronic cerebrospinal venous insufficiency in patients with multiple scle-
rosis is related to altered cerebrospinal fluid dynamics. Funct Neurol
6. Doepp F, Schreiber SJ, von Mu ¨nster T, Rademacher J, Klingebiel R,
Valdueza JM.How does the blood leave the brain? A systematic
ultrasound analysis of cerebral venous drainage patterns. Neuroradiology
7. Doepp F, Paul F, Valdueza JM, Schmierer K, Schreiber SJ.
cervical venous congestion in patients with multiple sclerosis. Ann Neu-
rol 2010; 68:173–183.
8. Chung CP, Lin YJ, Chao AC, et al.
changes with aging. Ultras Med Biol 2010; 36:1776–1782.
9. Sundstro ¨m P, Wåhlin A, Ambarki K, Birgander R, Eklund A, Malm J.
Venous and cerebrospinal fluid flow in multiple sclerosis: a case-control
study. Ann Neurol 2010; 68:255–259.
10. Stoquart-Elsankari S, Lehmann P, Villette A, et al.
MRI study of physiologic cerebral venous flow. J Cereb Blood Flow
Metabol 2009; 29:1208–1215.
11. Zivadinov R, Galeotti R, Hojnacki D, et al.
detection of internal jugular vein anomalies in multiple sclerosis: a pilot
longitudinal study. AJNR Am J Neuroradiol 2011; 32:938–946.
12. Zivadinov R, Lopez-Soriano A, Weinstock-Guttman B, et al.
venography for characterization of the extracranial venous system in
Anomalous venous blood flow and iron deposi-
A prospective open-label
The value of cerebral Doppler
The severity of
Jugular venous hemodynamic
Value of MR venography for
Use of MR
patients with multiple sclerosis and healthy control subjects. Radiology
13. Furukawa S, Nakagawa T, Sakaguchi I, Nishi K.
internal jugular vein studied by autopsy. Rom J Leg Med 2010; 18:125–
14. Tartiere D, Seguin P, Juhel C, Laviolle B, Malledant Y.
diameter and cross-sectional area of the internal jugular veins in adult
patients. Crit Care 2009; 13:R197.
15. Montgomery DC. Design and analysis of experiments, 5th edition: New
York: Wiley, 2000.
16. Seoane E, Rhoton AL Jr.Compression of the internal jugular vein by
the transverse process of the atlas as the cause of cerebellar hemor-
rhage after supratentorial craniotomy. Surg Neurol 1999; 51:500–505.
17. Khan O, Filippi M, Freedman MS, et al.
insufficiency and multiple sclerosis. Ann Neurol 2010; 67:286–290.
18. Siskin GP, Haskal ZJ, McLennan G, et al.
agenda for evaluation of interventional therapies for chronic cerebrospi-
nal venous insufficiency: proceedings from a multidisciplinary research
consensus panel. J Vasc Interv Radiol 2011; 22:587–593.
19. Zaharchuk G, Fischbein NJ, Rosenberg J, Herfkens RJ, Dake MD.
Comparison of MR and contrast venography of the cervical venous
system in multiple sclerosis. AJNR Am J Neuroradiol 2011; 32:1482–
20. Chung CP, Hsu HY, Chao AC, Sheng WY, Soong BW, Hu HH.
global amnesia: cerebral venous outflow impairment--insight from the
abnormal flow patterns of the internal jugular vein. Ultras Med Biol 2007;
21. Hsu HY, Chao AC, Chen YY, et al.
venous flow in transient monocular blindness. Ann Neurol 2008; 63:247–
22. Zivadinov R, Marr K, Cutter G, et al.
ificity of chronic cerebrospinal venous insufficiency in MS. Neurology
The diameter of the
Estimation of the
Chronic cerebrospinal venous
Development of a research
Reflux of jugular and retrobulbar
Prevalence, sensitivity, and spec-
68 ? Abnormal IJV Flow Distribution in MS with Venous Abnormalities Haacke et al ? JVIR
APPENDIX II: BASELINE CORRECTION FOR FLOW MEASUREMENTS
Although the PC sequence was built to remove phase shifts caused by many practical imaging imperfections, eddy currents and
transient effects could induce phase shifts in the final phase images. This can be considered as a baseline shift and needs to be
properly removed. To determine the baseline, the processor drew at least three regions of interest in the muscle area close to the
major vessels. The baseline phase values were calculated as the spatial average phase values inside these regions of interest for each
time point during the cardiac cycle. The baseline corrected phase value inside the vessels was achieved by subtracting these baseline
phase values at corresponding time points.
APPENDIX III: UNWRAPPING OF FLOW ALIASING
With the chosen Venc of 50 cm/s, there was no aliasing in the veins for a majority of the cardiac cycle. There might be phase
aliasing in a few time points during the cardiac cycle in the arteries (carotid arteries and vertebral arteries) and the IJVs in systole,
when the blood flow was the fastest. However, the fastest blood flow rarely exceeded 100 cm/s, which means that the resulting phase
wrapping can be corrected with a phase unwrapping algorithm. We implemented such an algorithm that examines the phase value
inside the vessel during the whole cardiac cycle. For the arteries (veins), any phase value that was less (greater) than 0.3? (?0.3?)
was increased (decreased) by 2?, assuming that the nominal arterial phase value is positive. The choice of 0.3? (?0.3?) guarantees
the unwrapping of velocities as high as 85 cm/s and avoids incorrect unwrapping of reversed flow with a velocity less than 15 cm/s.
APPENDIX IV: FLOW MEASUREMENTS
Flow measurements included the cross-sectional vessel lumen area, the volume flow rate, the distribution of arterial and venous
blood flow in different vessels, the mismatch between arterial and venous flow (Equation 1), the ratio of blood flow between the
two dominant veins that carry the most amount of venous blood (Equation 2), and reflux flow for the IJVs (Equation 3). Note that,
because a large portion of the cases show no reflux and some cases exhibit very low FIJV, cases with zero reflux flow or with
individual FIJV less than 2 mL/s were excluded when calculating the reflux measures.
Equation 1: ?Arterial flow - Venous flow?/Arterial flow?100
Equation 2: Fsd/Fd
Equation 3: ?Reverse flow?/?Forward flow? ? 100
APPENDIX I: MR Sequence Parameters
Site 1 (n ? 100) Site 2 (n ? 100)
Sequence 3D CE MR
340 ? 255
384 ? 288
0.9 ? 0.9
2D TOF MR
320 ? 256
512 ? 204
0.63 ? 0.63
2D PC MR
256 ? 256
448 ? 448
0.57 ? 0.57
3D CE MR
340 ? 255
288 ? 224
0.59 ? 0.59
2D TOF MR
220 ? 160
288 ? 192
0.43 ? 0.43
2D PC MR
160 ? 160
256 ? 256
0.63 ? 0.63
Flip angle (°)
No. of partitions (Nz)
FOV (mm ? mm)
Imaging matrix (Nx ? Ny)
Resolution (mm ? mm)
Slice thickness (mm)
iPAT (phase encoding)
Reconstructed no. of images
Note.—An arterial saturation band with a width of 40 mm and a separation of 10 mm from the excited slice was applied during 2D
TOF MR venographic acquisition for both sites. Also note that the technicians carrying out the scans might have changed some of
the parameters slightly for better image quality or patient constraints depending on the situation. However, such changes will not
significantly affect the results presented in this paper. CE ? contrast-enhanced, FOV ? field of view, NA ? not applicable, iPAT ?
integrated parallel acquisition technique, PC ? phase-contrast, 3D ? three-dimensional, TOF ? time of flight, 2D ? two-
dimensional, Venc ? encoding velocity. Nx represents the number of sampled points in the readout direction and Ny represents
the number of phase encoding steps.
Volume 23 ? Number 1 ? January ? 2012 68.e1
APPENDIX V: EXPERIMENTAL DESIGN FOR INTER- AND INTRAPROCESSOR
To assess the inter- and intraprocessor variability for both FIJV measurements and IJV cross-sectional area measurements, the
following statistical tests with a randomized complete block design were performed. Four patients were randomly selected from all the
patients (two from site 1 and two from site 2) for variability in FIJV measurement. Another four patients were randomly selected from
patients with apparent stenosis determined from 3D CE MR arteriovenography data for variability in stenosis measurement. In these four
patients, two had upper-level (ie, C2/C3) stenosis and the other two had lower-level (ie, C5/C6) stenosis. The flow data and the 2D TOF
MR venography data were processed by three randomly chosen processors four times independently in 1 day. The same processors then
repeated the processing on the same cases after 2 weeks. The FIJV rate as a percentage of the Fta rate was calculated for each processing
to assess variability in the flow measurements. The cross-sectional area of the RIJV was measured on the 2D TOF MR venography data
at the corresponding C2/C3 or C5/C6 levels. The cross-sectional area was calculated at each stenotic site to assess variability in the
measurements. A three-way analysis of variance was performed with the three factors being processor, patient, and processing day. The
processor factor represents interprocessor variability and the processing day factor represents intraprocessor variability.
APPENDIX VI: Flow Measurements Table for 200 MS Patients from 2 Separate Sites (100 Subjects from Each Site)
Heart rate (/s)
Flow rate (mL/s)
Reflux LIJV (%)
Reflux RIJV (%)
Site 1 (n ? 100)
46.76 ? 10.31
70.61 ? 9.49
Site 2 (n ? 100)
48.81 ? 10.87
71.10 ? 12.92
?3.97 ? 2.56
?6.36 ? 2.84
6.22 ? 1.36
6.30 ? 1.27
1.72 ? 0.71
1.47 ? 0.62
?10.33 ? 3.40
?14.66 ? 2.62
15.73 ? 2.75
?3.21 ? 2.28
?5.05 ? 2.93
5.73 ? 1.26
5.92 ? 1.32
1.34 ? 0.72
1.18 ? 0.62
?8.25 ? 3.51
?12.34 ? 3.29
14.17 ? 2.81
?25.56 ? 16.21
?40.99 ? 18.13
39.41 ? 3.94
40.09 ? 4.33
50.37 ? 4.90
49.63 ? 4.90
?66.54 ? 20.94
70.40 ? 19.53
?23.21 ? 17.15
?35.08 ? 19.06
40.49 ? 4.05
41.76 ? 3.97
49.91 ? 4.76
50.09 ? 4.76
?58.30 ? 22.94
66.15 ? 22.21
58.78 ? 38.08
78.42 ? 47.34
32.02 ? 7.39
33.98 ? 7.72
13.15 ? 4.32
12.26 ? 4.38
221.42 ? 81.15
91.49 ? 17.07
6.06 ? 12.45
0.52 ? 0.24
0.47 ? 0.28
3.61 ? 5.28
1.74 ? 1.92
47.01 ? 38.94
57.00 ? 51.73
36.02 ? 11.40
35.12 ? 9.20
13.28 ? 5.86
12.73 ? 5.68
203.00 ? 100.40
97.37 ? 22.51
13.18 ? 14.88
0.52 ? 0.26
0.42 ? 0.28
2.44 ? 3.23
1.27 ? 1.30
Note.—Values presented as means ? SD. AVM ? arterial/venous mismatch, CSA ? cross-sectional area, Fd ? dominant venous
flow rate, FD ? flow distribution, Fdj ? flow through dominant jugular vein, Fsd ? subdominant venous flow rate, Fsdj ? flow through
subdominant jugular vein, LCCA ? left common carotid artery, LIJV ? left internal jugular vein, LVA ? left vertebral artery, MS ?
multiple sclerosis, RCCA ? right common carotid artery, RVA ? right vertebral artery, RIJV ? right internal jugular vein, tA ? total
arteries, tIJV ? total internal jugular veins, tLA ? total left-sided arteries, tRA ? total right-sided arteries, tV ? total veins.
68.e2 ? Abnormal IJV Flow Distribution in MS with Venous AbnormalitiesHaacke et al ? JVIR
APPENDIX VII: Flow Measurements Table for Nonstenotic and Stenotic Patient Populations with MS
No. of pts. (NS/S)
Heart rate (/s)
Flow rate (mL/s)
Reflux LIJV (%)
Reflux RIJV (%)
Site 1 Site 2
Mean (NS/S)SD (NS/S) Mean (NS/S)SD (NS/S)
Note.—Values presented as means ? SD. AVM ? arterial/venous mismatch, CSA ? cross-sectional area, Fd ? dominant venous
flow rate, FD ? flow distribution, Fdj ? flow through dominant jugular vein, Fsd ? subdominant venous flow rate, Fsdj ? flow
through subdominant jugular vein, LCCA ? left common carotid artery, LIJV ? left internal jugular vein, LVA ? left vertebral artery,
MS ? multiple sclerosis, NS ? nonstenotic, RCCA ? right common carotid artery, RVA ? right vertebral artery, RIJV ? right internal
jugular vein, S ? stenotic, tA ? total arteries, tIJV ? total internal jugular veins, tLA ? total left-sided arteries, tRA ? total right-sided
arteries, tV ? total veins.
Volume 23 ? Number 1 ? January ? 201268.e3