J. Biomedical Science and Engineering, 2010, 3, 1093-1098 JBiSE
doi:10.4236/jbise.2010.311142 Published Online November 2010 (http://www.SciRP.org/journal/jbise/).
Published Online November 2010 in SciRes. http://www.scirp.org/journal/jbise
Labeled Hepasphere behavior during venous drainage
simulation at 1.5T*
Hassan Jassar, François Langevin
UMR6600-Centre of Advanced Medical Imaging, University of Technology of Compiegne, Compiegne, France.
Email: Hassan.jassar@utc.fr
Received 19 August 2010; revised 25 September 2010; accepted 30 September 2010.
Stability of the magnetic resonance (MR) contrast
agent inside vascular occlusion agents is important
for their localization with magnetic resonance imag-
ing (MRI). The aim of this paper is to study the be-
haviour of the superparamagnetic iron oxide (SPIO)
within Hepaspheres microparticles (MP) by MRI
when they are submitted to negative pressure in-
duced by venous drainage of a tumor. Therefore, a
venous drainage model was established and three
parameters were taken into account according to
physiologic parameters in tumors: pH, temperature
and flow blood rate. Four cycles of pumping were
performed with the presence of labeled Hepas-
pheres with Endorem®. Several MR images of MP
and perfusion liquid were taken before and after
pumping. Endorem® release was determined after
correction of non-uniformity intensities in MR images.
Intensity variation according to spatial position, coil
and MR acquisition parameters was studied. Labeled
microparticles (LB*MP) appeared as black spots in
MRI images whatever duration and pH. Our model
demonstrates the stability of the SPIO inside the oc-
clusion agent during time. Moreover, the proposed
correction method proves the reduction of the inten-
sity non-uniformity in MRI images.
Keywords: MRI; Venous Drainage Model; SPIO; En-
dorem®; Hepaspheres; Intensity Non-Uniformity
Embolization with MP consists in stopping blood flow
and starving tissues of oxygen and nutrients [1]. For
instance, Hepaspheres are dry MP in original state and
become non-biodegradable microspheres after swelling
within ionic fluids [2]. They are used as occlusion agents
in hepatocellular carcinoma and arteriovenous malfor-
mations [3] and may release drugs in situ by diffusion
[4]. Labelling these MP with superparamagnetic iron
oxide (SPIO) to localize them has already been demon-
strated [5]. However, the stability of SPIO within MP
along time remains an important question, as a release of
the contrast agent from MP could avoid the detection of
their real position in clinical applications. Vessel occlu-
sion provokes negative pressure on MP, especially with
existing arteriovenous shunt around necrotic tissues or
tumors [6,7].
Venous drainage model was performed in this work to
determine SPIO behaviour within Hepaspheres by MRI
during time. Several physiological factors of tumors
were taken into consideration in the model: flow, tem-
perature and pH. Simulation value of the flow was cho-
sen much higher than standard one, as pessimistic condi-
tions. This work included correction of the MR intensity
non-uniformity for quantitative analysis reproducibility
on MR images, leading to the estimation of released
SPIO from the MP.
2.1. Labelling Hepaspheres MP with Endorem®
SAP-MS (sodium acrylate and vinyl alcohol copolymer)
or Hepaspheres provided by Biosphere Medical SA
France were used as occlusion agents in hepatocellular
carcinoma and arteriovenous malformations [2-4]. Parti-
cle sizes of Hepaspheres in dry state is calibrated in 50
m increments ranging from 50 to 200 m (50-100,
100-150 and 150-200 m) [2]. Their diameters after
swelling are approximately 2 to 3.5 times larger than
their original size [8].
500 l (25% v/v) of Endorem® was diluted into 1.5 ml
(75% v/v) of saline solution. Solution was poured re-
spectively into a bottle containing dry Hepaspheresof
150200 m size (Ref: V705HS). After two hours of
Endorem® absorption, the preparation was then poured
into a glass column with porous filter (porosity: 20 m–
*This research was supported by the Centre of Advanced Medical Im-
aging at the University of Technology of Compiegne.
H. Jassar et al. / J. Biomedical Science and Engineering 3 (2010) 1093-1098
Copyright © 2010 SciRes. JBiSE
Flex 1.0 30 cm- Ref: 420401-1030). Labeled Hepas-
pheres with Endorem® were washed by gravity in this
column: 10 ml of saline was added to the Hepaspheres
during 15 minutes, to let saline leave entirely the column.
The washing process was repeated four times.
A few MP were picked up and then fixed between two
transparent gel layers of a Petri dish for using as a con-
trol (before pumping).
2.2. Venous Drainage Model
In vitro venous drainage model consisted of closed cir-
cuit and composed of a peristaltic pump (Ref. 40578
Fisher Bioblock Scientific), a thermostat bath (Julabo 5
liters, Ref: ED-5A/B) within immersed a glass column
with porous filter (Figure 1). Labeled Hepaspheres
with Endorem® were placed in the column and submitted
to 20 ml of the saline. According to physiological pa-
rameters of hepatocarcinoma, flow rate of the peristaltic
pump was adjusted to 10 ml/min, four times over-
evaluated blood flow through hepatic tumor of 16g [6];
tow pH, 6.0 and 7.0, were used in this model. The ther-
mostat bath was maintained at 37°C.
Four cycles of saline pumping was performed through
the closed circuit, two hours per cycle. Every two hours,
10 ml of perfusion liquid was filled into vials to dose
released SPIO by MRI. After each pumping cycle, some
LB*MP were picked up from the glass column and
dropped on a gel of a Petri dish, with MP control un-
submitted to pumping cycles.
Two pH solutions (6 and 7) were tested, providing
eight vials of perfusion samples. A new 20 ml of the
saline solution at one pH was added to the glass column
after each pumping cycle.
MR scans were performed on GE Signa® 1.5T Excite
11.0 scanner. Quantitative analyses of all MRI images
were done in our laboratory using an Advantage Win-
dows (4.1-GE) workstation. Values averaging of signal
Figure 1. Simulation of venous drainage model with the pres-
ence of labeled Hepaspheres. Perfusion liquid is filled into
vials at the bottom right side of the schema.
intensities were performed with MATLAB® 7.0 and
EXCEL® softwares.
2.3. Perfusion Liquid Imaging
Most intensity non-uniformities result from materials of
the MR scanner such as coil, its sensitivity and the ra-
diofrequency system [9-11]. Intensity non-uniformity
correction methods may be classified into two types, as
reported in literature [12]: empirical methods and post-
processing techniques. Empirical methods involve scan-
ning water or oil phantom prior to clinical examination
[9,11,13,14] to obtain an estimate on the scanner’s bias
field. Post-processing techniques are the most commonly
used approaches in the quantitative analysis of MR im-
ages [12,15-23]. These methods privilege modelling of
the effects but not the cause of the intensity non-uni-
formities. Therefore, it is difficult to appreciate the per-
tinence of these methods.
The proposed method was a hybrid method taking
advantage of the two categories: it first constituted a
reference from an image of a phantom with given hard-
ware and acquisition parameters. The independence of
the sample location in the analysed region was searched
by averaging the extracted values from phantom images
after successive rotation of the phantom or the perfusion
2.3.1. Coil and Imaging Parameters
Volume coil (Figure 2(c)), such as head coil (diameter
30 cm and length 40 cm, one channel), and 2D FSE-XL
T1w were used in all MR acquisitions to provide low
image-intensity non-uniformity [24-26]. Six contiguous
slices were obtained in coronal plane with the imaging
parameters: Repetition time/echo time = 500/15 ms,
bandwidth = 15.63 kHz, field of view = 180 180 mm,
matrix size = 128 128, slice thickness = 5 mm, number
of signal average = 6, and echo train length = 4.
2.3.2. Reference Phantom and Intensities Averaging
A phantom composed of eight vials were uniformly
filled with 10 ml of saline solution (0.9%), and equally
distributed into two square matrices, matrix-1 and ma-
trix-2 (Figure 2(a)). They were positioned on a rotating
wooden support, equidistant from the coil centre with
5.6 cm of radius (Figure 2(b)). 45° is the angle between
the two matrices and 8 cm is the distance between vials
of the matrix. The ninth vial was in central position.
MR images of the phantom were acquired in four di-
rections (superior «S», inferior «I», left «L» and right
«R») after 90° of the support rotation. For this purpose,
the phantom was brought out for each 90° rotation from
the scanner without changing the support position inside
the volume coil. Values of calibration in the first MR
acquisition were kept the same after each «Prescan» for
H. Jassar et al. / J. Biomedical Science and Engineering 3 (2010) 1093-1098
Copyright © 2010 SciRes. JBiSE
(a) (b) (c)
Figure 2. Schemas and photos of the saline solution phantom. (a) Eight vials were placed equidistantly to 5.6 cm
from the centre «c» and distributed into two matrices: «matrix-1» and «matrix-2» at 45°. At centre «c», a ninth vial
was common to the both matrices. (b) Glass vials were positioned on a rotating support of wood. (c) Fixed support
of polystyrene carried vials on the wooden support to volume coil and tunnel centre.
the four directions.
Mean signal intensities of the four vials in the ma-
trix-1 and matrix-2 were determined for a given direc-
tion. Dissimilarity of intensity values in the MR images
could be an indicator of the percentage of intensity
non-uniformity resulting from vials neighboring.
Intensity non-uniformity correction by signal averag-
ing of the same vial was performed in the four directions
after successive rotation of the phantom support.
The pertinence of this method was evaluated accord-
ing to Wicks’ formula [11]:
 (1)
where (G) is the percentage of non-uniformity, (
) is the
standard deviation of the sample vial intensities in the
MR image and (
) their mean.
Therefore, percentage of non-uniformity was calcu-
lated before (G) and after averaging (G).
2.3.3. Perfusion Liquid and Intensities Averaging
Eight vials of perfusion liquid were placed on the sup-
port equidistantly from the coil centre, and distributed
into two matrices, «matrix-1» for pH7 and «matrix-2»
for pH6 at 45°. The ninth vial containing intact saline
solution was common to both matrices. Perfusion liquid
imaging was performed likely to phantom acquisitions to
determine signal intensities
of vials on T1w image.
Intensities averaging
were performed for each vial
(pH6 and pH7) of the two matrices, in the four directions.
Then, contrast between signal intensities of samples and
saline solution was determined according to equation:
3.1. Hepaspheres Imaging
Figure 3 shows T1w images of two Petri dishes plunged
into water and containing labeled Hepaspheres with
SPIO. LB*MP submitted to pumping cycles (2h, 4h, 6h
and 8h) remains visible on MR images as black spots
whatever pumping time and pH of saline. This result was
similar to that of LB* MP unsubmitted to pumping cycles
(0h). This means that Endorem® was unreleased or
slightly released without affecting the signal intensity of
labeled LB*MP. Determined intensities of perfusion li-
quid provide more information on a trace of SPIO that pos-
sibly released from LB*MP, as shown in Subsection 3.3.
Figure 3. T1w images of Petri dishes after be-
ing plunged in water. These contained labeled
Hepaspheres with Endorem® that submitted
to four cycles of pumping at pH7 (left) and
pH6 (right). MR images were acquired in cor-
onal plane with 3D SPGR T1w acquisition of
parameters: TR/echo time = 35/5 ms, flip an-
gle = 45°, field of view = 140 140 mm, slice
thickness = 1 mm, number of slices = 28, matrix
size = 256 256, bandwidth = 15.63 kHz and
Nex = 1. Imaging was realized with the use of
a surface coil (phased array, 4 channels).
H. Jassar et al. / J. Biomedical Science and Engineering 3 (2010) 1093-1098
Copyright © 2010 SciRes. JBiSE
3.2. Phantom Imaging and Averaging Intensities
Figure 4(a) is T1w image of the phantom whose vials are
equidistantly distributed from the centre «c». Figure 4(b)
shows clearly the influence of neighbors proximity on
the signal intensity. Percentage difference between mean
signal intensities of «matrix-1» and «matrix-2» is sig-
nificant in the four directions (Table 1).
Figure 5 shows a non-uniformity percentage variation
of intensities G versus vial positions and phantom direc-
tions, before and after averaging. Before averaging, G
variation is important. It is less significant after averag-
ing (Figure 5(a), green curves). 0.094 is the average
absolute deviation of G (Figure 5(b)). This value
means low intensity non-uniformity for the equidistant
distribution of vials on the support.
Figure 4. T1w images of a saline phantom acquired by FSE-XL
T1. (a) MR images of eight vials equidistantly distributed from
the centre «C» into two matrices: «matrix-1» and «matrix-2» at
45°. (b) Vials’ intensities of matrix-1, matrix-2 and saline solu-
tion at the centre, in the four directions.
Table 1. Percentage difference between averaged intensities of
«matrix-1» against «matrix-2» at 45° according to vials’ posi-
tions and directions for a phantom of saline solution.
Directions % difference of vial’s intensities
1 2,19%
2 1,56%
3 1,75%
4 2%
Position index of vials «n» on the su
Position index of vials «n» on the su
Figure 5. (a) Non-uniformity percentage of signal intensities
(%) determined on each vial (n = 0 to 8) of the saline phantom
before (Gphant.(i,n), red, blue, yellow and pink colours) and after
averaging (
G, green colour). The point 0 corresponds to
the non-uniformity percentage of signal for a vial in the central
position. (b) Average absolute deviation of non-uniformity
percentage averaging (
G) against mean of
G for
each element of the phantom. Zero point corresponds to the
vial in common to the two matrices.
3.3. Perfusion Liquid Signal
The signal at pH7 is greater than that of saline solution
at the two first cycles of pumping, 2h and 4h. 22% is the
percentage of the signal decrease between 2h and 6h of
pumping time, then the signal becomes stable (Figure 6).
15% at 2h and 6% at 4h are the contrast of perfusion
liquid-saline solution; it is almost zero at 6h and 8h of
pumping time.
H. Jassar et al. / J. Biomedical Science and Engineering 3 (2010) 1093-1098
Copyright © 2010 SciRes. JBiSE
Figure 6. Representation of perfusion liquid signal (right) and
liquid-saline contrast (left) in function of hours of pumping, at
pH6 and pH7. Each signal value was averaged in the four di-
When the signal of perfusion liquid is greater than that
of saline solution the contrast is important, proving that
Endorem® has been slightly released from Hepaspheres.
However, when the signal is equivalent to that of saline
solution, the contrast becomes zero after 6h of pumping
time. Therefore, the Endorem® release does not exist.
The signal of perfusion liquid at pH6 after eight hours
of pumping is close to that of saline solution. 3% is the
contrast of perfusion liquid against saline solution. This
means an absence of Endorem® in the perfusion liquid
during all pumping time.
For the three physiological parameters values of hepato-
cellular carcinoma, pH, temperature and flow rate, we
demonstrated that the labeled Hepaspheres with En-
dorem® could be detected by MRI in the proposed condi-
tions. These are visible in conventional MRI (1.5T) as
black spots (negative contrast or low signal) and distin-
guished from the surrounding signal of gel whatever the
hours of pumping and the pH.
Perfusion liquid-saline contrast difference between
pH7 and pH6 is 14.8% at 2h and 11.5% at 4h of pump-
ing. It then becomes 2.5%. This implies a release of En-
dorem® from Hepaspheres in the first hours of pump-
ing, only at pH7.
Then, releasing reaches the stability when the signal
of perfusion liquid becomes very close to that of saline
solution. This can be demonstrated by zero release of the
contrast agent after several hours of pumping, although,
Hepaspheres’ properties let drug absorption and deliv-
ery. This observation may be depending on the spongy
behaviour of Hepaspheres. Very low release of En-
dorem® at pH6 shows that the behaviour of Hepas-
pheres could be modified by pH decreasing.
For the proposed intensity non-uniformity correction
method and the distribution of vials to the volume coil
centre, we demonstrated that the average absolute devia-
tion of the non-uniformity percentage was 0.094.
This method reduced the intensity non-uniformities in
function of position, coil and acquisition parameters, and
provided an accurate measure of the perfusion liquid
signal. Consequently, quantification of released Endo-
rem® could be determined precisely.
Finally, we noticed that the preliminary process of
four hours of pumping provides visible Hepaspheresin
MRI images.
This study determines the behaviour of labeled
Hepasphereswith Endorem® when submitted in a ve-
nous drainage model to negative pressure and some
physiological parameters of Hepatocarcinoma. Using
1.5T scanner for Hepaspheres imaging demonstrates a
permanent MR detection of Hepaspheres whatever the
proposed conditions that can be found nearly to tumors.
This result provides stable labeled Hepaspheres with
SPIO for using in MR interventional application.
Authors thank Philippe Robert of Guerbet for providing the SPIO. We
thank Philippe Reb of Biosphere Medical for providing Hepaspheres.
[1] Laurent, A., Wassef, M., Chapot, R., Wang, Y., Houdart,
E., Feng, L., Tran Ba Huy, E. and Merland, J.J. (2005)
Partition of calibrated tris-acryl gelatin microspheres in
the arterial vasculature of embolized nasopharyngeal an-
giofibromas and paragangliomas. Journal of Vascular
and Interventional Radiology, 16(4), 507-513.
[2] Osuga, K., Hori, S., Kitayoshi, H., Khankan, A., Okada,
A., Sugiura, T., Murakami, T., Hosokawa, K. and Naka-
mura, H. (2002) Embolization of high flow arteriovenous
malformations: Experience with use of superabsorbent
polymer microspheres. Journal of Vascular and Interven-
tional Radiology, 13(11), 1125-1133.
[3] De Luis, E., Bilbao, J.I., de Ciércoles, J.A.G.J., Cuesta, A.
M., Rodríguez, A.M. and Lozano, M.D. (2008) In vivo
evaluation of an new embolic spherical particle (HepaS-
phere) in a kidney animal model. Cardiovascular and
H. Jassar et al. / J. Biomedical Science and Engineering 3 (2010) 1093-1098
Copyright © 2010 SciRes. JBiSE
Interventional Radiology, 31(2), 367-376.
[4] Osuga, K., Khankan, A.A., Hori, S., Okada, A., Sugiura,
T., Maeda, M., Nagano, H., Yamada, A., Murakami, T.
and Nakamura, H. (2002) Transarterial embolization for
large hepatocellular carcinoma with the use of superab-
sorbent polymer microspheres: initial experience. Jour-
nal of Vascular and Interventional Radiology, 13(9), 929-
[5] Jassar, H. (2009) Detectability of vascular occlusion
materials for MRI follow-up. Ph.D. Dissertation, Com-
piegne University of Technology, Compiegne.
[6] Vaupel, P., Rallinowski, F. and Okunieff, P. (1989) Blood
flow, oxygen and nutrient supply, and metabolic micro-
environment of human tumors: A review. Cancer Re-
search, 49(23), 6449-6465.
[7] Tancredi, T., McCuskey, P.A., Kan, Z. and Wallace, S.
(1999) Changes in rat liver microcirculation after ex-
perimental hepathic arterial embolization: Comparison of
different embolic agents. Radiology, 12(1), 117-181.
[8] Khankan, A.A., Osuga, K., Hori, S., Morii, E., Murakami,
T. and Nakamura, H. (2004) Embolic effects of superab-
sorbent polymer microspheres in rabbit renal model:
comparison with tris-acryl gelatin microspheres and
polyvinyl alcohol. Radiation Medicine, 22(6), 384-390.
[9] Simmons, A., Arridge, S.R., Barker, G.J. and Williams,
S.C. (1994) Sources of intensity nonuniformity in spin
echo images at 1.5 T. Magnetic Resonance in Medicine,
32(1), 12-18.
[10] Haacke, E.M., Brown, R.W., Thompson, M.R. and
Venkatesan, R. (1999) Magnetic resonance imaging.
Physical Principles and Sequence Design, Wiley-Liss,
New York.
[11] Wicks, D.A.G., Barker, G.J. and Tofts, P.S. (1993) Cor-
rection of intensity nonuniformity in MR images of any
orientation. Magnetic Resonance Imaging, 11(2), 183-
[12] Zhou, L.Q., Zhu, Y.M., Bergot, C., Laval-Jeantet, A.M.,
Bousson, V., Laredo, J.D. and Laval-Jeantet, M. (2001) A
method of radio-frequency inhomogeneity correction for
brain tissue segmentation in MRI. Computerized Medical
Imaging and Graphics, 25(5), 379-389.
[13] Axel, L., Constantini, J. and Listerud, J. (1987) Intensity
correction in surface-coil MR imaging. American Jour-
nal of Roentgenology, 148, 418-420.
[14] Simmons, A., Arridge, S.R., Barker, G.J., Cluckie, A.J.,
and Tofts, P.S. (1994) Improvements to the quality of
MRI cluster analysis. Magnetic Resonance Imaging,
12(8), 1191-1204.
[15] Styner, M., Brechbuhler, C., Szekely, G. and Gerig, G.
(2000) Parametric estimate of intensity inhomogeneities
applied to MRI. Medical Imaging, 19(3), 153-165.
[16] Lai, S.H. and Fang, M. (1999) A new variational shape-
from-orientation approach to correcting intensity inho-
mogeneities in magnetic resonance images. Medical Im-
age Analysis, 3(4), 409-424.
[17] Pham, D.L. and Prince, J.L. (1999) A generalized EM
algorithm for robust segmentation of magnetic resonance
images. 33rd Annual Conference on Information Sciences
and Systems, Baltimore, 558-563.
[18] Narayana, P.A. and Bourthakour, A. (1995) Effect of
radio frequency inhomogeneity correction on the repro-
ducibility of intra-cranial volumes using MR image data.
Magnetic Resonance in Medicine, 33(3), 396-400.
[19] Vorkurka, E.A., Thaker, N. and Jackson, A. (1999) A fast
model independent method for automatic correction of
intensity nonuniformity in MRI data. Journal of Mag-
netic Resonance Imaging, 10(4), 550-562.
[20] Brinkmann, B.H., Manduca, A. and Robb, R.A. (1998)
Optimized homomorphic unsharp masking for MR gray-
scale inhomogeneity correction. IEEE Medical Imaging,
17(2), 161-171.
[21] Likar, B., Viergever, M.A. and Pernus, F. (2000) Retro-
spective correction of MR intensity inhomogeneity by
information minimization. MICCAI Conference on Medi-
cal Image Computing and Computer-Assisted Interven-
tion, 1935/2000, Pittsburgh, 375-384.
[22] Cohen, M.S., DuBois, R.M. and Zeineh, M.M. (2000)
Rapid and effective correction of RF inhomogeneity for
high field magnetic resonance imaging. Human Brain
Mapping, 10(4), 204-211.
[23] Han, C., Hatsukami, T.S. and Yuan, C. (2001) A multi-
scale method for automatic correction of intensity non-
uniformity in MR images. Proceedings of International
Society for Magnetic Resonance in Medicine, 13(3), 428-
[24] Hayes, C.E., Edelstein, W.A., Schenck, J.F., Otward, M.
M. and Eash, M. (1985) An efficient, Highly Homoge-
neous Radiofrequency Coil for whole-body NMR Imag-
ing at 1.5 T. Journal of Magnetic Resonance, 63, 622-
[25] Simmons, A., Tofts, P.S., Barker, G.J. and Arridge, S.R.,
(1994) Sources of intensity non uniformity in spin echo
images at 1.5T. Magnetic Resonance in Medicine, 32(1),
[26] Barker, G.J., Simmons, A., Arridge, S. R. and Tofts, P.S.,
(1998) A simple method for investigating the effects of
non-uniformity of radiofrequency transmission and ra-
diofrequency reception in MRI. The Britch Journal of
Radioliogy, 71(841), 59-67.