The dynamics of brain and cerebrospinal fluid growth in normal versus hydrocephalic mice.

Abstract

Hydrocephalus has traditionally been quantified by linear measures of ventricular size, with adjunct use of cortical mantle thickness. However, clinical outcome depends on cognitive function, which is more directly related to brain volume than these previous measures. The authors sought to quantify the dynamics of brain and ventricular volume growth in normal compared with hydrocephalic mice.
Hydrocephalus was induced in 14-day-old C57BL/6 mice by percutaneous injection of kaolin into the cisterna magna. Nine hydrocephalic and 6 normal mice were serially imaged from age 2-12 weeks with a 14.1-T MR imaging unit. Total brain and ventricle volumes were calculated, and linear discriminant analysis was applied.
Two very different patterns of response were seen in hydrocephalic mice compared with mice with normative growth. In one pattern (3 mice) brain growth was normal despite accumulation of CSF, and in the second pattern (6 mice) abnormal brain enlargement was accompanied by increased CSF volume along with parenchymal edema. In this latter pattern, spontaneous ventricular rupture led to normalization of brain volume, implying edema from transmantle pressure gradients. These 2 patterns of hydrocephalus were significantly discriminable using linear discriminant analysis (p < 0.01). In contrast, clinically relevant measurements of head circumference or frontal and occipital horn ratios were unable to discriminate between these patterns.
This study is, to the authors' knowledge, the first serial quantification of the growth of brain and ventricle volumes in normal versus hydrocephalic development. The authors' findings demonstrate the feasibility of constructing normative curves of brain and fluid growth as complements to normative head circumference curves. By measuring brain volumes, distinct patterns of brain growth and enlargement can be observed, which are more likely linked to cognitive development and clinical outcome than fluid volumes alone.

Full text

Page 1

J Neurosurg: Pediatrics / Volume 6 / July 2010
J Neurosurg: Pediatrics / Volume 6 / July 2010
J Neurosurg Pediatrics 6:000–000, 2010J Neurosurg Pediatrics 6:1–10, 2010
1
1
H
with the rapid advancement of neuroimaging technology,
shunt-valve design, and surgical techniques, outcomes
in patients with hydrocephalus have changed very little
in the past 50 years.10 This lack of progress may be due
to the fact that current treatment protocols, aimed at de-
creasing excess fluid in the intracranial cavity, have also
changed little. The fluid diversionary shunt was the first
effective treatment developed and is still the most widely
used form of treatment today.18
It has been known since the first models of experimen-
tal hydrocephalus created by Dandy7 that hydrocephalus
ydrocepHalus is a disorder resulting in the ac-
cumulation of intracranial CSF due to abnormal
CSF production, circulation, or absorption.31 Even
is accompanied by pathophysiological changes in brain
morphology. Since then, studies have shown that hydro-
cephalus can result in a thinning of the cortex, as well as
an increase in the water content and loss of myelin in the
periventricular white matter.8,23,25 Despite known morpho-
logical changes of brain tissue in the development of hydro-
cephalus, experimental work has focused on measurements
of the ventricular system—mainly linear measurements of
ventricular size alone7 or ratios of ventricle to brain thick-
ness and area.5,19,22,29,30 When studies have included linear
measurements of cortex on an imaging section, the au-
thors have often not taken into account other areas of the
brain or total brain volumes.4,17 Head circumference mea-
The dynamics of brain and cerebrospinal fluid growth in
normal versus hydrocephalic mice
Laboratory investigation
Jason G. Mandell, M.s.,1,2 ThoMas neuberGer, Ph.d.,3 Corina s. draPaCa, Ph.d.,1
andrew G. webb, Ph.d.,4 and sTeven J. sChiff, M.d., Ph.d.1,3,5
1Center for Neural Engineering, Department of Engineering Science and Mechanics, 2Department of
Bioengineering, 3Huck Institutes of the Life Sciences, and 5Departments of Neurosurgery and Physics,
Pennsylvania State University, University Park, Pennsylvania; and 4C. J. Gorter Center for High Field MR
Imaging, Department of Radiology, Leiden University Medical Center, Leiden, The Netherlands
Object. Hydrocephalus has traditionally been quantified by linear measures of ventricular size, with adjunct use
of cortical mantle thickness. However, clinical outcome depends on cognitive function, which is more directly related
to brain volume than these previous measures. The authors sought to quantify the dynamics of brain and ventricular
volume growth in normal compared with hydrocephalic mice.
Methods. Hydrocephalus was induced in 14-day-old C57BL/6 mice by percutaneous injection of kaolin into the
cisterna magna. Nine hydrocephalic and 6 normal mice were serially imaged from age 2–12 weeks with a 14.1-T MR
imaging unit. Total brain and ventricle volumes were calculated, and linear discriminant analysis was applied.
Results. Two very different patterns of response were seen in hydrocephalic mice compared with mice with
normative growth. In one pattern (3 mice) brain growth was normal despite accumulation of CSF, and in the second
pattern (6 mice) abnormal brain enlargement was accompanied by increased CSF volume along with parenchymal
edema. In this latter pattern, spontaneous ventricular rupture led to normalization of brain volume, implying edema
from transmantle pressure gradients. These 2 patterns of hydrocephalus were significantly discriminable using linear
discriminant analysis (p < 0.01). In contrast, clinically relevant measurements of head circumference or frontal and
occipital horn ratios were unable to discriminate between these patterns.
Conclusions. This study is, to the authors’ knowledge, the first serial quantification of the growth of brain and
ventricle volumes in normal versus hydrocephalic development. The authors’ findings demonstrate the feasibility of
constructing normative curves of brain and fluid growth as complements to normative head circumference curves. By
measuring brain volumes, distinct patterns of brain growth and enlargement can be observed, which are more likely
linked to cognitive development and clinical outcome than fluid volumes alone. (DOI: 10.3171/2010.4.PEDS1014)
Key words      •      hydrocephalus      •      magnetic resonance imaging      •      kaolin      •     
cerebrospinal fluid      •      cerebral ventricle      •      development      •      brain growth
Abbreviation used in this paper: LDA = linear discrimination
analysis.
This article contains some figures that are displayed in color
on line but in black and white in the print edition.

Page 2

J. G. Mandell et al.
2
J Neurosurg: Pediatrics / Volume 6 / July 2010
surements are a poor substitute for intracranial brain and
fluid volumes and are most effective during the first 10–24
months of life.1,32 Ventricular horn indices, ventricle/brain
ratios, and cortical thickness measurements are often used
clinically as substitutes for head circumference.24,27 All of
these measures, however, fail to fully characterize morpho-
logical changes in hydrocephalic development and do not
capture total brain volume.
This study attempts to understand brain and fluid
volume dynamics in normal development and to contrast
these with an experimental model of hydrocephalus. We
define hydrocephalus as the accumulation of CSF at great-
er than 3 SDs above normative averages in the developing
brain. To meet this definition, we need normative growth
curves, which we will define in this paper. By defining nor-
mal volume dynamics, we can explore the feasibility of an
improved framework for a more rational design of therapy
than is currently provided by the use of head circumfer-
ence or linear measurements from brain imaging. Our
hypothesis is that achieving normal brain volume growth
is likely more important than fluid volume per se in the
achievement of normal cognitive function in children. The
ultimate goal in the treatment of pediatric hydrocephalus is
to create an intracranial environment for the brain to prop-
erly develop. Therefore, assessing both dynamic brain and
fluid volumes will be important in improving the diagnosis
and management of pediatric hydrocephalus.
Methods 
All animal procedures were approved by Penn State
University’s institutional animal care and use committees.
Induction of Hydrocephalus
Fourteen-day-old C57BL/6 mice (Harlan Labs) were
anesthetized (2% isoflurane in 0.5 L/minute O2). Ten mi-
croliters of sterilized kaolin solution, composed of 250 mg
kaolin and 50 μl Magnevist contrast agent (Bayer Health-
care) per milliliter of 0.9% saline, were injected percutane-
ously using aseptic technique into the cisterna magna with
a 100-μl Hamilton syringe and 30-gauge needle. Animals
were observed daily postinjection and body weights were
obtained prior to imaging. A total of 35 mice from 6 litters
were injected.
Magnetic Resonance Imaging
All imaging was performed on an 89-mm vertical
wide-bore 14.1-T unit (Varian, Inc.) with Direct Drive
technology. The gradient system had an inner diameter
of 55 mm and maximum field strength of 550 mT/m. A
T2-weighted spin echo pulse sequence (TR 5000 msec,
TE 30 msec) with a custom-built surface resonator (2 ×
1.7 mm) was used to achieve high contrast between the
brain tissue and fluid. Partial Fourier imaging was used,
with 75% k-space coverage (96 points) in the phase-en-
code direction; 160 complex data points were acquired
in the readout direction, and the data were zero-filled to
320 × 256. Slice thickness was 200 μm, while in-plane
resolution was between 81 and 94 μm, depending on the
FOV used. The slice acquisition was interleaved with no
gap between the slices. Four averages were acquired for
a total scan time of 32 minutes. Control mice were lit-
termates that were not injected. The control mice were
imaged once per week from 2 to 4 weeks of age, then ev-
ery other week from 4 to 12 weeks of age. Hydrocephalic
mice underwent imaging just before the kaolin injection
at 2 weeks of age, 2 days postinjection, and then every
3–5 days whenever possible.
Image Processing and Volumetric Analysis
The images were imported into AMIRA 4.1 (Visage
Imaging) for processing and segmentation. A 2D 3-by-3
median filter was applied to the images before segmenta-
tion. The AMIRA system uses semiautomatic segmen-
tation tools such as region growing, edge tracing, and
connected thresholding tools, as well as a manual pixel
selection tool, all of which were used to segment data in
this study. The total brain volume was generated by seg-
menting the surface area for each slice, beginning ros-
trally with the olfactory bulb region and ending caudally
in the brainstem on the slice showing the most ventral
portion of the cerebellum. Fluid from the lateral, third,
and fourth ventricles was included in this segmentation.
An atlas14 was used to determine correct tissue and fluid
classification, and the results were checked by a neurosur-
gery expert and an MR imaging expert.
In addition to calculating brain and fluid volumes of
the hydrocephalic and normal mice, we calculated the
frontal and occipital horn ratio24 and head circumference
from the images. The frontal and occipital horn ratio was
calculated as the sum of the maximal width of the frontal
horn and occipital horn of the lateral ventricles divided
by twice the maximum brain width taken in an axial slice
at the level of the cerebral aqueduct. On the same axial
slice, head circumference was estimated as the circum-
ference of the brain and fluid in the cranial cavity. These
measures were calculated for the final data set acquired
for each normal and hydrocephalic mouse. Head circum-
ference was also calculated for normal 3-week-old mice
to compare with the hydrocephalic mice of the same age.
However, it was not possible to measure the frontal and
occipital horn ratio in the mice at this age because the
fluid could not be observed in the occipital horn in a brain
this immature.
Data Analysis
In 1936 Roland Fisher12 created a method of multivari-
able discrimination to help classify data that had more than
1 measurement and that came from more than 1 group of
items. Fisher’s problem was motivated by iris flowers. He
had petal and sepal length and width measurements of
each of 3 species (50 samples each). He was able to find
the optimal way of adding these 4 measurement variables
together so that he could clearly show that these sets of
measurements could separate and classify each species
type. Indeed, the method provides a recipe to measure a
new item, weight the measurements, and optimally clas-
sify the likely species for this out of sample data. A similar
biological example, with only a few items, was shown by
Bernhard Flury using midges measured for antenna and

Page 3

J Neurosurg: Pediatrics / Volume 6 / July 2010
Dynamics of brain and CSF growth in hydrocephalic mice
3
wing length.13,15 We have refined this method, to take into
account modern numerical computer algorithms28 (Fisher
did all of his work on a hand calculator), and we employ
this numerically stable form of LDA in what follows.
We used LDA to test the classification of hydrocepha-
lic data into 2 separate groups: 1) hydrocephalic mice with
abnormal brain enlargement and 2) hydrocephalic mice
with brain growth consistent with normal controls. In this
study the variables used for analysis are brain volume and
ventricle volume. For each hydrocephalic mouse, the final
brain and ventricle volume measurements were normal-
ized by the age-matched control brain and ventricle vol-
umes. This resulted in a 9 × 2 matrix, Y, consisting of a
single brain and single ventricle volume measurement for
each of 9 hydrocephalic mice. The data from each mouse
were classified into either Group 1 for hydrocephalic mice
with abnormal brain enlargement or Group 2 for hydro-
cephalic mice with normal brain growth.
Wilks test statistic, W, was used to test for the signifi-
cance of the classification. This likelihood ratio tests the
hypothesis, H0, that each group mean, µk, is equal:
H0 : μ1 = μ2 = ... = μk [1]
In addition to testing the significance of the rejec-
tion of H0, the Wilks statistic is also used to test random
combinations of regrouping the data with a bootstrapping
method.11,13,28 This is employed to prove that the grouping
defined prior to the LDA is unlikely to have occurred by
chance. The Wilks statistic for each permutation of the
data is compared with the Wilks statistic for the origi-
nally defined classification. The bootstrap probability, Pb,
is the probability that the original classification would oc-
cur randomly and is given by
where Nless is the number of groupings found with a Wilks
statistic less than or equal to the original W statistic cal-
culated, and Nperm is the number of permutations. This
calculation includes the original grouping as one of the
permutations in addition to those calculated in the ran-
domization of the bootstrap algorithm; thus adding one
more permutation with a Wilks statistic equal to the orig-
inal grouping. This is accomplished by adding 1 to both
the numerator and denominator of Equation 2.
The LDA was also used to test the ability of normal-
ized brain volume, normalized ventricle volume, frontal
and occipital horn ratio, and head circumference to dis-
criminate between the patterns of observed hydrocepha-
lic development.
Results
Ventricular enlargement was seen in the first postin-
jection MR images obtained in hydrocephalic mice—that
is, within 2 to 3 days of injection. If the injection did
not result in a clearly hydrocephalic mouse by the first
imaging session, subsequent imaging revealed that the
mouse never became hydrocephalic. The hydrocephalic
mouse’s head often became visibly dome shaped, and
the mice generally weighed less and were more lethargic
than their littermates (Fig. 1). Volumetric reconstructions
of 3 mice—a 12-week-old normal mouse, a 4-week-old
moderately hydrocephalic mouse, and a 4-week-old se-
verely hydrocephalic mouse—are shown in Fig. 2. Meet-
ing the criteria of hydrocephalus, the substantial increase
in fluid can be seen in these reconstructions. Of the 35
mice injected, 9 successfully acquired hydrocephalus, 17
survived the injection procedure but did not acquire hy-
drocephalus, and 9 did not survive the injection.
Each hydrocephalic mouse exhibited differences in
the growth of brain and fluid. The 3 major categories of
development are shown in Fig. 3: normal controls (Rows
A and B), hydrocephalus without edema (Row C), and
hydrocephalus with edematous brain (Rows D and E).
Examples of mild, moderate, and severe hydrocephalus
are also shown in these examples (Rows D, C, and E, re-
spectively). Rows A and B show normal mice at ages 4
and 12 weeks. Row C shows a mouse with hydrocephalus
of moderate severity at 24 days of age; the lateral, third,
and fourth ventricles are all enlarged. Row D shows the
hydrocephalic mouse with the smallest ventricular vol-
ume; at 26 days all ventricles were larger than normal,
and substantial edema was present in the forceps minor
and forceps major of the corpus callosum, the external
capsule, and around the dorsal hippocampal commis-
sure. Edema was seen in the images as increased signal
intensity due to longer T2 signal values. As seen in Row
D of Fig. 3, hydrocephalus often generates a cavum sep-
tum pellucidum, or fifth ventricle.6,8,16 Row E shows the
most severely hydrocephalic mouse; at 28 days of age
the remarkable ventricular size was accompanied by
periventricular edema as well as fused lateral ventricles.
Interestingly, the fourth ventricle is not seen, suggesting
the obstruction in this mouse was at the aqueduct of Syl-
vius above the cisterna magna. In this particular case, the
ventricular system later ruptured, leading to a normaliza-
tion of brain and ventricle volume, as well as the loss of
periventricular edema (Fig. 4). The images not only show
decreased brain volume from the disappearance of edema
but a 40% increase in cortical thickness (measured from
the middle slice of the upper and lower rows shown in
Fig. 4). Unfortunately, this mouse did not survive suffi-
Fig. 1. Photograph of 4-week-old normal (left) and hydrocepha-
lic (right) mice from the same litter. Note the smaller size and dome-
shaped head of the hydrocephalic mouse.
[2]

Page 4

J. G. Mandell et al.
4
J Neurosurg: Pediatrics / Volume 6 / July 2010
ciently long after this catastrophic ventricular rupture to
permit further serial imaging.
Two very different patterns of response were seen
in hydrocephalic mice compared with normative growth
curve data (Fig. 5). In one pattern (3 mice), brain volume
growth was normal despite the accumulation of CSF and
head enlargement. In this pattern, the mice showed no
signs of periventricular edema. In the second pattern (6
mice), brain volume enlarged abnormally as parenchymal
edema increased. As defined by hydrocephalus, ventricle
volume was greater than normal in both patterns. In this
second pattern of pathological brain enlargement, the 1
case of a spontaneous rupture of the ventricular system
led to a normalization of brain volume, ventricle volume,
and cortical thickness.
A ratio of the final brain and intraventricular CSF
volumes for each hydrocephalic mouse to the estimated
age-matched normal volumes was calculated for the LDA
(Table 1). The normal volume estimations were calcu-
lated from a natural log least-squares fit to the normal
growth curve data. The coefficient of multiple determina-
tion, R2, is the ratio of the regression sum of squares to the
total sum of squares.26 The R2 values calculated for the
normal brain and ventricle volume fitted curves are 0.82
and 0.89, respectively. The groups were determined based
on the presence (Group 1) or absence (Group 2) of brain
edema. This classification corresponded with abnormal
brain enlargement (Group 1) or brain growth consistent
with normal (Group 2).
The matrix Y is the hydrocephalic brain (first column)
and ventricle (second column) volumes normalized to the
age-matched normal data. The data are classified into Y1
and Y2 corresponding to Groups 1 and 2, respectively:
Fig. 2. Volumetric reconstruction of the segmented MR images obtained in a 12-week-old normal mouse (Column A), a
moderately hydrocephalic mouse at 4 weeks (Column B), and a severely hydrocephalic mouse at 4 weeks (Column C). Blue
designates brain and red indicates ventricular fluid. A lateral view (Row 1), dorsal view (Row 2), and angled views, as seen from
the anterior (Row 3) and posterior (Row 4), of the same mice through each column is shown. Note the dramatic increases in fluid
and the more rounded shapes of the hydrocephalic brains.

Page 5

J Neurosurg: Pediatrics / Volume 6 / July 2010
Dynamics of brain and CSF growth in hydrocephalic mice
5
Fig. 3. Coronal T2-weighted MR images and related drawing. The locations of the 3 slices are indicated in the diagram at the
top. Rows A and B: The same normal mouse is shown at 4 weeks and 12 weeks of age, respectively. The lateral and third
ventricles are just seen at 4 weeks and are well defined by 12 weeks of age. Row C: A moderately hydrocephalic mouse with
normal brain volume (lacking signs of periventricular edema) at 24 days of age. Rows D and E: The mildest and most severely
hydrocephalic mice at 26 days and 28 days of age, both with substantial periventricular edema. Although edema might be due to a
transcortical pressure gradient, it is not dependent on the size of the ventricles alone. The drawing of the mouse brain is reprinted
with permission from Elsevier (Franklin KBJ, Paxinos G: The Mouse Brain in Stereotaxic Coordinates, 3rd Ed., 2007).

Page 6

J. G. Mandell et al.
6
J Neurosurg: Pediatrics / Volume 6 / July 2010
The means yj and covariance matrices ̂ ψj, for groups
j = 1,2 are calculated as follows:
Note that ̂ ψj are not unbiased13 and are calculated
with, N1 = 6 and N2 = 3. The total variability ψ is the sum
Fig. 4. Coronal T2-weighted MR images of the same severely hydrocephalic mouse at 4 weeks of age (upper) and 5 weeks
of age (lower). Each slice is 200-µm thick, and every third slice is shown. The lateral ventricles ruptured between the imaging
sessions on week 4 and 5. After the rupture, the intraventricular fluid volume substantially decreased and extraventricular fluid
volume increased. Cortical thickness increased 40% (measured from the third slice) despite a drop in total brain volume. Signs of
periventricular edema are no longer present, suggesting that a transcortical pressure gradient was responsible for the presence
of edema.
Fig. 5. Line graphs demonstrating the development of brain and ventricle volumes in all mice showing the development for
each mouse individually (A and B) and the data averaged for each group (normal [6 mice], hydrocephalic without brain edema
[3 mice], and hydrocephalic with brain edema [6 mice]) (C and D). In C and D, the volume measurements of the solitary survivor
of hydrocephalus without brain edema are shown without averaging. Ventricle volumes of all hydrocephalic mice are significantly
larger than those with normal brains, but brain volumes are only different from normal in those hydrocephalic mice with edema-
tous brain. In no case was the total brain volume of a hydrocephalic mouse smaller than normal. The data points with asterisks oc-
curred after the ventricular rupture shown in Fig. 4. After the rupture, the brain and ventricle volumes decrease toward the normal
curves. The data points marked with asterisks were excluded from the averaged graphs. The error bars in the averaged graphs
display 1 SD above and below the mean. In D, the normal ventricle volume curve is plotted at 1 SD above the mean (the error bars
are still centered around the mean) to adequately show the large differences in volume compared with the hydrocephalic mice.

Page 7

J Neurosurg: Pediatrics / Volume 6 / July 2010
Dynamics of brain and CSF growth in hydrocephalic mice
7
of the within-group, ̂ ψW and the between-group, ̂ ψB, vari-
abilities.
The prior probabilities of a measurement belonging
to Group 1 or 2 are ̂ π1 = 2/3 and ̂ π2 = 1/3. Since the rank
of ̂ ψB is m = min(p,k – 1) = 1, where p is the number of
variables (2), and k is the number of groups (2), there is at
most 1 canonical discriminant function. The 2 eigenval-
ues of ̂ ψB are λ1 = 2.104 and λ2 ≈ 0. The value of λ1 can
be interpreted as the ratio of normalized between-group
variance to within-group variance, meaning the normal-
ized between-group variance is approximately 2 times
the within-group variance. The value of λ2 confirms that
the data can be best classified by a linear combination of
only the first discriminant function. The canonical vari-
ate, z, where z = Y * ̂ γ, is defined by the coefficients of the
first discriminant function, ̂ γ:
where y1 is the normalized brain volume measurement
and y2 is the normalized ventricle volume measurement
for an individual mouse.
The hydrocephalic mice were clearly separated by
the normalized brain volume and, to a lesser extent, by
the normalized ventricle volume (Fig. 6). All of the data
points have posterior probabilities close to 100%, except
for the 2 hydrocephalic mice with abnormal brain enlarge-
ment plotted close to the discrimination line, which have
posterior probabilities of 71 and 76%. The Wilks statistic
calculated for the entire data set was 0.32, less than the
value of 0.36 expected by the chi-square distribution for
W (DF = 2, p = 0.01). Normalized brain volume alone
successfully discriminates between the 2 patterns of hy-
drocephalus (W = 0.34, p < 0.01, chi-square distribution),
whereas normalized ventricle volume alone cannot dis-
criminate these 2 patterns of hydrocephalus (W = 0.99).
The classification of data into hydrocephalus with
excessive brain enlargement or normal brain volume
growth was checked by randomly mixing the group as-
signments with a bootstrap method. Figure 7 shows the
results for 1000 random permutations of the data. The
results are shown as a histogram of the frequency of the
Wilks statistic for each permutation, and the W values of
the original data are marked with an asterisk. There were
only 29 other ways found to regroup the data to obtain a
better W statistic, leading to a 3.0% probability that the
prior classification was obtained by chance.
Measurements of frontal and occipital horn ratio and
head circumference are shown in Fig. 8. The frontal and
occipital horn ratio was able to separate normal and hydro-
cephalic mice (W = 0.23, p < 0.01 by chi-square distribu-
tion) but was not able to discriminate between the 2 pat-
terns of hydrocephalus, irrespective of whether they were
analyzed together (W = 0.90) or separated by age (W = 0.96
at 3 weeks and W = 0.68 at 12 weeks). Head circumference
was able to separate normal and hydrocephalic mice at 12
weeks of age (W = 0.16, p < 0.01 by chi-square distribu-
tion), but was unable to separate the mice at 3 weeks of
age (W = 0.84). Head circumference also failed to separate
the 2 patterns of hydrocephalus when analyzed separately
by age (W = 0.66 at 3 weeks and W = 0.58 at 12 weeks) or
when analyzed all together (W = 0.81).
Discussion
This study is, to our knowledge, the first serial quan-
tification of the growth of brain and cerebral ventricle
volumes in normal and hydrocephalic development. Hy-
drocephalus was induced by a percutaneous injection of
kaolin into the cisterna magna, and mice were imaged
with high-field MR imaging. Despite the common use of
ventricular size, cortical thickness, and ratios of ventricle
to linear brain measurements, brain volume has not been
used in an attempt to understand the development of hy-
drocephalus in animals or humans.
Two very different patterns of response were seen in
the hydrocephalic mice, each correlating with the absence
or presence of edematous brain. In the first pattern, the hy-
drocephalic mice displayed brain growth consistent with
normal control mice and did not have any obvious signs of
edematous brain. In the second pattern, the hydrocephalic
TABLE 1: Summary of data used for the Fisher LDA
Hydrocephalic (mm3)
Brain
Normal (mm3)
Brain
Hydrocephalic/Normal
BrainGroupCSFAge (days) CSF CSF
1
1
1
1
1
1
2
2
2
526.5
524.3
612.2
618.2
687.7
667.8
482.6
477.8
580.1
62.86
60.88
48.56
350.3
55.54
158.9
27.55
69.65
181.7
22
22
23
28
85
85
18
24
85
483.0
483.0
486.3
501.2
585.1
585.1
467.8
489.5
585.1
1.89
1.89
2.16
3.36
10.16
10.16
0.66
2.42
10.16
1.09
1.09
1.26
1.23
1.18
1.14
1.03
0.97
0.99
33.29
32.24
22.48
104.10
5.47
15.64
41.71
28.77
17.89

Page 8

J. G. Mandell et al.
8
J Neurosurg: Pediatrics / Volume 6 / July 2010
mice showed brain volumes increasing faster than normal
and presented with substantial edematous brain, seen as in-
creased T2 signal in the MR images. There is a well-char-
acterized linear relationship between T2 signal value and
water content.2,3 Additionally, in recent studies of kaolin-
induced hydrocephalus with infant rats and mice, increased
T2 signal was confirmed as edema by histological meth-
ods.5,19 In our study, the increased T2 signal is therefore
most consistent with extracellular edema. The edematous
tissue was seen most notably in the periventricular white
matter, as reported elsewhere.5,9,17,19,23 White matter is made
up of bundles of axons and oligodendrocytes and lacks the
astrocytic processes that hold together gray matter struc-
tures. This may allow fluid to enter and expand the extra-
cellular spaces of these structures.23 In this second pattern,
the most severe case of hydrocephalus resulted in a spon-
taneous ventricular rupture, allowing much of the fluid to
escape the ventricular system, acting therapeutically much
like an endoscopic ventriculostomy. After the rupture, the
brain volume and ventricular volume decreased toward the
normal curve, and the edema was no longer seen. Addition-
ally, cortical mantle thickness increased while brain volume
decreased. Cortical thickness can therefore change inverse-
ly to that of the brain volume it is supposed to estimate. This
case demonstrates experimental evidence of a correlation
between abnormal brain size and edema in the hydrocepha-
lic brain and suggests that a transmantle pressure gradient
was responsible for the presence of edema. However, the
presence of edema is not dependent on ventricle size alone,
as both the mildest and most severely hydrocephalic mice
both showed signs of edematous brain.
In no case did the hydrocephalic brain volume grow
at a rate slower than normal, suggesting that in these
young mice, as in humans,32 the skull was able to expand
to compensate for the increased ventricle and brain sizes.
Furthermore, these data demonstrate a clear time window
in early hydrocephalus when the brain can be expanded
without atrophy and volume loss.
The differences in the 2 responses of hydrocephalic
brain volume growth were confirmed using LDA and a
bootstrap method.12,13,26,28 The weighting of the discrim-
inant function shows that the groups are best classified
by a combination of both normalized brain and ventricle
volume. Nevertheless, ventricle volume alone is not an ef-
fective discriminator between these 2 hydrocephalic dy-
namics, demonstrating the added value of brain volume
measurements by MR imaging or CT.
Current clinical measurements of head circumfer-
ence and frontal and occipital horn ratio were also used
to analyze the data. Head circumference, as estimated in
this study by brain circumference, was able to discrimi-
nate between normal and hydrocephalic mice, but only in
the 12-week-old mice. Discrimination was not possible at
3 weeks even though the ventricle volumes were already
20–90 times the volume of normal ventricles. Based on the
results presented here, the rate at which changes in head
circumference show signs of hydrocephalus is on a time
scale much slower than one would desire in clinical prac-
tice. The frontal and occipital horn ratio was able to dis-
criminate between normal and hydrocephalic mice at both
3 and 12 weeks of age because frontal and occipital horn
ratio is an effective estimator of ventricle volume.24 Howev-
er, neither measure was able to discriminate between the 2
patterns of hydrocephalus. In this animal model, brain vol-
Fig. 6.  Graph showing hydrocephalic mice classified by the linear 
discrimination function. The brain and ventricle volumes are normalized
by the age-matched normal volumes. The black line dividing the plot is
the calculated cutoff point for classification. The open circle and square
show the mean values of each group. Although the data are best clas-
sified by a combination of the brain and ventricle volumes, the brain 
volume alone is able to discriminate between the 2 groups, whereas
ventricle volume alone is not.
Fig. 7. The bootstrap results for 1000 permutations of reclassification of the data plotted as a histogram of the Wilks statistic.
The Wilks statistics for the original data are marked with an asterisk. Only 2 unique permutations (total 29) of the data gave better
Wilks statistics, yielding a 3% probability that the data classified were obtained by chance.

Page 9

J Neurosurg: Pediatrics / Volume 6 / July 2010
Dynamics of brain and CSF growth in hydrocephalic mice
9
ume was the only effective discriminator of the 2 patterns
of hydrocephalus. Additionally, the frontal and occipital
horn index, which correlates inversely with cortical mantle
thickness, consistently increases in the hydrocephalic mice
whether the brain is increasing in size or not. Normal brain
growth is the target of clinical treatment, but in those hy-
drocephalic mice that showed normal brain growth there
was still a decreased mantle thickness. We speculate that
it is likely, however, that the hydrocephalic brain with in-
creased volume due to edema will lead to worse cognitive
development and clinical outcomes.
This present study suggests that neonatal hydrocepha-
lus can have very different patterns of volume growth. Cur-
rent evaluation methods show decreases in cortical thick-
ness and increases in ventricle to brain ratios4,5,7,17,19,22,29 but
do not reveal brain volumes directly. Our finding that the
brain increases in size faster than normal is impossible to
observe with measures currently in clinical use.
Total brain volume is of substantial importance in
understanding hydrocephalus, as the overall goal of treat-
ing pediatric hydrocephalus is to enable normal brain
development, which is not equivalent to decreasing fluid
volumes. The patient’s brain growth, water content, and
turgor should be as great a concern as the volume of CSF
within the cranial cavity. Although ventricular volume
has been considered to be the gold standard for follow-
ing hydrocephalus,24 it does not offer a direct indication
of cognitive function. We have learned to be more cau-
tious in treating patients with achondroplastic dwarfism
because most have relatively preserved cognitive function
in the setting of mildly to moderately dilated ventricles.20
Because many children with enlarged ventricles appear
to develop normally, both with and without achondropla-
sia, the standard for diagnosis and treatment becomes the
response to treatment itself.21,31 Although hydrocephalus
has historically been seen as a simple imbalance of CSF
production and absorption, more recently it is being rec-
ognized as a more complex brain disorder composed of
neuronal and glial injury.31
Conclusions
Our findings demonstrate the feasibility of construct-
ing normative curves of brain and fluid growth as comple-
ments, or alternatives, to head circumference normative
curves. By observing brain volumes in this animal model,
we were able to measure distinct patterns of normal brain
growth and abnormal brain enlargement compared with
normative growth, measures that are more likely linked
to cognitive development than fluid volumes alone. In-
deed, both excessive brain volume due to edema and
eventual decrease in brain volume due to the prolonged
effects of elevated pressure are states to avoid through
suitable clinical management. In order for brain volume
assessment to be implemented clinically in the diagno-
sis and management of hydrocephalus, the measurement
must be automated with a reliable and accurate method.
A dynamic model of brain and CSF growth in normal and
hydrocephalic humans would offer the possibility for a
more rational design of therapy aimed at optimizing brain
growth and associated cognitive capacity.
Disclosure
The authors report no conflict of interest concerning the mate-
rials or methods used in this study or the findings specified in this
paper.
Author contributions to the study and manuscript preparation
include the following. Conception and design: all authors. Ac qui si-
tion of data: Mandell, Neuberger. Analysis and interpretation of data:
all authors. Drafting the article: Schiff, Mandell. Critically revising
the article: all authors. Reviewed final version of the manuscript and
approved it for submission: all authors. Statistical analysis: Schiff,
Man dell, Drapaca, Webb. Administrative/technical/material sup-
port: Schiff, Neuberger, Drapaca, Webb. Study supervision: Schiff,
Neuberger, Drapaca, Webb.
Fig. 8. Frontal and occipital horn ratio and head circumference plotted as a function of age for normal and hydrocephalic
mice. The frontal and occipital horn ratio is able to discriminate between normal and hydrocephalic mice at 12 weeks. It was not
possible to measure the frontal and occipital horn ratio for normal mice at 3 weeks, but the 12-week normal frontal and occipital
horn ratio is still significantly different than that in the hydrocephalic mice at 3 weeks. Head circumference is able to discriminate
between normal and hydrocephalic mice at 12 weeks but not at 3 weeks. Neither the frontal and occipital horn ratio nor head
circumference is able to discriminate between the 2 patterns of hydrocephalus.

Page 10

J. G. Mandell et al.
10
J Neurosurg: Pediatrics / Volume 6 / July 2010
References
1. American Academy of Pediatrics Committee on Practice and
Ambulatory Medicine, Bright Futures Steering Committee:
Recommendations for preventive pediatric health care. Pedi-
atrics 105:645–646, 2000
2. Barnes D, McDonald WI, Johnson G, Tofts PS, Landon DN:
Quantitative nuclear magnetic resonance imaging: characteri-
sation of experimental cerebral oedema. J Neurol Neurosurg 
Psychiatry 50:125–133, 1987
3. Bederson JB, Bartkowski HM, Moon K, Halks-Miller M,
Nishimura MC, Brant-Zawadski M, et al: Nuclear magnetic
resonance imaging and spectroscopy in experimental brain
edema in a rat model. J Neurosurg 64:795–802, 1986
4. Boillat CA, Jones HC, Kaiser GL, Harris NG: Ultrastructural
changes in the deep cortical pyramidal cells of infant rats with
inherited hydrocephalus and the effect of shunt treatment.
Exp Neurol 147:377–388, 1997
5. da Silva Lopes L, Slobodian I, Del Bigio MR: Characteriza-
tion of juvenile and young adult mice following induction of
hydrocephalus with kaolin. Exp Neurol 219:187–196, 2009
6. Dandy WE: Congenital cerebral cysts of the cavum septi
pellucidi (fifth ventricle) and cavum vergae (sixth ventricle).
Arch Neurol Psychiatry 25:44–66, 1931
7. Dandy WE: Experimental hydrocephalus. Ann Surg 70:129–
142, 1919
8. Del Bigio MR: Future directions for therapy of childhood hy-
drocephalus: a view from the laboratory. Pediatr Neurosurg
34:172–181, 2001
9. Del Bigio MR, Crook CR, Buist R: Magnetic resonance im-
aging and behavioral analysis of immature rats with kaolin-
induced hydrocephalus: pre- and postshunting observations.
Exp Neurol 148:256–264, 1997
10. Drake JM: The surgical management of pediatric hydrocepha-
lus. Neurosurgery 62 Suppl 2:633–642, 2008
11. Efron B, Tibshirani RJ: An Introduction to the Bootstrap. 
Monographs on Statistics and Applied Probability, Vol 57.
New York: Chapman & Hall/CRC, 1993
12. Fisher RA: The use of multiple measurements in taxonomic
problems. Ann Eugen 7:179–188, 1936
13. Flury B: A First Course in Multivariate Statistics. New York:
Springer, 1997
14. Franklin KBJ, Paxinos G: The Mouse Brain in Stereotaxic 
Coordinates, ed 3. New York: Academic Press, 2007
15. Grogan WL Jr, Wirth WW: A new American genus of pre-
daceous midges related to Palpomyia and Bezzia (Dipteria:
Ceratopogonidae). Proc Biol Soc Wash 94:1279–1305, 1981
16. Hankin MH, Schneider BF, Silver J: Death of the subcallosal
glial sling is correlated with formation of the cavum septi pel-
lucidi. J Comp Neurol 272:191–202, 1988
17. Harris NG, Jones HC, Williams SCR: MR imaging for mea-
surements of ventricles and cerebral cortex in postnatal rats
(H-Tx strain) with progressive inherited hydrocephalus. Exp 
Neurol 118:1–6, 1992
18. Kestle JR: Pediatric hydrocephalus: current management. Neu-
rol Clin 21:883–895, vii, 2003
19. Khan OH, Enno TL, Del Bigio MR: Brain damage in neonatal
rats following kaolin induction of hydrocephalus. Exp Neurol
200:311–320, 2006
20. King JAJ, Vachhrajani S, Drake JM, Rutka JT: Neurosurgi-
cal implications of achondroplasia. A review. J Neurosurg
Pediatr 4:297–306, 2009
21. Lewin R: Is your brain really necessary? Science 210:1232–
1234, 1980
22. Li J, McAllister JP II, Shen Y, Wagshul ME, Miller JM, Egnor
MR, et al: Communicating hydrocephalus in adult rats with
kaolin obstruction of the basal cisterns or the cortical suba-
rachnoid space. Exp Neurol 211:351–361, 2008
23. McLone DG, Bondareff W, Raimondi AJ: Brain edema in the
hydrocephalic hy-3 mouse: submicroscopic morphology. J
Neuropathol Exp Neurol 30:627–637, 1971
24. O’Hayon BB, Drake JM, Ossip MG, Tuli S, Clarke M: Frontal
and occipital horn ratio: a linear estimate of ventricular size
for multiple imaging modalities in pediatric hydrocephalus.
Pediatr Neurosurg 29:245–249, 1998
25. Raimondi AJ, Bailey OT, McLone DG, Lawson RF, Echev-
erry A: The pathophysiology and morphology of murine hy-
drocephalus in Hy-3 and Ch mutants. Surg Neurol 1:50–55,
1973
26. Rencher AC: Methods of Multivariate Analysis, ed 2. New
York: John Wiley & Sons, 2002
27. Rizvi R, Anjum Q: Hydrocephalus in children. J Pak Med 
Assoc 55:502–507, 2005
28. Schiff SJ, Sauer T, Kumar R, Weinstein SL: Neuronal spa-
tiotemporal pattern discrimination: the dynamical evolution
of seizures. Neuroimage 28:1043–1055, 2005
29. Slobodian I, Krassioukov-Enns D, Del Bigio MR: Protein and
synthetic polymer injection for induction of obstructive hy-
drocephalus in rats. Cerebrospinal Fluid Res 4:9, 2007
30. Tada T, Kanaji M, Kobayashi S: Induction of communicating
hydrocephalus in mice by intrathecal injection of human re-
combinant transforming growth factor-beta 1. J Neuroimmu-
nol 50:153–158, 1994
31. Williams MA, McAllister JP, Walker ML, Kranz DA, Bergs-
neider M, Del Bigio MR, et al: Priorities for hydrocephalus re-
search: report from a National Institutes of Health-sponsored
workshop. J Neurosurg 107 (5 Suppl):345–357, 2007
32. Zahl SM, Wester K: Routine measurement of head circumfer-
ence as a tool for detecting intracranial expansion in infants:
what is the gain? A nationwide survey. Pediatrics 121:e416–
e420, 2008
Manuscript submitted January 6, 2010.
Accepted April 1, 2010.
Portions of this work were presented in poster form at the 50th
Experimental Nuclear Magnetic Resonance Conference, Pacific
Grove, California, March 29–April 3, 2009.
Portions of this work were presented as an oral presentation at
the 38th Annual Meeting of the AANS/CNS Section on Pediatric
Neurological Surgery, Boston, Massachusetts, December 2, 2009.
Address correspondence to: Steven J. Schiff, M.D., Ph.D., 212
Earth–Engineering Sciences Building, University Park, Penn syl-
vania 16802. email: sschiff@psu.edu.