Brain Research Protocols 8 (2001) 45–57

www.elsevier.com / locate / bres

Protocol

A simple stereologic method for analysis of cerebral cortical

microvessels using image analysis
Panya S. Manoonkitiwongsa a , Paul J. McMillan b , Robert L. Schultz b ,
a a,
Catherine Jackson-Friedman , Patrick D. Lyden *
a
Department of Neurosciences, University of California, San Diego, School of Medicine, and Neurology, Veterans Administration Medical Center,
3350 La Jolla Village Drive, San Diego, CA 92161, USA
b
Department of Pathology and Human Anatomy, Division of Human Anatomy, Loma Linda University, Loma Linda, CA, USA

Accepted 8 June 2001

Abstract

Previous methods for determining morphological features of vascular networks in cerebral cortex were subject to arbitrary variation and
bias. Unbiased estimates of vessel number, volume, surface area and length can be obtained using stereology but these techniques tend to
be tedious and time-consuming. Stereologic protocols generally require micrographs that have to be analyzed manually for intersections of
vessels on grid points or lines. In this report, we provide a simpler and more precise method for measuring morphological features of
cerebral cortical microvessels. Images of microvessels in 1 mm toluidine blue stained sections were captured using a popular image
analysis software package. Luminal surfaces of endothelial cells were automatically traced using commonly available features; the
two-dimensional data of vessels (diameter, area, perimeter and number of vessels) were automatically computed and transferred to a
spreadsheet. Three-dimensional features were then determined using basic stereologic equations. The method eliminates the need for
manual measurements and is particularly time- and cost-effective for quantitative studies where numerous images have to be evaluated.
 2001 Elsevier Science B.V. All rights reserved.

Theme: Cellular and molecular biology

Topic: Staining, tracing, and imaging techniques

Keywords: Image analysis; Stereology; Cerebral cortex; Microvessel; Rat

1. Type of research • Surface density (SV ), defined as the total surface of

The protocol presently described is suitable for:
(a) Quantitation of cerebral cortical microvessel features
which include:
• Profile density (NA ), defined as the total number of
microvessel profiles per unit area of tissue ([ of
microvessels / mm 2 ).
• Volume fraction (VV ), defined as the total microvessel
volume per unit volume of tissue (%).
• Length density (LV ), defined as the total length of
microvessels per unit volume of tissue (mm / mm 3 ).
*Corresponding author. Tel.: 11-619-543-7760; fax: 11-619-543-
7771.
E-mail address: plyden@ucsd.edu (P.D. Lyden).
microvessels per unit volume of tissue (mm 2 / mm 3 ).
• Diameter (D), defined by the mean short-axis of the
lumina of microvessels (mm).
(b) Quantitation of microvessel features of organs with
isotropic vessels (vessels having the same properties in all
directions of space) or organs with anisotropic vessels
(vessels having different properties in different directions
of space) where the sections are isotropic, uniform random
(IUR).
2. Time required
Total time required from tissue preparation to analysis
of 240 images (four animals; six Epon blocks per animal;

1385-299X / 01 / $ – see front matter  2001 Elsevier Science B.V. All rights reserved.
PII: S1385-299X( 01 )00087-3
46 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
one semithin section per Epon block and 10 images per
semithin section) by a trained operator is 5 days.
2.1. Tissue preparation and embedding in Epon: 3.5
days ( four animals; 24 Epon blocks)
(a) Intracardial perfusion followed by overnight immer-
sion fixation: 24 h.
(b) Washing of tissues with buffer: 8 h.
(c) Osmication: 3 h.
(d) Rinsing, dehydration, infiltration and embedding in
Epon: 24 h.
(e) Polymerization of the Epon: 24 h, 608C.
2.2. Preparation of one semithin section of an Epon
block, selection and capture of 10 microvessel images:
40 min
(a) Trimming of an Epon block, cutting and staining of a
1-mm section: 30 min.
(b) Systematic random sampling of 10 cortical regions
in the 1-mm section and imaging of microvessels by IPP:
10 min.
2.3. Morphometric and stereologic procedures: 7 min
for 10 images
(a) Tracing of the luminal surfaces of microvessels and
transfer of data to spreadsheet: 5 min.
(b) Calculation of microvessel morphological features
using stereological equations: 2 min.
2.4. Statistical analysis: 35 min (240 images from 24
Epon blocks)
(a) Creating a spreadsheet summarizing all data obtained
(profile density, volume fraction, surface density, length
density and diameter of microvessels from the frontal,
middle and occipital regions of the cortex, along with
medial-dorsal and lateral cortices of each region): 30 min.
(b) Analysis of Variance (ANOVA) using Statistical
Packages for the Social Sciences (SPSS) comparing the
frontal, middle and occipital regions of the cortex, along
with medial–dorsal and lateral cortices of each region: 4 h.
3. Materials
3.1. Animals
Four adult male Sprague–Dawley rats, purchased from
Harlan, San Diego, CA, at the age of 6 weeks, were used
in this study. The animals were fed commercial Rat Chow
pellets and tap water ad libitum, and housed in paper
shavings in a room maintained at 22628C and relative
humidity of 5565% with alternating 12-h light–dark
schedule (lights on 07:00–19:00 h). The animals were
maintained under routine laboratory conditions for at least
2 weeks before use and weighed from 250 to 450 g.
3.2. Special equipment
• Interactive Image Analysis System (Image Pro-Plus,
version 4.0, Media Cybernetics, MD, USA) installed
on a Pentium II PC.
• Videocamera (Sanyo CCTV Camera, VDC-2524,
Chatsworth, CA, USA) attached to a light microscope
(Olympus CH2, Model BHT, Olympus Optical Co.,
Ltd., Tokyo, Japan).
• Excel software (Microsoft, Redmond, WA, USA).
• Ultramicrotome (MT-2, Ivan Sorvall, Inc., CT, USA).
• Perfusion pump (Harvard Apparatus, Boston, MA,
USA).
• Rat brain mold (1 mm slicer) (Braintree Scientific Inc.,
Braintree, MA, USA).
3.3. Standard chemicals and materials
• Urethane (Sigma, St. Louis, MO, USA).
• Sterile saline solution (Baxter Healthcare, Corp., Deer-
field, IL, USA).
• Paraformaldehyde, glutaraldehyde, sodium cacodylate,
Epon 812, dodecenyl succinic anhydride (DDSA),
nadic methyl anhydride (NMA), 2, 4, 6-Tris (di-
methylaminomethyl) phenol (DMP-30), propylene
oxide, osmium tetroxide, Azure II, small Stendor / Prep
dish, 133 inch light microscope slides, ]12 inch double-
sided tape, 22322 mm cover glass (Electron Micro-
scopy Sciences, Fort Washington, PA, USA).
• Ethyl alcohol (200-proof; Aaper Alcohol & Chemical
Co., Shell Byville, KY).
• Toluidine blue, Better Equipment for Electron Micro-
scopy (BEEM) capsules (Ted Pella Inc., Redding, CA,
USA).
• Glass beakers (300 ml), hot plate (VWR Scientific,
Cerritos, CA, USA).
4. Detailed procedure
4.1. Tissue preparation
(A) Anesthetize animals with 50% aqueous urethane
(0.003 ml / g body weight; i.p. injection) and perfuse
intracardially with 15 ml saline followed by 300 ml of a
fixative containing 2% paraformaldehyde and 3% glutaral-
dehyde in 0.1 M sodium cacodylate (pH 7.4). Both the
saline and fixative solutions are used at room temperature.
Insert the cannula into the ascending aorta so that the tip of
the cannula is slightly inferior to the arch of the aorta.
Allow the pressure at the tip of the cannula to approach
normal rat arterial pressure of 100–120 mmHg or slightly
 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57 47

higher during the first 15 s of perfusion. After 15 s, reduce
the perfusion to 30 ml / min to conserve fixative. The
duration of the perfusion is about 8 min.
(B) Remove the brains from the skulls immediately after
perfusion and immerse the whole brains in the fixative
overnight (immersion fixation).
(C) For proper infiltration with the embedding medium
while allowing for a relatively large tissue size, trim the
cortical regions of interest into 43231 mm 3 blocks. These
tissue dimensions give a relatively large area of cortex for
microvessel analysis and include all six layers of the cortex
with the long dimension perpendicular to the cortical
surface. However, for our studies we used whole coronal
slices of brain tissues (Fig. 1). With coronal slices, any
region of the brain could be cut out following osmication,
dehydration and infiltration (just prior to embedding).
Using a rat brain mold (a device similar to the common
household ‘egg slicer’), cut the entire rat brain equally
from front to back into coronal slices of preset thickness.
We used a mold that cuts the brain into 1-mm-thick
coronal slices. We compared microvessel networks of the
left frontal, middle and occipital cortices, including the
medial–dorsal (parietal motor) and the lateral (parietal
somatosensory) regions of the frontal, middle, and occipi-
tal cortices. We accomplished this by obtaining slices from
2, 7 and 13 mm post-frontal pole of the brain. Selecting
these cortical areas allow for an understanding of the
manner in which cortical microvessels are distributed
across the brain from front (frontal) to back (occipital) and
from the top (dorsal) to the side (lateral) of the cortex.
(D) Rinse the tissue blocks or slices in four changes of
0.1 M sodium cacodylate buffer, pH 7.4 (2 h each change).
(E) Postfix in 0.1 M sodium cacodylate buffered 1%
osmium tetroxide, pH 7.4, for 3 h.
(F) Rinse with distilled water (two changes; 15 s each
change).

Fig. 1. Diagram demonstrating the sequential steps for obtaining brain slices and cortical tissues for analysis of microvessels. (A) Coronal slices were
obtained from the frontal, middle and occipital regions (2, 7 and 13 mm post-frontal pole, respectively) of each brain. (B) In the coronal slice (diagram
shown is of a slice taken from 7 mm post-frontal pole), two areas of the left cortex were evaluated: the medial–dorsal (parietal motor) and lateral (parietal
somatosensory) cortices. (C) In the 43231 mm 3 tissue section obtained from each cortical region, 10 images were selected using systematic random
sampling from cortical Layers II and V. (D) In each image, the microvessels appear as white, circular-like profiles against the darker stained neuropil.
48 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
(G) Dehydrate, infiltrate and embed in plastic resin.
Dehydration and infiltration were done at room tempera-
ture.
• 50% ethanol, 15 min; 75% ethanol, 15 min; 95%
ethanol, 15 min; 100% ethanol, 30 min (2X); pro-
pylene oxide, 30 min (2X); Epon / propylene oxide
(1 / 1), 6 h; Epon / propylene oxide (2 / 1), 6 h; full
Epon, 6 h.
• Embed the tissue blocks in BEEM capsules.
• Place the label for block identification in the BEEM
capsule. Use pencil to write the tissue identification
on the label. Pens are not recommended since Epon
would remove the ink.
• After attaching the double edge sticky tape to a
slide, place the tissue blocks on the tape. Space the
tissue blocks sufficiently apart from each other in a
row.
• Fill the BEEM capsules with Epon and invert the
capsules on to each of the tissue blocks on the
slide.
• If brain slices were processed (as we did in our
study), then the 43231 mm 3 blocks embedded in
the inverted capsules on the slide would have been
taken from the cortical regions of interest (Fig. 1).
• Polymerize the Epon; 24 h, 608C.
4.2. Preparation of semithin sections
(A) Trim the Epon blocks so that the trapezoidal block
face includes the entire tissue piece. Cut 1-mm sections
from the Epon blocks with the ultramicrotome. We took
only one section from each Epon block.
(B) Stain sections with aqueous toluidine blue (1%
toluidine blue plus 1% azure II in distilled water). For best
staining of sections, warm the toluidine blue stain to
approximately 508C in a small Stendor / Prep dish placed
on the hot plate. Float the 1-mm sections on the warm stain
for 1–2 min. Use a wire loop to transfer the sections from
the knife boat to the stain. Immerse the wire loop into the
water and position the loop immediately below the section
so that section is in the center of the loop. Remove the
loop vertically from the water. Use a loop whose size is
just slightly larger than the section. This will allow the the
water in the loop (when the loop is lifted vertically from
the water) to have sufficient surface tension to keep the
section floating flat on the water in the loop. Transfer the
section in the loop to the stain by immersing the loop
directly down into the stain. The section would be removed
from the loop and float flatly on the stain. Since the lumens
of the blood vessels contain only Epon (which does not
pick up the stain), the vessels appear as conspicuous white,
circular-like profiles against the darker stained neuropil
(Fig. 1). Following staining, pick up the section from the
stain with the wire loop in the same manner as picking up
the section from the knife boat. Float the section on
distilled water in a 300-ml glass beaker by immersing the
loop directly down into the water in the same manner as
the loop is immersed into the stain when transferring the
section from the knife boat to the stain. Use two beakers so
the sections are rinsed twice. Pick up the section with the
loop from the second beaker and place the section on the
slide. This is accomplished by placing the loop on the slide
and then touching a small piece of filter paper to the
bottom of the loop wire. The downward suction caused by
the filter paper would land the section squarely and flatly
on the slide. Allow the section to dry on a hot plate (508C)
briefly. Place a cover slip on the section.
4.3. Use of IPP versus another image analysis software
We used IPP as our image analysis system since it is one
of the most popular systems used currently by inves-
tigators. The following detailed methods (Sections 4.4–
4.10) for calibrating the features of the software, along
with techniques for measuring morphologic characteristics
of microvessels, are unique to IPP. This does not, however,
exclude the use of other image analysis systems available
commercially. The objectives of each step for calibrating
and using particular features of IPP are mentioned in each
of the following section and / or subsection. If another
image analysis system is used, the features of that system
should be calibrated and used in the manner so as to
achieve the objectives stated.
4.4. Calibrating IPP features for morphometric analysis
IPP settings need to be properly established to enable
morphological evaluation of microvessel network. This
needs to be done only once. After the settings are
established, they could be saved and quickly loaded prior
to analysis.
4.4.1. Establishing grid mask, counting and data
collection settings
(A) Open any image (File / Open . . . ; select appropriate
file path to the image). An image has to be opened for the
menu options to be active.
(B) Open Measure / Count / Size from the primary menu.
In the Count / Size dialog box, choose Measure / Select
Measurements. Select the following in order:
Area 0 722 626.9
Axis (minor) 0 268 817.2
Perimeter 0 268 817.2
The two numbers for each parameter represent the low
(Start) and high (End) amounts for each parameter. We
arbitrarily chose these low and high numbers since they
allow sufficient range for each of the above parameters for
our study. For the high (End) numbers, the numbers
 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
represent the maximum limit of measure IPP would allow
for that particular parameter. Thus, for area, 722 626.9
mm 2 is the maximum area measurable. For axis (minor)
and perimeter, 268 817.2 mm is the maximum length
measurable for the parameters. It is advisable to first
determine the general size distributions of the objects of
interest. The high (End) numbers for each parameter could
then be appropriately set. Close the dialog box.
(C) Under Edit of the Count / Size dialog box, select
Draw / Merge Objects . . . In the Draw / Merge Objects . . .
dialog box, choose the Magic Wand option (rather than
Trace). Enter ‘45’ for Range and ‘3’ for Smooth. We chose
these numbers since they best allow IPP to automatically
trace the luminal surface of the microvessels with the kind
of staining we achieved with our staining method.
(D) Close the Draw / Merge Objects . . . dialog box. In
the Count / Size dialog box (which appears when the Draw /
Merge Objects . . . dialog box is closed), select
Options . . . In the Display options, we chose ‘Filled’ for
Outline Style and ‘Object [’ for Label Style, and green for
Label Color. In the Objects options, select the 4-Connect
and Fill Holes options. Select ‘None’ for Clean Borders.
These settings cause the microvessels to appear as filled
structures with the selected color once the vessels are
selected.
(E) Close the Options . . . dialog box. Save the settings.
In the Count / Size dialog box (which appears when the
Options . . . dialog box is closed), select File / Save Set-
tings and save the settings to an appropriate file. Before
analysis of the images, open the Count / Size . . . dialog
box (Measure / Count / Size . . . ) and load the settings (File /
Load Settings and then choose the file the settings were
saved in).
(F) Open Measure / Data Collector from the primary
menu. In the Layout tab of the Data Collector dialog box,
select the following Data Items from the Data Source to
the Selected Items box in the following order:
Data source Data items
1. Image Name
2. Count Size Object Count
3. Count Size Area (values)
4. Count Size Perimeter (values)
5. Count Size Axis (minor)
(G) Under the Export tab, select the Data List, With
Column Headers and Excel (DDE) options by clicking in
the small box beside each option. Click on the DDE
options . . . button. In the Dynamic Data Exchange dialog
box, enter ‘Excel’ for Target Program, appropriate file path
to the Excel spreadsheet for Path, ‘Active Sheet’ for Sheet,
‘1’ for Row and Column. Select the option ‘Append next
data set to the bottom’ by clicking in the small box beside
the option.
(H) Close the Dynamic Data Exchange dialog box.
Select Options in the Data Collector dialog box. Choose
49
‘Always append new data to the end of the last collected
block’, ‘Collect from any image active at time of collec-
tion’, ‘Row number’ and ‘Show module name in column
headers’ by clicking in the small box beside each option.
For Column Width and Significant Digits, we entered ‘20’
and ‘5’ characters, respectively. All these steps (F–H)
determine what data IPP would collect from the vessels,
where to transport the data to (the Excel spreadsheet) and
how the data would be displayed in Excel.
(I) Close the Data Collector dialog box. Open Process /
Grid Mask . . . from the primary menu. In the Grid
Mask . . . dialog box, press the ‘New’ button. This creates
a new grid file. Select the Grid tab and choose ‘Lines’ for
Objects and ‘Orthogonal’ for Layout. For Spacing, we
chose ‘19’ for both horizontal and vertical grid lines.
Select the Settings tab. We chose ‘6’, ‘74’, ‘85’ and ‘14’
for the left, right, top and bottom margins, respectively. All
these settings create a 29320 grid (580 grid points). Each
grid square is 535 mm 2 for our study (see Spatial
Calibration; Section 4.4.2). Enter ‘1’ for Width. Increasing
the numbers increases the width of each grid line. Choose
‘pixels’ for Measure Units.
4.4.2. Spatial calibration
(A) Place the slide with calibration scale on the micro-
scope. We used a scale with 25-mm spacing. Select the
403 objective lens and image the scale on IPP (Acquire /
Video / Digital).
(A) Open Measure / Calibration / Spatial . . . In the Spa-
tial Calibration dialog box, press the ‘New’ button. This
creates a new calibration file. Name the file as desired. For
Unit, choose ‘mm’.
(C) Click on the Image button of the Pixels / Unit option.
This creates a line on the image. Place the line on the scale
and enter the appropriate calibration unit in the Scaling
dialog box. For our study, we extended the line from one
edge of the 25-mm scale to the other and entered ‘25’ in
the Scaling dialog box. Leave the Aspect Ratio, Origin
(Pixels) and Angle Offset at default (‘1’, ‘0’, ‘0’, respec-
tively). Press the ‘OK’ button. IPP will use this spatial
calibration to measure the microvessels any time the file
with this calibration is active.
4.5. Determining final magnification seen on IPP
The final magnification imaged by IPP is dependent on
the particular: (1) objective lens of the microscope, (2)
relay lens (situated inside the light microscope between the
objective lens and the attachment for the videocamera) and
(3) model of the videocamera used. Thus, it should not be
assumed that the final magnification of IPP with a par-
ticular microscope set-up would produce the same final
magnification with another microscope set-up even though
the same objective lens is used. To determine the final
magnification of a particular microscope-IPP set-up, per-
form the following steps.
50 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
(A) Properly calibrate IPP (see Section 4.4.2 for the
‘Spatial Calibration’ procedure).
(B) If the image of the calibration scale has been saved
(from Section 4.4.2), open the image (File / Open . . . ;
select appropriate file path to the image). If the image of
the calibration scale was not saved, place the calibration
slide on the microscope stage. Select an objective lens
intended for use (we used 403 for our study; see Section
4.4.2). Capture the image of the calibration scale by IPP
(Acquire / Video / Digital).
(C) Place the mouse cursor anywhere on the image and
click the right mouse button. This opens a special menu
box allowing for different zooming and magnification
options. Chose the zoom power that best shows the objects
of interest (for our study of microvessels we chose the
zoom power of 100%. This allowed for sufficient viewing
of the microvessels. Zoom power of 200% creates an
image that covers too much of the monitor screen while
zoom power of 50% creates an image that is too small for
viewing the microvessels. Maximize the image by clicking
on the ‘maximize’ button on the upper right corner of the
image. This will allow for full viewing of the image with
the particular zoom power selected.
(D) Measure the length of the calibration scale, as
imaged by IPP, for the particular zoom power chosen (we
used a small transparent ruler). Convert the unit of the
length to the same unit as that on the calibration slide.
(E) Divide the length of the scale seen on the screen by
the actual length of the scale on the slide. This gives the
final magnification seen on IPP. With the 403 objective
lens, the 13 relay lens and the video camera model used in
our study, the length of the scale imaged by IPP at zoom
power of 100% is 2.9 cm (which is 29 000 mm). Since the
actual distance of each scale bar on the slide is 25 mm, the
final magnification imaged by IPP for our study is there-
fore 11603 (29 000 mm / 25 mm).
4.6. Systematic random sampling of microvessels in
Layers II and V of the cortex
(A) Place the slide with the tissue section in the
microscope.
(B) Observe the section under low power (43). Bring
any side edge of the trapezoidal-shaped section to the
center of the field of view. Arbitrarily choose the edge of
the section closest to the upper left corner of the field of
view as the starting point for sampling the first image from
the section. This edge will be either towards the superficial
aspect of the cortex (Layer 1) or towards the corpus
callosum (Layer VI) (Fig. 1). If the section edge is towards
the superficial aspect of the brain, move the section so that
the edge of the trapezoid is in line with Layer II. If the
edge is towards the corpus callosum, move the section so
that the edge of the trapezoid is in line with Layer V.
(C) Switch the objective lens to 403. Place the edge of
the section in the center of the field of view. Then move
the section slightly in line with the respective cortical layer
so that the neuropil fills the entire field of view. This is the
first cortical area for imaging microvessels. Since the
1-mm section was arbitrarily selected from the region of
interest, this is considered to be a random starting point.
(D) Image the section on IPP (Acquire / Video / Digital).
Label and save the image with an appropriate file name
(e.g. ‘tissue ID; 1st image’). Save the image as a ‘TIFF’
file. Make sure that the image is saved with the proper
spatial calibration (see Section 4.4.2), not in ‘pixels’.
(E) If the first image is captured in Layer II, move the
frame towards the corpus callosum by a set distance (i.e.
one, two or three frames). We chose a fixed interval of two
frames between images since, given our IPP set-up, this
distance would situate the frame approximately at Layer V.
This is the second cortical area for imaging the mi-
crovessels (Fig. 1). Label and save the image (e.g. ‘tissue
ID; 2nd image’).
(F) Move the section in line with Layer V by the set
interval distance (two viewing fields for our study) towards
the other edge of the section. This is the third area for
imaging. Label and save the image (e.g. ‘tissue ID; 3rd
image’).
(G) Move across the neuropil back towards the superfi-
cial cortex by the set interval distance. For our study, the
set interval of two viewing fields would situate the frame
at approximately Layer II (see Section 4.6 E). This is the
fourth area for imaging. Label and save the image (e.g.
‘tissue ID; 4th image’).
(H) Move the section in line with Layer II by the set
interval (two viewing fields for our study) in the same
direction as in step (F) above. This is the fifth area for
imaging. Label and save the image (e.g. ‘tissue ID; 5th
image’).
(I) Keep selecting the images in this manner until 10
images are captured (for our study, this would be five
images each from Layers II and V (Fig. 1). Such a
sampling of images with fixed intervals between each of
the images constitutes a systematic sampling procedure.
(J) We arbitrarily chose a distance of two viewing fields
between the vertical images since we wanted to sample
microvessels from the two limits of the cortex that contain
isotropic vessels (Layers II and V; vessels in Layers II–V
are isotropic). Capturing images only from Layers II and V
provided a simple way to systematically select the images.
It also serves as a simple example to illustrate the concept
of systematic sampling. Furthermore, our preliminary
studies capturing images from all layers between II–V
versus Layers II and V, as done in this study, showed
similar results (data not shown). If one prefers a sampling
procedure where all the layers between Layers II–V are
included, select the images from all the layers by applying
the principle of systematic random sampling discussed
above (i.e. randomly choose the first sample and then
 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
select the other samples with a set interval between the
samples).
For our study, the tissue slide on the microscope stage
was moved manually. We did not use a motorized micro-
scope scanning stage though such a device is more
recommendable than manual movement since the distance
between the frames could be programmed and thus more
accurately maintained.
4.7. Morphometric procedures
The following steps are done to analyze each image and
transfer the data to Excel. These steps may appear arduous
at first. However, these steps are greatly simplified when
macros are created (see Section 4.10).
(A) Open a new spreadsheet in Excel (File / New . . . ;
under the ‘General’ tab double click ‘Workbook’ icon).
(B) Switch to IPP (click on ‘IPP’ box in status bar).
Open the 1st image (File / Open . . . ; select pathway to file
where the 10 images are saved). If spatial calibration of the
image appears as ‘pixels’ in the status bar, change to
proper spatial calibration (Measure / Calibration / Spatial;
choose the proper spatial calibration (see Section 4.4.2)).
View the section. If the tissue is properly stained, mi-
crovessels are easily observed appearing conspicuously as
white circular-like structures against a darker neuropil
background (Fig. 1).
(C) Open Process / Grid Mask . . . Choose the appro-
priate grid mask (see Section 4.4.1) and overlay the grid on
the image by clicking on Apply button in the dialog box.
(D) Open the Draw / Merge Objects tools (Measure /
Count / Size . . . ; in the Count / Size . . . dialog box choose
Edit / Draw / Merge Objects). Choose the ‘Magic Wand’
option. Select the microvessels to be included in the
analysis by clicking on them with the mouse cursor. IPP
automatically outlines the luminal surface of the microves-
sel when its lumen is selected and gives its area and
boundary length. For our study we included all vessels
within the grid. Only the outer outline of the grid is needed
when using IPP since, with the IPP approach, the equations
for deriving the stereologic features of the microvessels do
not require intersections of the vessel profiles with the grid
points or lines (see equations used with the IPP method in
Section 4.9 B). The outer outline of the grid serves simply
to demarcate the area for selecting the vessels. The grid
points and lines inside the grid outline were used only in
the manual protocol since, with the manual approach, the
equations for deriving the stereologic features of the
vessels depend on intersections of the vessel profiles with
the grid points and lines (see the equations used with the
grid method in Section 6.3.2).
Microvessels were of all shapes (round, oblong, curved,
ovoid, short and long straight profiles). For those micro-
vessels whose profile touched any of the borders of the
51
grid, the exclusion rule should be followed. That is, any
microvessel profile that touches the left border and / or its
upward extension or the bottom border and / or the down-
ward extension of the right border is excluded. This
assures counting each microvessel profile landing on the
border of the grid only once. After the vessels of interest
have been selected, click the ‘O.K.’ button.
4.8. Transfer of data to spreadsheet and application of
stereologic equations
(A) Open Measure / Data Collector. Click on the ‘Collect
Now’ button. This gathers the measurements to the Data
Collector feature. Transfer the data to Excel by selecting
the Export tab and then clicking on the ‘Export Now’
button).
(B) Switch to Excel by clicking on the Excel box in the
status bar. Data obtained from the selected vessels are
shown in the spreadsheet. Columns A–E show the ID,
object count, area, perimeter and axis (minor) of each
vessel in the image, respectively.
(C) Switch back to IPP by clicking on the IPP box in the
status bar. Repeat steps 4.7 B to 4.8 A for the second and
the rest of the remaining images from the section.
4.9. Application of stereologic equations to data
(A) Switch to Excel by clicking on the Excel box in the
status bar. Data from all the images are shown in the
spreadsheet. Sum the data of all the images for columns B
to D. These columns contain the data for ‘Object Count’,
‘Area’ and ‘Perimeter’, respectively (column A contains
the ID’s of the images). For column C (Area), divide the
6 2
sum by 10 . This converts the value in mm to mm . For
2
column D (Perimeter), divide the sum by 10 . This
3
converts the value in mm to mm. Stereologic equations
assume the units of the values to be in mm.
(B) Apply the following stereologic equations (which
assume a random distribution and, in the case of SV and
LV , a random orientation) to obtain the NA , VV , SV and LV
of microvessels:
• Column B (Object Count) for profile density:
(NA ) 5 sum of column B / total area of the images (mm 2 )
For our study the total area, in mm 2 , of the 10 images
is 0.145 mm 2 (0.0145 mm 2 per image 310 images);
• Column C (Area) for volume fraction:
VV 5 (sum of column C / total area of the images [mm 2 ])
3100
52 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57

• Column D (Perimeter) for surface density:

5.1060.17 5.1460.17
5.3760.24 5.4260.26
5.3860.50 5.3360.30
Each mean6S.D. value from the medial–dorsal and lateral cortices is the average obtained from four animals. The value in each of the ‘Average’ columns is the combined mean6S.D. of the
Average
SV 5 (4 /p ) (sum of column D/ total area of images (mm 2 )

• For length density:

Medial–dorsal Lateral
cortex
LV 5 2 (NA )

8356105 8006100 5.1860.27

8006117 862694 5.4860.53
8626105 900678 5.2860.13
Diameter

(C) Calculate the average and S.D. of the values in

cortex
(mm)

column E (from the 10 images). This column contains the

data for ‘Axis (minor)’.
Average

4.10. Creating macros for quick measurements

Medial–dorsal Lateral

After some experience with using image analysis to

cortex

understand how the different features work, create macros

to significantly speed up the microvessel measurements
Length density
(mm/mm 3 )

and data collection procedures (Sections 4.7 and 4.8). For

12.7662.15 12.0861.62 7666161
12.3262.25 13.1161.93 9256127
13.1662.21 14.1061.61 938658

our study, we created four simple macros. More compli-

cortex

cated macros could be created to further reduce the effect

but they require knowledge of Auto-Pro, a Visual Basic
programming language for IPP.
Average

medial–dorsal and lateral cortical regions. Data were obtained from Layers II and V of the left cortex of each brain.

For our study, the first macro opens the Grid Mask and
the Count / Size dialog boxes and applies the grid on the
image (Section 4.7 C). The second macro opens the Data
Medial–dorsal Lateral

Collector, collects the data and transports the data to Excel

cortex

(Section 4.8 A–C). The third macro closes all the dialog
Surface density

boxes and the image but maintains connection with the

(mm 2 /mm 3 )

1.7660.33 1.6660.27 11.4162.65

1.8460.37 1.9260.30 13.9161.65
1.8360.41 2.0160.21 15.0561.68

Excel spreadsheet so data from more images could be

transported to the same spreadsheet (to gather a total of 10
cortex

images for the spreadsheet). The fourth macro clears all

the data gather by the Data Collector so the entire process
Average

could start all over from the beginning on another spread-

sheet.
(A) Open Macro / Record Macro . . . In the Record
Macro dialog box, name the macro as desired (e.g. Macro
Medial–dorsal Lateral Average Medial–dorsal Lateral
cortex

1 to apply grid). Press the ‘OK’ button. The macro would

automatically start to record every step performed in IPP
Volume fraction

from that point on until the ‘Stop Recording’ button is

417652 400650 1.5760.48
400658 431647 2.0060.24
431653 450639 2.1960.28

selected.
Morphological features of cortical brain microvessels

cortex

(B) Perform the sequential steps for analyzing the image

(%)

(Sections 4.7 and 4.8) and create the four macros as

mentioned above.
region post- ([ of microvessels/mm 2 )

cortex

5. Results
Cortical brain Numerical density

Data of the NV , VV , SV , LV and D of the microvessels are

383681
462664
469629
cortex

shown in Table 1. The Table shows the mean and S.D.

values of the medial–dorsal and lateral cortices from the
left frontal, middle and occipital cortical regions. No
frontal pole

significant difference exists between morphological fea-

Table 1

13 mm
2 mm
7 mm

tures of vessels of the three brain regions nor between the

medial–dorsal and lateral cortices of each of the regions.
 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
6. Discussion
6.1. Benefits of using the image analysis technique
Previous methods for determining morphological fea-
tures of cerebral cortical microvessels relied on the use of
stereologic grid techniques and vascular casts [4,6,10–
14,19,24]. Though data gathered from these techniques
have contributed much to the understanding of microvessel
three-dimensional architecture, these procedures tend to be
time-consuming and / or expensive. Traditional stereologic
grid methods generally required numerous photographs of
the microvessels. Grid points or lines were then overlaid
on the photographs and intersections of the vessels with
the points or lines manually counted. Though use of
vascular casts appears to be superior to the grid method by
providing an actual three-dimensional layout of the micro-
vessel network [5,15,16,25], the procedure requires use of
expensive equipment and specialized scanning electron
microscopy skills. Three-dimensional data of microvessels
obtained from the vascular casts are also similar to those
obtained from the grid methods. Thus, a new procedure
that not only simplifies the stereologic grid method but
also offers more precision in measuring the microvessel
morphological features is helpful, particularly for studies
where numerous images have to be evaluated. We therefore
took advantage of a popular image analysis system in
creating a simple stereologic method for analyzing mor-
phological features of cerebral cortical vessels.
6.2. Troubleshooting
6.2.1. Use of anesthetic agent and fixation method
Optimal preservation of nervous tissue morphology is
critically dependent on the use of proper anesthetic agents
and fixation method. We chose urethane as our anesthetic
agent since our preliminary studies revealed that use of this
anesthetic agent avoided formation of pyknotic neurons
commonly seen with halothane and ketamine–xylazine.
Urethane also does not cause the vessels to close or
collapse as with halothane. The amount of saline used prior
to the fixative solution is important. We used only 15 ml of
saline because our preliminary data revealed that a greater
volume results in pyknotic neurons and damaged neuropil
indicated by increased numbers of artifactual spaces in the
brain tissue. This volume of saline is also considered
optimal for washing out the blood from the vascular
system prior to fixation [21].
Our fixative solution is a half-strength modification of
the original Karnovsky’s fixative [8] and is similar to the
fixative used by other investigators studying cerebral
microvessels [6,7,12]. This fixative solution appears to
optimally preserve the morphology of the cortex. During
the first 15 s of perfusion, the pressure at the tip of the
53
cannula in the ascending aorta should be set to approach
normal rat arterial pressure of 100–120 mmHg or slightly
higher. Perfusion flow rate could then arbitrarily be
reduced to 30 ml / min to conserve fixative. The initial 15 s
of high pressure is needed not only to clear the mi-
crovessels of blood but also to keep the microvessels open
until the vascular system is fixed. Our preliminary studies
showed that slow delivery of fixative during the first 15 s
of perfusion resulted in many collapsed vessels. Rapid
delivery of a strong fixative to the nervous system follow-
ing a very brief saline wash is also critical in preventing
neuropil artifacts and formation of pyknotic neurons [21].
6.2.2. Use of 1 -m m sections and proper staining
One of the fundamental assumptions of stereology is
that the planar sections of the specimen represented true
planes. It assumes a section plane as a two-dimensional
point set of zero thickness. Similarly, line sections should
be one-dimensional point sets of zero thickness and test
points should be true points. These conditions are rarely
fulfilled in reality. Even ultrathin sections used in electron
microscopy are very thick slices when compared with the
resolution afforded by the microscope. The sections are in
thickness about three times the larger diameter of the
ribosome or 10 times thicker than the plasma membrane.
The effect of finite section thickness on stereological
measurements may be complex. The thickness of the
section tends to make a greater sample area of the structure
content available for study than would be hit by a true
planar section. This therefore would lead to overestimation
of the structure content if the basic stereological methods
are applied indiscriminately to the projected image of the
slice. Some components of structure may be projected on
top of each other so that their profiles can overlap. This
may lead one to miss some of the profiles, thus leading to
an underestimation of their density in the structure. The
contrast between components and matrix tends to become
reduced with increasing section thickness particularly if the
matrix is not completely translucent. This again has the
tendency to obscure some component profiles, particularly
if these are not totally opaque, or if they represent grazing
sections where the profiles do not extend all the way across
the section. All these effects will occur simultaneously in
any practical instance. This error introduced by finite
section thickness in estimating stereologic parameters of
objects is referred to as the ‘Holmes effect’.
We therefore chose to use 1-mm sections for our study.
Light microscopic semithin sections generally include
sections with a thickness between 0.5 and 2 mm. But we
preferred 1-mm sections since they offered sufficient ease
in cutting the sections and clear distinction between the
microvessels and the rest of the neuropil. Thicker sections
presented a poorer view of the microvessels in addition to
enhancing the Holmes effect. Thinner sections (0.5 mm
thick), though possessing more stereologic merit, unneces-
54 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
sarily complicate sectioning and offered no advantage in
identifying the microvessels.
Proper staining of the sections is important. In our study
we floated the semithin sections on a warm toluidine blue
solution (about 508C) for 1–2 min. Staining for longer than
2 min would cause the neuropil to appear too dark. Our
staining method is superior to the commonly used tech-
nique where the section is placed on the slide, covered
with a drop of toluidine blue and then heated on a hot
plate. With the latter method, staining of sections is uneven
and precipitates of the stain occur frequently in the section.
Proper staining allows the vessels to appear as conspicu-
ous white circular-like profiles against the darker stained
neuropil. This contrast is needed when using the ‘Magic
Wand’ option to automatically trace the lumen of the
vessels when the vessels are selected. The ability of IPP to
automatically trace the lumen of the vessels is dependent
on sufficient contrast between the lumen of the vessels and
the neuropil.
6.2.3. Tissue shrinkage
In most morphological studies, brains are usually fixed
and embedded in a medium. In the course of tissue
processing, the size of the brain and its components might
alter due to tissue shrinkage. The amount of shrinkage of
brain tissue varies depending on the preparatory methods.
With wax or celloidin embedding, as much as 50%
shrinkage would occur [3,20]. However, in resin embedded
tissue, shrinkage appears to be only a few percent [22].
With Epon embedding, shrinkage occurs by approximately
2% [9]. Since we used Epon as our embedding medium, it
was unnecessary to correct for the tissue shrinkage.
6.2.4. Microvessel orientation and stereology
Morphometric studies of microvessels generally involve
measuring the five stereologic features of microvessels
[23]. These are NA , VV , LV , SV and D. With IPP, the NA , VV ,
LV and SV of cerebral cortical vessels are easily and
precisely measured. The only microvessel parameter that
presents some problem is the measurement of diameter.
Cortical microvessel profiles are of different shapes though
those of Layers II–V tend to be circular. We choose to use
the minor axis of vessels as representative of the diameter.
IPP defines minor axis as ‘length of minor axis of ellipse
with same moments of order 0, 1, and 2 as object’. Our
preliminary studies indicated that the minor axis, in
general, best represented the average diameter of varying
vessel profile shapes resulting from different planes of
sections. We preferred not to apply the measure of ‘equiva-
lent diameter’ derived from the area since this would
assume that all microvessels are cut in cross-section.
The stereologic equations used in our study assume that
the microvessels are randomly distributed and oriented.
Isotropic vessels are vessels randomly oriented in space.
The vessels do not follow any particular direction; rather,
they have the same properties in all directions of space.
Anisotropic vessels, on the other hand, are vessels that
tend to be oriented in certain directions. The vessels have
different properties in different directions of space. Broad-
ly speaking, in stereology, only number and volume
estimation do not require some kind of isotropy. This
means that the specimen under study must be isotropic, or
‘vertical sections’ must be applied [1] or the sections must
be isotropic uniform random (IUR). Since cerebral cortical
microvessels of Layers II–V are isotropic [13,14,18], we
therefore did not need to use vertical or IUR sections.
Though our method was applied to isotropic vessels, the
method is applicable to tissues with anisotropic vessels as
well. For tissues with anisotropic vessels, we recommend
that sections be prepared using the isector method. The
isector is a simple and direct method for generating IUR
sections from specimens [17]. If the small specimens are
chosen from a larger organ, it is assumed that they have
been sampled from systematic or uniform random loca-
tions. On the other hand, if one assumes that the vessels
approximate cylinders then SV and LV can be estimated
from VV and average cross-sectional area and these are
independent of isotropy.
6.2.5. Determining final magnification on IPP
With the 403 objective lens, the type of relay lens and
videocamera model used in our study, the final magnifica-
tion imaged by IPP is 11603 (see Section 4.5). As
mentioned earlier, different videocamera models in combi-
nation with the particular type of relay lens used may
magnify the image differently even with the same objec-
tive lens. Depending on the final magnification of the
image, the minimum number of images that need to be
analyzed and the grid spacing may need to be changed.
Section 6.3.2 discusses the methods for determining the
minimum area of cortical tissue needed for evaluation). In
stereology, the grid spacing generally should not exceed
1.53 the diameter of the object of interest, particularly if
the object occupies (in VV ) only a small portion of the
tissue. Since the average diameter of cerebral cortical
vessels is approximately 5 mm, we therefore chose a grid
spacing of 5 mm (535 mm 2 each grid square).
We chose to use the 403 objective for our study since
this objective, in combination with our relay lens (13) and
videocamera model, allowed a large enough magnification
of the tissues to easily view the microvessels in IPP. As
mentioned earlier we chose to keep our IPP zoom power of
100% constant since it offered the best resolution and
image size for our study (see Section 4.5). For our
microscope set-up (relay lens and videocamera model), the
43, 103 or 203 objectives do not sufficiently magnify the
microvessels and therefore do not allow for greater ease in
viewing the microvessels. The 1003 objective overly
magnifies the images, thus demanding an increase in the
number of images to be analyzed. It also requires use of
immersion oil for proper resolution.
 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57 55

of IPP on digital images. Comparison of the data obtained

by Gross et al. and from our study suggests that large
numbers of cortical microvessels need not be evaluated to
obtain precise estimate of the microvascular features. The
data of Gross et al. were obtained from evaluating a total
of 21 049 microvessel profiles from five animals (approxi-
mately 4210 vessels per animal). A parietal sensorimotor
cortical area of 0.145 mm 2 per animal, as used in our
study, generally includes an average of 50 vessels. Yet our
data showed values quite similar to those reported by
Gross et al. They reported the following values: NA 5
32765; VV 51.260.05; SV 59.760.5; D55.160.02. They
did not measure the LV of the vessels. But given their
reported NA value of 327, the LV would have been
approximately 654 (LV 52[NA ]; assuming isotropic vessels;
see Section 4.9 B). This is supported by the work of
McMillan et al. [13] also conducted on adult Sprague–
Dawley rat sensorimotor cortex. McMillan et al. showed
the LV of vessels to be 6506173. They used optical
sections and the traditional grid method.
The differences in data between that of Gross et al. [6]
Fig. 2. Comparison of the coefficient of error (CE) of volume fraction
and ours (Table 1) stemmed from many reasons. We
(VV ) obtained from analyses of a cortical area of 0.0725 mm 2 (five
images), 0.145 mm 2 (10 images), 0.218 mm 2 (15 images) and 0.29 mm 2 differed in our analysis method of the vessels (manual grid
(20 images). No significant difference in CE occurs when a minimum method on photographs versus IPP automated features on
cortical area of 0.145 mm 2 (10 images) is used. digital images). Gross et al. analyzed approximately 4000
vessels per animal while we studied only 50 vessels per
animal. Gross et al. evaluated only vessels with diameters
6.3. Alternative and support protocols of 7.5 mm or less. This assured that the microvessels
evaluated were capillaries. We evaluated all the vessels
6.3.1. Number of images or total tissue area to use for within Layers II–V. Gross et al. used sections with a
analysis thickness of 2 mm while we used 1-mm sections.
We chose a cortical area of 0.145 mm 2 for analyzing the (B) We compared VV data obtained from analyzing a
microvessels for the following three reasons: cortical area of 0.0725 mm 2 (five images), 0.145 mm 2 (10
(A) The work done by Gross et al. [6] most closely images), 0.218 mm 2 (15 images) and 0.29 mm 2 (20
resembles our study. They used similar animals (adult male images). There was no significant difference in CE be-
Sprague–Dawley rats), tissue preparation methods and tween data derived from using 10, 15 and 20 images (Fig.
cortical area (parietal sensorimotor cortices) as we did in 2).
our study. Gross et al., however, used the manual grid (C) When performing point counting or using the line
method on photographs while we used automated features intersection method, one must consider how many points
need to be counted or intersections measured to obtain
adequate results. As data are added, the S.E. decreases to
Table 2 reach a value that changes little with the addition of more
Comparison of microvessel data obtained using Image-Pro Plus versus the data. For isotropic random objects that occupy a very small
manual grid method area of the section (VV of 2% for cortical microvessels), a
Animals Image-Pro Plus method Manual method S.E. (e 2A ) of 15% or less is generally acceptable. In such
VV SV D VV SV D circumstances, the minimum number of points (P) to be
counted, where x equals the estimate of the percentage of
1 2.06 14.50 5.60 2.24 16.00 5.54
2 2.12 13.54 6.17 2.14 14.21 6.07 the sample occupied by the structure of interest, is given
3 2.14 15.75 5.11 2.05 16.07 5.39 by the equation:
4 1.64 11.84 5.02 1.57 12.14 5.17
2
P 5 (1 2 x) /(e 2A 3 x) [2]
A cortical area of 0.145 mm (10 images) of the medial–dorsal (parietal
motor) cortex, 7 mm post-frontal pole, of each animal was used. VV , 5 (1 2 0.02) /([0.15] 2 3 0.02)
volume fraction (%); SV , surface density (mm 2 / mm 3 ); D, diameter (mm). 5 2178 points
The numerical (NA ) and length (LV ) densities of vessels are not shown
since both measures are identical using each method. Correlation co-
efficients (r) of the data obtained from the two methods are as follows: Our grid contains 580 points (a 29329 double square
VV , r50.844; SV , r50.913; D, r50.947. lattice with each square being 535 mm 2 ). Since we used
56 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
10 images the total number of points was 5800, twice the
required minimum number.
6.3.2. Comparison of IPP method with traditional grid
method
Data obtained from use of IPP were similar to those
obtained from traditional manual counting. Table 2 shows
the data derived from both methods on the same vessels.
The slight variation in data obtained with the manual
counting method and the IPP approach stemmed from the
fact that, with the manual method, data were manually
obtained from point and line intersections and the diameter
of vessels was manually measured. With the IPP approach,
the area, perimeter and diameter of microvessels were
automatically measured by IPP.
The VV obtained by the manual method is determined by
the equation:
VV 5 Pvessels /Ptotal
where Pvessels are points or coincident with (overlaying) the
vessels (commonly termed ‘hits’). Ptotal is the total number
of hits or points overlying the tissue (580 points per image
for our study). In the manual method, the parallel line test
system was used for determining SV with the equation:
SV 5 2 3 IL
where IL is the density of intersections on a test line length
(i.e. the number of intersections of test line with the
surface contour of the vessels divided by total length of the
test lines). The NA and LV of vessels are not shown in the
Table since both values are identical for the two methods.
The same vessels were selected and LV 52(NA ).
The similarity between data obtained with the manual
grid method and those obtained by IPP reveals that the IPP
method is a reliable alternative to the manual grid method.
The speed, ease, accuracy and efficiency of IPP make it a
superior alternative to the manual grid method. For each
image, the manual method requires about 3 min to analyze
the VV , SV and D after the vessels are selected (NA ). For 10
images, the time required is thus approximately 30 min.
More cumbersome features of IPP have to be used to
manually tag the vessels to determine the number of
appropriate hits for determining VV and SV . The D has also
to be manually measured. However, with the IPP ap-
proach, these manual challenges are eliminated reducing
the time required for evaluating the 10 images to 5 min.
7. Essential references
Original papers: Refs. [6,14,17,21,23].
8. Quick procedure
(A) Tissue preparation: fixation and embedding of brain
tissue in Epon.
(B) Cut 1-mm sections and stain with toluidine blue.
(C) Calibrate the IPP features for microvessel mor-
phometry.
(D) Select cortical areas in the section using systematic
random sampling procedure.
(E) Perform morphometric analysis of the microvessels
in the images.
(F) Transfer raw data to Excel and apply the stereologic
equations.
(G) Create a summary spreadsheet of the stereologic
data and perform statistical analysis.
Acknowledgements
Supported by the Veterans Affairs Medical Research
Service and a Grant-in-Aid from the American Heart
Association.
References
[1] A.J. Baddeley, H.J.G. Gundersen, L.M. Cruz-Orive, Estimation of
surface area from vertical sections, J. Microsc. 142 (1986) 259–276.
[2] J.J. Bozzola, L.D. Russell, in: Electron Microscopy, 2nd Edition,
Jones and Bartlett, Sudbury, MA, 1999, pp. 320–340.
[3] K.R. Brizzee, J. Vogt, X. Kharetcho, Postnatal changes in glial
neuron index with a comparison of methods of cell enumeration in
the white rat, Prog. Brain Res. 4 (1964) 136–149.
[4] R.H. Christofferson, B.O. Nilsson, Microvascular corrosion casting
with analysis in the scanning electron microscope, Scanning 10
(1988) 43–63.
[5] H.M. Duvernoy, S. Delon, J.L. Vannson, Cortical blood vessels of
the human brain, Brain Res. Bull. 7 (1981) 519–579.
[6] P.M. Gross, N.M. Sposito, S.E. Pettersen, J.D. Fenstermacher,
Differences in function and structure of the capillary endothelium in
gray matter, white matter and a circumventricular organ of rat brain,
Blood Vessels 23 (1986) 261–270.
[7] N. Harik, S.I. Harik, N.T. Kuo, K. Sakai, R.J. Przybylski, J.C.
LaManna, Time-course and reversibility of the hypoxia-induced
alterations in cerebral vascularity and cerebral capillary glucose
transporter density, Brain Res. 737 (1996) 335–338.
[8] M.J. Karnovsky, A formaldehyde–glutaraldehyde fixative of high
osmolarity for use in electron microscopy, J. Cell Biol. 27 (1965)
137A.
[9] H. Kushida, A study of cellular swelling and shrinkage during
fixation, dehydration, and embedding in various standard media, J.
Electron Microsc. (Tokyo) 11 (1962) 135–138.
[10] J.C. LaManna, L.M. Vendel, R.N. Farrell, Brain adaptation to
chronic hypobaric hypoxia in rats, J. Appl. Physiol. 72 (1992)
2238–2243.
[11] A. Lametschwandtner, U. Lametschwandtner, T. Weiger, Scanning
electron microscopy of vascular corrosion casts; technique and
applications: updated review, Scanning Microsc. 4 (1990) 889–941.
[12] S.Z. Lin, N. Sposito, S. Pettersen, L. Rybacki, E. McKenna, K.
Pettigrew, J. Fenstermacher, Cerebral capillary bed structure of
normotensive and chronically hypertensive rats, Microvasc. Res. 40
(1990) 341–357.
[13] P.J. McMillan, J.O. Archambeau, A. Gokhale, M. Archambeau, M.
Oey, Morphometric and stereologic analysis of cerebral cortical
microvessels using optical sections and thin slices, Acta Stereol. 13
(1) (1994) 33–38.
[14] V. Mironov, M.A. Hritz, J.C. LaManna, A.G. Hudetz, S.I. Harik,
 P.S. Manoonkitiwongsa et al. / Brain Research Protocols 8 (2001) 45 – 57
Architectural alterations in rat cerebral microvessels after hypobaric
hypoxia, Brain Res. 660 (1994) 73–80.
[15] E.D.F. Motti, H.G. Imhof, M.G. Yasargil, The terminal vascular bed
in the superficial cortex of the rat, J. Neurosurg. 65 (1986) 834–846.
[16] K. Nakai, H. Imai, I. Kamei, Microangioarchitecture of rat parietal
cortex with special reference to vascular sphincters. Scanning
electron microscopic and dark field microscopic study, Stroke 12
(1981) 653–659.
[17] J.R. Nyengaard, H.J.G. Gundersen, The isector: a simple and direct
method for generating isotropic, uniform random sections from
small specimens, J. Microsc. 165 (1992) 427–431.
[18] G. Pawlik, A. Rackl, R.J. Bing, Quantitative capillary topography
and blood flow in the cerebral cortex of cats: an in vivo microscopic
study, Brain Res. 208 (1981) 35–58.
[19] D. Ribatti, L. Mancini, L. Roncali, B. Nico, M. Bertossi, Mor-
phometric analysis on the effects of hypoxia during the central
nervous system vasculogenesis, Int. J. Microcirc. Clin. Exp. 8
(1989) 135–142.
57
[20] E. Robins, D.E. Smith, K.M. Eydt, The quantitative histochemistry
of the cerebral cortex. 1. Architectonic distribution of ten chemical
constituents in the motor and visual cortices, J. Neurochem. 1
(1956) 54–67.
[21] R.L. Schultz, E.F. Whitter, Improved procedures for pineal gland
fixation for electron microscopy, J. Pineal Res. 6 (1989) 267–284.
[22] E.R. Weibel, Stereological principles for morphometry in electron
microscopic cytology, Int. Rev. Cytol. 26 (1969) 235–302.
[23] E.R. Weibel, Measuring through the microscope: development and
evolution of stereological methods, J. Microsc. 155 (1989) 393–403.
[24] H.R. Weiss, E. Buchweitz, T.J. Murtha, M. Auletta, Quantitative
regional determination of morphometric indices of the total and
perfused capillary network in the rat brain, Circ. Res. 51 (1982)
494–503.
[25] Y. Yoshida, F. Ikuta, Three-dimensional architecture of cerebral
microvessels with a scanning electron microscope: a cerebrovascular
casting method for fetal and adult rats, J. Cereb. Blood Flow Metab.
4 (1984) 290–296.