The Stereology and 3D Volume Analyses in Nervous Tissue - Animal Models of Acute Neurological Injuries: Injury and Mechanistic Assessments - page 32

3. Practical
Considerations in
Using Stereology

The concern with traditional section-by-section “2D” (thin sections)
counting methods is the presence of several inherent biases. The
main problems with “2D” counting and sampling include deter-
mining what and how many sections to sample and defi ning repre-
sentative sections as well as regions of interest. What to sample
refers to maintaining regional consistency throughout the sections
sampled, which may be established by anatomical or regional
landmarks aided by immuno- or counter-staining. A stereological
analysis of the number of serotonin (5-HT) neurons in the dorsal
and ventral raphe ( 6 ) exemplifi ed the importance of regional
consistency, yet it points out that regional sampling has some
caveats. Both stereology and manual counting were performed and
compared (MR, DR, Fig. 3 ). In this case, blood vessels, which sur-
round the raphe throughout the sections, provided an appropriate
anatomical boundary (Fig. 3a , b arrows).
After defi ning the ROI, the Stereo Investigator program (and
some, but not all others) provides an automated, systematic, and
unbiased means to estimate the sampling population within the
ROI. Stereology programs should take into account user prefer-
ences for the sampling fraction (these are normally determined in
a pilot study ):

3.1. Less Regional
Variation and
Increased Sampling
Effi ciency

1. The section interval desired to count from each animal (this
determines the proportion of sections through the ROI to be
sampled).
2. The spacing of the counting frames per section, known as the
grid size (this determines the proportion of the cross-sectional
area sampled). The number of sampling sites per section will
vary with the size of the ROI on each section and the random
start point—it will normally be a different number of sampling
sites on each section.
3. The disector height and guard zones sizes (this determines the
proportion of the section thickness that will be sampled).
The program then automatically and randomly provides the
number of counting frames, which will normally vary somewhat
from section to section. If an exhaustively sectioned ROI would
lead to too many sections, greater than 50 for example, being able
to sample a subset, e.g., every third or fourth section, rather than
a higher sampling interval can improve effi ciency signifi cantly with
minimal impact on the precision of the resulting estimates. The
number of sections sampled impacts the coeffi cient of error (CE),
which refl ects variability in the estimated mean ( 7 ). Differing levels
of precision may be needed for any given study, thus viable CE val-
ues may vary signifi cantly. Manual counting on the other hand, using