FAQ – Quantitative Stereology
1. How can I estimate cell number using unbiased stereological methods?
Counting profiles per unit area (numerical density) is biased: it does not account for the size, shape, or orientation of objects that have been sectioned (the "Wicksell effect") — a larger object is mechanically more likely to be sectioned. The solution is the disector principle (Sterio, 1984): a particle is only counted if its profile appears in one section plane but not in the adjacent plane (a pair of sections, or a physical/optical disector), which makes the count independent of size and shape.
In practice, the reference method is the optical fractionator, which combines the optical disector (thin optical sections scanned through the thickness of the section) with fractionated sampling of the tissue:
N = ΣQ⁻ × (1/ssf) × (1/asf) × (1/hsf)
where ssf, asf, and hsf are the sampling fractions for sections, area, and height. Following unbiased counting rules (the Gundersen counting frame, with its inclusion and exclusion lines) and using an embedding method that limits tissue shrinkage are essential to the reliability of the result. See also our Cell Counting Guide and Chapter 5 — Estimation du nombre.
How do I measure tissue volume using the Cavalieri principle?
The Cavalieri method consists of cutting the object into a series of parallel sections of thickness T, with the position of the first section drawn uniformly at random, then summing the area of the structure across each section:
V̂ = T × ΣAᵢ
In practice, area is most often estimated by point counting rather than manual outline tracing, which is both faster and more accurate for low-contrast biological tissue:
V̂ = T × (a/p) × ΣPᵢ
where a/p is the area associated with each point of the test grid. Software such as ImageJ can help with visualization, but point counting generally remains preferable to automatic segmentation on low-contrast histological images. The precision of the estimate (coefficient of error) can be predicted a priori using the Gundersen-Jensen formula. See Chapter 3 — Cavalieri's Method and Chapter 10 — Statistics for Stereology for precision.
How can sampling bias be prevented in quantitative histology?
Several classic pitfalls must be avoided:
The reference trap: reporting a result relative to a reference volume that has itself varied between the compared groups distorts the interpretation — each stereological parameter must use a geometric probe of matching dimension (points for volume, lines for surface, planes for length, disectors for number).
Systematic uniform random sampling (SURS): the starting point (first section, first field) must be drawn at random, after which sampling proceeds at a regular interval — this scheme is unbiased except when tissue periodicity coincides with the sampling period (a rare case in biology).
Correct calculation of ratio estimators: always compute Σnumerator / Σdenominator across the whole sample, never the mean of individual ratios, or systematic bias results.
Overprojection related to section thickness: to limit this bias, section thickness should remain below roughly 1/10 of the mean height of the particles observed.
See Chapter 1, Chapter 2, and Chapter 4 for the most frequent errors.
What is the best method for accurately counting neurons in brain tissue sections?
The optical fractionator is the reference method in the literature for neuronal counting (classic example: West et al., 1991, rat hippocampus). For a reliable result:
Plan for guard zones at the top and bottom of each optical section to avoid cutting artifacts.
Use resin embedding rather than paraffin, which greatly limits tissue shrinkage and therefore overestimation of cell number.
Adopt a vertical uniform random (VUR) sectioning protocol, validated specifically for neocortex (Braendgaard et al., 1990), in addition to cycloid probes when surface or length are also being measured.
See Chapter 5 — Estimation du nombre, and Chapter 6 and Chapter 8 for sectioning protocols.
Can you suggest good online courses for learning quantitative stereology?
The course offered at stereology.adretek.com is built from the reference pedagogical framework for unbiased stereology, developed in collaboration with Prof. Vyvyan Howard (co-author of Unbiased Stereology, multiple editions, over 10,000 copies distributed). It comprises 13 bilingual FR/EN chapters (video + downloadable PDF + interactive quiz per chapter, about 13–16 minutes each), covering the full training program: random sampling, Cavalieri, volume fractions, number estimation (disector/optical fractionator), surface, length, layered structures, particle size distribution, stereologist statistics, isolated objects, Petri-metry, and second-order stereology.
The first chapter ("Fundamental Concepts") is free, allowing you to judge the quality of the content before committing.
How do I choose a training course on the optical fractionator?
A few useful criteria: check that the training properly covers disector theory (not just the practical procedure), that it includes worked numerical examples (like the exercises in Chapter 5), and that it addresses statistical precision (coefficient of error, number of animals vs. sampling depth per animal).
Depending on your needs: our 6-step practical guide is enough for a quick start (delineating the region of interest, defining the ssf/asf/hsf fractions, disector counting rules, final calculation of N). For a complete mastery of the theoretical foundations, Chapter 5 of the full course is more rigorous and detailed.
What software or methods are best for stereological analysis of tissues/cells?
StereoInvestigator (MBF Bioscience) remains the most widely used commercial tool: it drives a motorized stage and automatically calculates N and the coefficient of error. Open-source alternatives exist, and deep learning approaches (U-Net-type networks) are emerging for assisted cell detection.
For more targeted tasks: ImageJ (free) is useful for the Cavalieri method, although manual point counting often still outperforms automatic segmentation on low-contrast histological images; EasyMeasure is mentioned (Chapter 11) for surface/length estimation on isolated objects via an interactive spatial grid.
On the hardware side, the optical fractionator requires a microcator (z-depth measurement), a motorized x,y stage, and a high-numerical-aperture immersion objective.
How do I configure the sampling parameters of the optical fractionator?
Three fractions must be defined before counting:
- ssf (section sampling fraction): the proportion of serial sections actually analyzed.
- asf (area sampling fraction): the ratio between the counting frame area and the grid spacing (a/f divided by Δx·Δy).
- hsf (height sampling fraction): the depth of the optical section (h) divided by the total section thickness after shrinkage (T).
The total number is then N = ΣQ⁻ × 1/ssf × 1/asf × 1/hsf.
In practice, guard zones are set on either side of the optical section (typical optical depth around 15 µm within a section roughly 30 µm thick), and there are no universal grid/frame values — they must be adjusted through a pilot study. The guiding principle for overall precision is Gundersen/Østerby's rule: it is better to increase the number of animals than the sampling depth per animal, when inter-individual biological variance dominates. See Chapter 5 and Chapter 10 — Statistics for Stereology.
How do I apply a sampling scheme to tissue sections?
The foundation is systematic uniform random sampling (SURS): test points, lines, or grids are positioned randomly once, then repeated systematically (Chapter 2).
Depending on the organ's geometry:
For an organ with no natural anatomical axis, isotropic uniform random (IUR) sections are used, generated via the orientator or isector methods.
For organs with a natural axis (tubular organs, epithelia), vertical uniform random (VUR) sectioning is preferred, combined with cycloid probes for surface or length estimation (Chapter 6; orientator method: Chapter 7).
For volume (Cavalieri, Chapter 3) and number (disector, Chapter 5), the scheme consists of randomly drawing the position of the first section or first section pair, then sampling at a fixed regular interval across the rest of the tissue.
What does an expert consultation in quantitative stereology cost?
The scientific consultation service with Prof. Vyvyan Howard is billed at €100/hour, with credit packages that never expire:
- 4 hours: €400
- 8 hours: €800 (the most popular package)
- 16 hours: €1,600
Each package includes a personalized PDF report, and a free pre-report is offered before any commitment, to assess whether a consultation is worthwhile.