Stereological methods estimate three dimensional quantities from sampled sections, images or focal planes. The method depends on the quantity you need: volume, volume fraction, particle number, particle size, surface area or length. These are different measurements, not interchangeable ways of describing how much structure appears in an image.
The practical task is to match the biological or materials question to an estimator, then prepare and sample the specimen accordingly. This guide compares the main methods and their requirements. The linked method guides cover calculations, counting rules and implementation in greater detail.
Choose the Method by the Quantity You Need
Start with a measurable question. “Has this tissue changed?” is too broad. “Has the total number of identified cells changed within this region?” gives you a target population, a reference region and an endpoint.
| Measurement target | Method | Main observation | Preparation requirement |
|---|---|---|---|
| Regional or organ volume | Cavalieri estimator | Section areas or point hits | Parallel sections at known intervals with a random start |
| Component volume fraction | Point counting | Points hitting the component and reference space | Representative sections and fields |
| Particle numerical density | Physical or optical disector | Particle appearances or disappearances | Paired sections or usable optical depth |
| Total particle number | Optical fractionator | Disector counts within known sampling fractions | Sampled thick sections with measurable depth |
| Mean cell or particle volume | Nucleator | Distances from an internal point to the boundary | Suitable particle selection and directional sampling |
| Surface area | Line intersection methods | Intersections between test lines and boundaries | A compatible section orientation and test system |
| Fiber or vessel length | Space balls | Intersections with virtual spherical surfaces | Usable image depth and identifiable linear structures |
The distinction between total quantities and densities deserves attention before any counting begins. A density describes an amount per unit reference volume; it does not establish the total amount without further information. This distinction is part of the ATS/ERS standards for quantitative structural assessment.
Volume Estimation: Cavalieri and Point Counting
Cavalieri Estimates the Volume of a Defined Region
The Cavalieri estimator combines the areas of a structure on parallel sections with the distance between sampled section planes. A systematic series begins at a uniformly random position within the first sampling interval and continues across the full extent of the target.
For equally spaced planes, the calculation is:
Estimated volume = section interval × sum of sampled areas.
Areas can be measured by tracing boundaries or estimated with a calibrated point grid. The Gundersen and Jensen study of systematic sampling develops this volume estimator alongside methods for assessing sampling efficiency.
Consider an illustrative series with a summed area of 24 mm2 and a sampling interval of 0.5 mm. The estimated volume is 12 mm3. The interval is the distance between the sampled planes, not automatically the thickness of one mounted section.
Before choosing this method, ask whether the entire region can be followed through the series and whether its boundary can be identified consistently. Use the Cavalieri volume estimation guide for section selection, area measurement and interval handling.
Point Counting Estimates How Much of a Reference Space a Component Occupies
Point counting classifies test points rather than tracing every visible boundary. With representative sampling and a common point grid, the estimated component volume fraction is the number of points hitting that component divided by the number hitting the reference space.
In a hypothetical assessment, 80 component hits among 400 reference hits give a volume fraction of 0.20, or 20%. If the corresponding reference volume is 12 mm3, the estimated component volume is 2.4 mm3. Without the reference volume, the result remains a fraction.
Define the denominator explicitly. “Collagen fraction of the tissue” and “collagen fraction of the entire specimen, including empty spaces” are different endpoints. The point counting guide covers reference boundaries, point classification and grid selection.
Particle Number: The Disector and Optical Fractionator
The Disector Counts Events in Three Dimensions
A visible cell profile is not the same thing as a sampled cell. Larger particles can intersect more section planes than smaller particles, so counting profiles alone does not provide a size independent estimate of particle number. The disector instead uses a defined counting event between section planes or through optical depth.
With a physical disector, corresponding fields in two sections are compared. Under a chosen counting direction, a particle present in the reference section but absent from the lookup section contributes a count if it meets the counting frame rules. An optical disector applies the same three dimensional principle by following particles through focal planes within a thick section.
The disector principle guide explains the counting logic and the choice between physical and optical implementations. Neither approach means counting every profile visible inside a rectangle.
For a simple numerical density estimate, divide accepted particle counts by the total sampled disector volume. Converting that density to total number requires a matching reference volume. Alternatively, a disector can be incorporated into a fractionator sampling design.
The Optical Fractionator Estimates Total Number
The optical fractionator combines optical disector counts with known fractions of the section series, section area and section thickness. Rather than multiplying numerical density by regional volume, it scales sampled counts by the inverse of the sampled fraction. This combination was established in the original optical fractionator study by West, Slomianka and Gundersen.
For a design with constant sampling fractions:
Estimated total number = accepted counts ÷ (section sampling fraction × area sampling fraction × thickness sampling fraction).
Suppose a hypothetical study samples one in ten sections, one twentieth of the area and half the usable section thickness. The combined fraction is 1/400. A count of 250 gives an estimated total of 100,000 particles. This example illustrates the arithmetic, not a recommended sampling schedule.
The optical fractionator guide addresses sampling fractions and thickness variation. Do not apply a single convenient thickness value without checking whether it represents the sampled sites.
Also decide what qualifies as one particle. If nuclei serve as counting units, the endpoint is initially nuclear number. Interpreting it as cell number requires a justified relationship between nuclei and cells in the target population.
Particle Volume: The Nucleator
The nucleator estimates individual particle volume from boundary distances measured along appropriately sampled directions from an internal reference point. It answers a size question, rather than the number question addressed by the disector.
To estimate a number weighted mean volume, particle selection must not favor large objects. Uniform particle sampling with a disector can provide that selection. Directional sampling must also satisfy the estimator’s requirements; rotating rays within an arbitrarily oriented section is not automatically sufficient. These conditions follow from Gundersen’s original nucleator paper.
Keep the measured compartment explicit. A study may target nuclear volume, cell body volume or another bounded particle. Those choices need different boundary definitions, even when the same internal landmark is visible.
A useful planning question is whether the stain shows both the sampling landmark and the boundary you intend to measure. If only the nucleus is clear, a cell body volume endpoint needs further preparation work. The nucleator guide covers particle selection, intercept measurements and interpretation of the resulting mean.
Surface Area: Count Intersections, Control Orientation
Surface estimation uses intersections between test lines and the section profiles of surfaces. The observation is a line crossing a boundary, not the area enclosed by that boundary. Suitable designs can estimate surface density, which can then be related to the matching reference volume.
Orientation is central to method selection. One approach uses isotropic uniform random sections. Another uses vertical uniform random sections with a compatible cycloid test system. Vertical sections retain a chosen axis while randomizing rotation around it; the test system accounts for that directional constraint. The original method is developed in Baddeley, Gundersen and Cruz-Orive’s paper on surface estimation from vertical sections.
Choose the orientation scheme before cutting the specimen. For an epithelial study, write down whether the target is the luminal surface, basal surface or another interface. For a materials study, identify which phase boundary counts. The surface area estimation guide explains probe selection and boundary counting.
Length Estimation: Space Balls
Space balls estimate the length of linear structures, such as fibers or vessels, by counting intersections with virtual spherical or hemispherical surfaces. The probe surface supplies the required range of orientations, so the tissue itself need not be cut in isotropic orientations.
As the observer moves through an image stack or thick section, the virtual surface appears as circles of changing diameter. The task is to identify crossings of the structural centerline with that surface. Counting every instance where a thick fiber touches the probe can overcount intersections, a problem addressed in the methods paper Space Balls Revisited.
Before committing to this approach, inspect whether the target structures remain identifiable throughout the required depth. Define how crossings, branches and closely adjacent fibers will be handled. The space balls length estimation guide covers probe geometry and routes from sampled intersections to length estimates.
Combine Methods When the Question Requires It
A study need not rely on one estimator. Plan each measurement around the claim it is intended to support, rather than collecting several endpoints because they are available in the software.
Consider a hypothetical treatment comparison. Region A has an estimated volume of 10 mm3 and contains 100,000 target cells. Region B has an estimated volume of 8 mm3 and also contains 100,000 target cells. Their numerical densities are 10,000 and 12,500 cells per mm3, respectively.
Reporting density alone would show a 25% increase in Region B. Reporting total number would show no difference. Neither calculation is wrong; they answer different questions. Pairing regional volume with total number makes the distinction visible rather than leaving the reader to guess.
For a proposed study of tissue enlargement, consider separate endpoints for regional volume, total cell number and mean cell size. Specify in advance which comparisons address enlargement through more cells, larger cells or changes in other tissue compartments. Avoid treating one measurement as a substitute for all three.
Check Feasibility Before Choosing the Final Protocol
Use a pilot to answer practical questions before processing the full collection. Can observers identify the target consistently? Can the region be delineated throughout the required section series? Are boundaries visible at the intended magnification? Can the preparation support the depth measurements required by the chosen probe?
For optical methods, examine mounted thickness, staining through depth and the placement of counting zones. Do not treat the microtome setting as a complete description of the finished section. The guide to section thickness, guard zones and tissue shrinkage addresses these preparation constraints.
Keep the pilot record practical: retain example images, note ambiguous classifications and document rejected fields. Set rules for damaged or missing sections before group comparisons begin. A written rule is easier to apply consistently than an improvised decision at each microscope field.
Finish method selection with a short specification: the target quantity, reference region, counting or measurement unit, sampling design and preparation requirements. If those items agree, the estimator has a clear job. If they do not, revise the study before collecting more measurements.