The disector principle estimates the number of particles in three dimensions by counting events across depth, rather than counting every profile visible in a section. A particle might be a cell nucleus, an organelle or another clearly identifiable object. The defining question is whether that object appears or disappears between observation planes.
This approach avoids the size bias of ordinary profile counts: a large particle should not receive more counting opportunities simply because it intersects more sections. Introduced in Sterio’s 1984 paper on unbiased particle estimation, the disector provides a counting principle, not a complete sampling plan. Reliable estimates still require suitable specimens, unambiguous identification and correctly placed sampling sites.
Why Counting Profiles Is Not Counting Particles
A profile is the two dimensional intersection of a particle with a section. One particle can produce profiles in several successive sections. A particle with a complicated shape may even produce several separate profiles within one section.
Particles with greater extent perpendicular to the section plane have more opportunities to be intersected. Consequently, an increase in profiles per unit area might reflect larger particles rather than more particles. Changing their orientation or shape can also change the profile count without changing their number. These distinctions are illustrated in the original methodological paper A Brief Introduction to Particle Number Estimation.
Consider a hypothetical experiment in which nuclei enlarge after treatment but no cells divide or die. Counting nuclear profiles could suggest a numerical increase. The measurement has answered “how many intersections are visible?” when the biological question was “how many nuclei exist?” Those questions are not interchangeable.
How the Disector Counting Rule Works
A physical disector compares two parallel section planes separated by a known distance. One serves as the reference section, the other as the lookup section. Corresponding tissue locations must be identified in both.
Under the disappearance convention, count a particle when it is present in the reference section but absent from the lookup section, provided it satisfies the counting frame rules. The accepted count is conventionally written Q−. Particles visible in both sections do not contribute to that directional count.
| Reference section | Lookup section | Decision |
|---|---|---|
| Particle present | Same particle present | Do not count |
| Particle present | Particle absent | Count if the frame rules allow it |
| Particle absent | Particle present | Do not count in this direction |
The frame has inclusion and exclusion edges. Accept eligible reference profiles within the frame or touching an inclusion edge, but reject those touching an exclusion edge or its prescribed extension. A particle touching both an inclusion edge and an exclusion edge is excluded. These rules prevent arbitrary decisions at field boundaries. The combined presence, absence and frame rules are demonstrated in a physical disector study of dopaminergic neurons.
During protocol development, keep annotated examples of each decision. Include awkward boundary cases, not just the easy particles in the middle of a field. Another observer should be able to apply the written rule without having to guess what the first observer intended.
What Counts as One Particle?
Define the counting unit before collecting data. Counting nuclei estimates nuclear number. Calling that result “cell number” requires a justified relationship between nuclei and cells. Multinucleated cells make the distinction especially clear.
A nucleolus can be useful when the target cells reliably contain one, but counting every nucleolus is not a valid substitute for counting cells when nucleolar number varies. The choice of nucleus or nucleolus, and the requirement for a dependable counting unit, are addressed in Boyce and Gundersen’s particle counting methodology.
Write the operational definition in terms an observer can use: the required label, morphology and relationship to nearby structures. Specify how to handle touching nuclei, uncertain staining and fragmented profiles. Do not create a new exception halfway through counting because an inconvenient object appears.
Physical and Optical Disectors
Physical Disector
The physical version uses separate histological sections. Its practical challenge is matching the same particle across the pair. The workflow therefore needs dependable section order, image alignment and recognition criteria.
The separation must be small enough to avoid missing particles that could lie entirely between the observed planes. Knowing the distance is not sufficient if the observations cannot establish what happened within that distance. For implementation details, including section pairing and matching, use the guide to the physical disector.
Optical Disector
The optical version follows particles through successive focal levels inside a thick section. Instead of comparing two separately mounted sections, the observer moves through a defined depth and records the first appearance of the chosen counting feature.
Objects already visible at the initial exclusion plane are not counted. Eligible objects first encountered within the counting depth are accepted under the frame and depth boundary rules. The observer must follow the focal sequence, not count everything visible in a projection of the stack. A projection removes the depth information that makes the method work.
The optical disector guide covers focal plane recognition, depth measurement and practical setup. The choice between versions should follow the specimen and identification requirements, rather than assuming that one method is automatically preferable.
Calculating Numerical Density
For equal sized disectors placed within the reference compartment, numerical density is estimated by dividing the accepted count by the total sampled volume:
N̂V = ΣQ− / (n × a × h)
Here, n is the number of disectors, a is the counting frame area and h is the disector height. When valid sampled volumes differ, use their sum in the denominator rather than assuming that every site contributes the same volume.
The height is the distance between observation planes, not automatically the nominal thickness selected on a microtome. Incorrect height produces an incorrect volume and therefore an incorrect density. The geometry, density formula and consequences of unverified section thickness are detailed in research on section thickness measurement for stereological analyses.
Worked Example
Suppose a hypothetical study examines 100 disectors, each with a frame area of 2,500 µm2 and a height of 10 µm. Assume every frame lies fully within the reference compartment and all sites are valid. The observer records 150 accepted events.
The volume of one disector is:
2,500 µm2 × 10 µm = 25,000 µm3
The total sampled volume is:
100 × 25,000 µm3 = 2,500,000 µm3 = 0.0025 mm3
The estimated numerical density is therefore:
150 / 0.0025 mm3 = 60,000 particles/mm3
These values illustrate the arithmetic; they are not recommended settings. Suitable height and frame area depend on the specimen and counting unit.
Notice what happens if the recorded height is mistakenly entered as 20 µm while the count remains unchanged. The calculated sampling volume doubles and the estimated density falls to 30,000 particles/mm3. Perfect particle recognition would not rescue that calculation.
Counting Both Directions
A physical section pair can be evaluated again with the reference and lookup roles reversed. If both directional counts are pooled for density estimation, both directional sampling volumes must also enter the denominator.
For example, if one direction yields six events and the reverse direction yields four, the pooled numerator is ten. For equal frame areas and heights, its denominator is 2ah, not ah. Doubling the observations without doubling the corresponding sampling effort would double the estimated density incorrectly.
Numerical Density Is Not Total Number
Numerical density describes particles per unit reference volume. Total number describes how many particles the entire defined region contains. To obtain total number through a density and volume approach, multiply the density estimate by a compatible estimate of reference volume:
N̂ = N̂V × V̂ref
In the worked example, a compatible reference volume of 2 mm3 would give an estimated total of 120,000 particles. A reference volume of 1 mm3 would give 60,000, despite identical numerical density.
An alternative is to combine disector counting with known sampling fractions. The original optical fractionator study established this combination of an optical counting probe and systematic fractionator sampling for estimating total neuronal number.
Choose the endpoint before preparing specimens. “Are there fewer cells?” and “are cells less densely packed?” may demand different interpretations. A region can change volume as well as particle number; density alone does not separate those changes.
Sampling Must Support the Counting Rule
A defensible counting event does not make a conveniently chosen field representative. Decide where sections and fields will be sampled before inspecting them for abundant or easily recognized particles.
For a conventional uniform design, use a randomized starting position followed by the prescribed sampling interval. Spread observations across the defined reference region rather than concentrating effort in attractive fields. The guide to systematic uniform random sampling covers this sampling structure.
Build the pilot around practical questions. Can observers identify the counting unit throughout the usable depth? Can corresponding particles be matched? Are many sampled sites empty? Does a small part of the region account for most counts?
Keep empty but valid sites in the sampling record. Distinguish them from sites that cannot be assessed because of damage or imaging failure. Decide how such failures will be handled before the main study, rather than replacing them with nearby fields that happen to contain countable objects.
Where Disector Studies Can Go Wrong
Unbiased in principle does not mean immune to implementation errors. A validation study in frog sensory ganglia found that the tested physical disector protocol produced variable and systematically low estimates compared with a calibrated reconstruction approach. Increasing sampling improved repeatability without removing the observed bias. This study of physical disector reliability and validity supports testing the actual protocol in the tissue of interest, not treating the method’s name as a guarantee.
Make quality checks address different failure modes separately. Recounting the same images tests consistency, but does not establish that poorly stained particles were visible. Checking image alignment does not establish that the section separation was appropriate. Repeating a calculation does not validate its inputs.
For optical work, examine whether usable tissue depth, surface damage and particle visibility support the selected counting interval. Guard zones should be justified by the preparation rather than copied as a universal setting. The guide to section thickness, guard zones and tissue shrinkage addresses these preparation issues.
What to Record and Report
A useful methods record should allow another researcher to reconstruct both the counting decision and its denominator. Record the reference region, particle definition, sampling schedule, frame area, disector height and measurement method. State which boundary rules were used and whether physical pairs were counted in one or both directions.
Report accepted counts and sampling effort for each specimen, alongside exclusions and the reason for each exclusion. For density estimates, identify the reference volume and units. For total number estimates, state how density and volume were combined or which sampling fractions were applied.
The practical test is straightforward: can a reader tell what was counted, where it could have been counted, and how the observations became the reported estimate? If any of those steps is missing, the final number is harder to evaluate than it needs to be.