The physical disector estimates particle number by comparing two separate, parallel tissue sections a known distance apart. Rather than counting every visible profile, it counts objects present in one section but absent from the other. With appropriate sampling and counting rules, this provides an estimate of number per unit volume without assuming that particles have a particular size or shape.
The method turns on three practical questions: are the sections correctly matched, is their separation known, and does an apparent disappearance represent the end of an object rather than a preparation artifact? The broader disector principle covers the statistical basis. Here, the focus is preparing section pairs, making counting decisions and calculating defensible results.
How the Physical Disector Works
A physical disector consists of a reference section, a lookup section and a counting frame applied to corresponding tissue locations. Their separation supplies the third dimension missing from a single section. The original method, introduced in Sterio’s 1984 paper on unbiased particle counting, established a three dimensional counting probe using parallel sections with known separation.
In the usual disappearance convention, a particle qualifies when its profile is sampled in the reference section and the same particle is absent from the lookup section. This event is recorded as Q−. Objects visible in both sections do not contribute to that directional count.
The distinction matters because a profile is not a particle. One nucleus may produce profiles in several successive sections. Counting those profiles independently gives that nucleus several opportunities to enter the result. A correctly applied disector instead samples its termination along the sectioning direction.
Apply the Counting Frame Before Recording an Event
Use a counting frame with inclusion and exclusion boundaries. A reference profile qualifies if it lies within the frame or touches an inclusion line, provided it does not touch an exclusion line or its prescribed extension. Apply the same convention throughout the study.
| Reference section | Lookup section | Decision |
|---|---|---|
| Eligible profile within the counting frame | Same particle absent | Record one Q− event |
| Eligible profile within the counting frame | Same particle present | Do not count |
| Profile touches an exclusion boundary | Present or absent | Exclude from this count |
| No corresponding particle | Particle present | Do not count in this direction |
| Profile identifiable | Correspondence uncertain | Resolve or flag; do not assume absence |
The lookup search must follow the object, not just its original pixel position. A particle can shift laterally between sections. A corresponding profile outside the projected counting frame is still evidence that the particle continues.
Define the Population and Sampling Plan
Write an operational definition of the object before cutting study material. Are you estimating nuclei, cells, synaptic junctions or another discrete structure? State which visible feature identifies that object and what distinguishes it from neighboring structures.
Do not label a nuclear count as a cell count without establishing the relationship. A population containing multinucleated cells needs a counting strategy that addresses those cells explicitly. Likewise, counting stained fragments is not automatically equivalent to counting the structures that produced them.
Define the reference compartment just as carefully: an entire organ, an anatomical subdivision or a tissue class within a region. Use a probability sampling design to choose blocks, section pairs and fields. Systematic uniform random sampling provides a practical framework for distributing observations after a random start.
Keep two intervals separate in the protocol. The distance between members of a pair determines disector height. The distance between sampled pairs determines how observations are distributed through the specimen. These settings answer different questions and should never share an ambiguous label such as “section interval.”
For a pilot, record time spent locating fields, matching profiles and resolving uncertain objects. Use those records to decide whether the bottleneck is sampling coverage, image quality or object identification before simply increasing the count.
Choose and Verify Section Separation
The disector height, h, is the distance between corresponding observation planes in the original section sequence. It is not the space between slides and is not necessarily the sum of the thicknesses of the two mounted sections.
For equally thick sections, corresponding planes in adjacent sections are one section interval apart. Sections 10 and 12 are two intervals apart. At a verified interval of 1 µm, their separation is therefore 2 µm, not 3 µm.
The denominator of a density estimate depends directly on this distance. Do not accept a nominal microtome setting without checking its suitability for the preparation and estimator. Experimental work on histological section thickness measurement demonstrates methods for verifying section thickness and explains its role in physical disector estimates.
Document whether the estimator uses calibrated block advance or measurements made on processed sections. Those measurements describe different stages of preparation; choose a consistent dimensional basis rather than mixing them.
Keep the Pair Close Enough to Interpret
Particles lying entirely between the two observation planes cannot be detected by comparing those planes alone. Excessive separation can also make correspondence uncertain. Conversely, very small separation may produce few counting events, increasing the workload.
A separation around one quarter to one third of particle height is a commonly used starting point, not a universal guarantee. A comparison of disector estimates with reconstructed synapse counts examined the consequences of changing that separation. The practical requirement is to validate it for the objects being counted.
During the pilot, inspect additional consecutive sections around candidate pairs. Pay particular attention to the smallest target objects and to structures that split into several profiles or become difficult to recognize. A satisfactory average size does not settle every identification problem.
Match Sections Without Manufacturing Disappearances
Acquire enough surrounding tissue to recognize landmarks and track particles across the pair. Begin with regional alignment, then inspect correspondence locally. A good match at one corner of an image does not establish a good match everywhere.
Serial sections can rotate, stretch, fold or lose tissue. Registration software can assist alignment, but it cannot restore a missing piece of specimen. Review the underlying images rather than accepting an overlay as proof that every object corresponds.
Physical disectors are useful in electron microscopy, where paired thin sections can resolve small structures. A primary study of synapse quantification in the mouse dentate gyrus combined serial sections with controlled sampling of microscope fields. Both anatomical localization and field selection were part of the method, not optional preparation steps.
For your own protocol, save paired images with frame overlays and event annotations. Record uncertain matches separately. If one observer calls a profile “absent” while another finds its continuation nearby, the disagreement should lead to a clearer identification rule, not an average of two incompatible decisions.
Calculate Numerical Density
For uniformly sampled disectors within a defined reference compartment, estimate numerical density by dividing the accumulated count by the accumulated sampled reference volume:
Estimated NV = ΣQ− / ΣVdis
When every counting frame lies fully within that compartment, each disector volume is its frame area multiplied by its height. With identical frames and heights, the denominator is n × a × h, where n is the number of directional disectors and a is frame area.
If frames cross a compartment boundary, account for the sampled compartment area using the chosen design. Do not divide a count from a thin tissue compartment by an image volume that also contains background or another tissue.
Worked Example
Suppose an illustrative study uses 100 directional disectors, each with a frame area of 2,500 µm2 and a height of 1 µm. All frames lie within the reference compartment. The accumulated count is 125 events.
Sampled volume = 100 × 2,500 × 1 = 250,000 µm3
Estimated density = 125 / 250,000 = 0.0005 particles/µm3
Since 1 mm3 contains 109 µm3, the result is 500,000 particles/mm3. These are demonstration values, not recommended sampling settings.
The denominator also shows the effect of measurement error. If the true height were 0.8 µm but the calculation used 1 µm, the reported density would be 20% below the value calculated with the correct height, holding the count and area constant.
Counting in Both Directions
You may reverse the roles of the two sections and repeat the procedure. This produces a second directional count from the same pair. A published physical fractionator study of bladder epithelial cells used this approach and explicitly adjusted its estimator for counting in both directions.
For density estimation with equal sampled areas, sum both directional counts and divide by twice the single-direction volume. Alternatively, average the counts and retain the single-direction denominator. Adding both counts without adjusting the denominator doubles the estimate incorrectly.
Keep the two counts distinguishable in the data file. They reuse the same tissue and must not be treated as independent biological specimens.
Numerical Density Is Not Total Number
A physical disector density estimate answers “how many per unit volume?” It does not, by itself, answer “how many in the whole region?”
For an illustrative comparison, consider two regions that each contain 100,000 particles. If one occupies 1 mm3 and the other 0.8 mm3, their densities are 100,000 and 125,000 particles/mm3. The second is denser without containing more particles.
One route to total number is to multiply estimated numerical density by an estimate of the matching reference volume. The Cavalieri method for volume estimation can supply that volume within an appropriate study design. The compartment and processing state must match those used for density estimation.
A physical fractionator takes a different route: it combines physical disector counting with known sampling fractions to estimate total number. Do not substitute a fractionator formula into a density study after collection unless the required sampling fractions were established and recorded.
Quality Checks Before Expanding the Study
Build a small audit set containing straightforward events, boundary cases, uncertain matches and damaged fields. Ask observers to apply the written rules without discussing each image first. Resolve disagreements before counting the full sample.
Predetermine how to handle missing sections and damaged sites. Record every exclusion and its reason. Replacing an awkward field with a more convenient neighbor changes the sampling procedure; any replacement rule should be part of the design rather than improvised at the microscope.
Keep valid empty fields. A zero count is an observation, not a failed image. By contrast, a field that cannot be interpreted because its lookup section is missing is not a valid zero.
Also separate more counting from better counting. Increasing the number of observations does not repair inconsistent particle definitions, unverified distances or systematic matching mistakes.
Physical or Optical Disector?
The physical method compares separate sections. The optical disector follows objects through focal planes within a thicker section. Their counting principle is related, but their preparation and quality control demands differ.
For a practical choice, test whether the physical preparation provides recognizable particles, reliable serial correspondence and a defensible separation. For an optical preparation, assess visibility through depth and whether usable section thickness supports the intended probe. Choose from pilot evidence rather than assuming one method is always faster or more accurate.
What to Report
A methods section should allow another laboratory to reconstruct both the sampling and the counting decisions. The guide to reporting stereological methods and results covers the wider reporting framework. For a physical disector study, include:
- The target particle, identification feature and reference compartment.
- Block, section-pair and field sampling procedures, including random starts.
- Section preparation, pair separation and its verification method.
- Frame dimensions, boundary rules and alignment procedure.
- Counting direction, accumulated events and the estimator denominator.
- Missing material, exclusions, uncertainty handling and specimen-level results.
The most useful final check is traceability: every reported estimate should lead back to identifiable section pairs, annotated counting events and a documented sampled volume or sampling fraction.