Swedish Interactive Thresholding Algorithm

An adaptive static-perimetry algorithm family that continuously updates a visual-field model from patient responses and stops testing each location when a specified threshold-certainty criterion is met.

Version

v1 · 2026-08-30 · History

Domain-specific #

2908

Origin domain

ophthalmic diagnostics

Subdomain

automated perimetry

Aliases

SITA

Core Idea

The Swedish Interactive Thresholding Algorithm, or SITA, is a family of computerized static-perimetry strategies for estimating visual sensitivity at many locations in a patient's visual field. Rather than applying an independent fixed staircase at every location, SITA maintains a comprehensive model informed by normal and glaucomatous fields, updates that model as responses arrive, estimates both threshold and uncertainty at each location, and ends testing at a location when a predetermined certainty level has been reached. The original developers explicitly presented this coupled model-update and certainty-stopping architecture as a way to reduce test time without reducing result quality.^[1]

The identity is therefore not simply “a faster eye test.” It is the reusable algorithmic organization of prior field knowledge, spatially related threshold evidence, sequential stimulus responses, online error estimates, stopping rules, response-reliability estimation, and a final whole-test recomputation. SITA Standard, SITA Fast, and SITA Faster alter parameter settings and test-economy choices while remaining members of the recognized family.

Structural Signature

• Subject and eye: a patient supplies psychophysical responses for a specified visual field.
• Spatial test pattern: multiple retinal or field locations require sensitivity estimates.
• Stimulus-response loop: the perimeter presents luminance stimuli and records seen/not-seen responses.
• Field model: normal-age and disease-informed knowledge constrains plausible spatial threshold patterns.
• Interactive update: every response updates estimates during the same test rather than only after collection.
• Threshold estimate: each location receives an estimated differential light sensitivity, normally expressed in decibels.
• Uncertainty estimate: the algorithm tracks how precisely each local threshold is currently known.
• Location-specific stopping: presentation at a location ends when its designated certainty condition is satisfied.
• Reliability estimation: false-positive or related behavior is estimated without simply duplicating the traditional catch-trial burden.
• Pacing and scheduling: the next stimulus and its timing are chosen to use the patient's responses efficiently.
• Final recomputation: completed-test responses can be used to update the final set of local estimates.
• Variant policy: Standard, Fast, and Faster preserve the family identity while changing speed–precision tradeoffs and implementation details.

What It Is Not

SITA is not the visual field itself, the perimeter hardware, or the grid designation such as 24-2 or 10-2. It is a threshold-estimation strategy executed within compatible automated perimetry. It is not glaucoma diagnosis: the result supplies evidence that a clinician interprets alongside reliability, anatomy, history, and other examinations.

It is not any adaptive staircase, any Bayesian psychophysical test, or any algorithm that stops early. Those ingredients are broader than the SITA family. Nor is SITA Fast merely an informal instruction to administer SITA quickly; it is a defined variant. SITA Faster is a later strategy whose modifications produce further time savings and whose results need not be interchangeable without attention to disease severity and longitudinal transition.

Scope of Application

The primary scope is standard automated perimetry used to detect or monitor visual-field loss, especially in glaucoma and glaucoma suspicion. The same role package applies across many patients, field locations, severity levels, and recurring examinations. The 1997 construction paper compared simulations for normal and glaucomatous fields and described a family rather than one fixed patient trace.^[1]

Clinical evaluations established that the algorithmic choices recur as test strategies. In normal subjects, SITA testing took about half the time of Humphrey Full Threshold testing while threshold agreement and variability remained acceptable.^[2] The newer SITA Faster strategy was evaluated against SITA Fast and SITA Standard across glaucoma and glaucoma-suspect patients, demonstrating that the family can evolve while preserving its adaptive perimetric purpose.^[3] Exact clinical interchangeability, however, is an empirical matter rather than an invariant guaranteed by the family name.

Clarity

Imagine a field location for which the initial model predicts near-normal sensitivity. A bright stimulus is seen, so the algorithm can raise its estimate; a dimmer stimulus is missed, narrowing the plausible interval. Responses at neighboring locations and the field model help determine what evidence is worth gathering next. When the estimated error reaches the variant's stopping criterion, that location no longer consumes presentations.

If a patient's responses contradict the initial expectation, SITA must update rather than preserve the prior. This is why “prior-guided” does not mean “prior-determined.” A completed test is a spatial set of estimates supported by the patient's own response sequence, not a normative field imputed in place of measurement.

Manages Complexity

Conventional exhaustive thresholding multiplies a lengthy psychophysical search by dozens of locations. Patient fatigue can itself degrade attention and reliability, so test economy is part of measurement quality rather than mere convenience. SITA compresses the joint problem: exploit shared field structure, allocate presentations according to remaining uncertainty, estimate response reliability from ordinary test behavior, and stop gathering low-value evidence.

This organization also separates clinical output from internal search history. The clinician needs a coherent field result, local thresholds, and reliability information, not a manually interpreted log of every stimulus. Final recomputation integrates the complete response body into the reported estimates.^[1]

Abstract Reasoning

At each location \(j\), let \(\hat{t}_j\) denote the current threshold estimate and \(u_j\) its uncertainty. A response \(r\) to a stimulus of level \(s\) changes the evidence for plausible values of \(t_j\), and the field model couples that update to contextual information. Testing continues while \(u_j\) exceeds the strategy's allowed uncertainty and stops when its certainty rule is satisfied.

That schematic notation is intentionally not a claim that every SITA variant exposes one public Bayesian formula. The licensed inference is architectural: changing priors, update rules, uncertainty criteria, pacing, or reliability handling can change test duration and threshold behavior. A valid comparison must therefore hold the field pattern and patient population in view rather than treating all SITA-labeled outputs as numerically identical.

Knowledge Transfer

SITA illustrates a domain-bound instance of adaptive sequential measurement: combine prior structure with incoming observations, focus resources where uncertainty remains, and terminate when the required precision has been achieved. This skeleton transfers to other psychophysical and diagnostic tests only if the receiving setting has comparable sequential responses, estimable uncertainty, a defensible prior model, and explicit stopping criteria.

The name SITA should not transfer metaphorically to every adaptive test. Its literal identity retains automated perimetry, spatial visual-field sensitivities, disease-informed field modeling, and the recognized variant lineage. The substrate-neutral residue belongs to Algorithm, Measurement, and Adaptive Control.

Examples

1. SITA Standard: the accuracy-oriented family member used for routine automated threshold perimetry.
2. SITA Fast: a faster strategy with altered stopping and efficiency choices.
3. SITA Faster: a later variant evaluated in a multicenter study; mean test time was shorter than both Fast and Standard in that study.^[3]
4. Normal field: estimates settle near age-appropriate sensitivities, with responses updating rather than merely copying the normative model.
5. Glaucomatous field: local depressions require the model and stopping process to accommodate an abnormal spatial pattern.
6. Nonexample: a clinician manually comparing two printed field reports is interpretation of outputs, not execution of SITA.

Structural Tensions

• Speed vs. precision. Fewer stimuli reduce fatigue but can change uncertainty and effective dynamic range. Diagnostic: compare variant-specific stopping criteria, variability, and duration rather than accepting “faster” as uniformly equivalent.
• Prior structure vs. patient evidence. The model improves efficiency but must yield to discordant responses. Diagnostic: verify that response updates can move estimates away from the prior field.
• Local threshold vs. spatial field. Each location has an output, yet context across locations informs acquisition. Diagnostic: distinguish the reported local value from the joint model used to estimate it.
• Algorithm family vs. proprietary implementation detail. The published architecture is stable even when every parameter is not public. Diagnostic: require the model-update, uncertainty, and stopping package, not possession of one device.
• Autonomous abstraction vs. Algorithm plus Measurement. Generic parents explain procedure and evidence gathering but not SITA's perimetric field model and variant lineage. Diagnostic: subtract the parents and require interactive spatial threshold estimation with location-specific certainty stopping.

Structural–Framed Character

The structural core is an adaptive, uncertainty-governed sequential estimator. The frame is ophthalmic static perimetry: luminance stimuli, visual-field locations, sensitivity in decibels, normal and glaucomatous field knowledge, patient response reliability, and named SITA variants.

This combination supports a domain-specific abstraction. Removing the perimetric field model produces a broader algorithm class, while retaining it yields a repeatable identity across tests, patients, and strategy variants.

Structural Core vs. Domain Accent

Structural core: maintain estimates and uncertainties, update them from observations and contextual structure, allocate another observation only where useful, and stop when a precision target is met.

Domain accent: monocular visual-field testing, differential light sensitivity, spatial test grids, normal-age and glaucomatous patterns, false-response behavior, patient pacing, and Standard/Fast/Faster parameterizations.

Instantiates / Related Primes

SITA is a strict specialization of Algorithm: it is a defined procedure mapping a perimetric test configuration and sequential patient responses to local threshold estimates and reliability information. Measurement is constitutive context and Adaptive Control is structurally related, but adding either as another direct parent would obscure the minimal algorithmic identity.

Relationships to Other Abstractions

Current abstraction Swedish Interactive Thresholding Algorithm Domain-specific

Parents (1) — more general patterns this builds on

• Swedish Interactive Thresholding Algorithm is a kind of Algorithm Prime

  SITA is a strict specialization of Algorithm: it is a defined procedure mapping a perimetric test configuration and sequential patient responses to local threshold estimates and reliability information.

Hierarchy paths (2) — routes to 2 parentless roots

• Swedish Interactive Thresholding Algorithm → Algorithm → Function (Mapping)

Show alternative path (1)

Neighborhood in Abstraction Space

Swedish Interactive Thresholding Algorithm sits in a sparse region of the domain-specific corpus (95^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

• Psychophysical Scaling — 0.78
• Kriging — 0.78
• Depth Perception — 0.76
• Simon Effect — 0.76
• Electronystagmography — 0.76

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With

• Standard automated perimetry: the broader measurement practice in which SITA can be used.
• Humphrey Full Threshold: an older threshold strategy used as a comparator.
• SITA Fast / SITA Faster: variants within, not aliases for the whole family.
• Visual-field index or mean deviation: summaries computed from a field result, not threshold-search algorithms.
• Glaucoma diagnosis: a clinical judgment to which perimetry contributes.
• Generic Bayesian adaptive testing: a broader statistical architecture lacking SITA's ophthalmic identity.
• Short-wavelength automated perimetry: a stimulus/test modality rather than the SITA update-and-stop procedure.

References

[1] Boel Bengtsson, Jonas Olsson, Anders Heijl, and Holger Rootzén, “A New Generation of Algorithms for Computerized Threshold Perimetry, SITA,” Acta Ophthalmologica Scandinavica 75 (1997), 368–375, DOI: 10.1111/j.1600-0420.1997.tb00392.x, PMID 9374242. registry ↩a ↩b ↩c

[2] Boel Bengtsson, Anders Heijl, and Jonas Olsson, “Evaluation of a New Threshold Visual Field Strategy, SITA, in Normal Subjects,” Acta Ophthalmologica Scandinavica 76 (1998), 165–169, DOI: 10.1034/j.1600-0420.1998.760208.x, PMID 9591946. registry ↩

[3] Boel Bengtsson et al., “A New SITA Perimetric Threshold Testing Algorithm: Construction and a Multicenter Clinical Study,” American Journal of Ophthalmology 198 (2019), 154–165, DOI: 10.1016/j.ajo.2018.10.010, PMID 30336129. registry ↩a ↩b