Ethmoid Roof (Keros Classification): Global CT and CBCT Meta-analytic Synthesis

The ethmoid roof and olfactory fossa form the thinnest segment of the anterior skull base and are a well-recognised locus minoris resistentiae during functional endoscopic sinus surgery (FESS) and extended endonasal skull-base procedures.¹¹ The lateral lamella of the cribriform plate is particularly fragile, and its depth determines how close the surgeon works to intracranial contents.

Keros originally classified olfactory fossa depth into three types based on the height of the lateral lamella: type I (1–3 mm), type II (4–7 mm) and type III (8–16 mm), with increasing risk of skull-base injury from type I to III.¹ Subsequent CT-based studies have confirmed that Keros II usually predominates, but the proportion of deep type III fossae varies widely between populations.

A recent systematic review and meta-analysis of lateral nasal wall and anterior skull-base variants highlighted significant geographical differences in Keros distribution and ethmoid roof asymmetry, with implications for complication risk in FESS and extended approaches.² CT and CBCT have therefore become mandatory preoperative tools, providing a three-dimensional roadmap of the ethmoid bone, anterior ethmoidal artery corridor and skull base.⁹

Individual CT series from Brazil, Egypt, Turkey, India and Pakistan show broadly similar patterns of Keros II predominance but differ in the relative proportions of types I and III, and in the frequency of inter-side discordance.^{3, 4} Quantifying these differences is important for understanding population-specific risk and for counselling surgeons operating in regions with a higher burden of deep fossae.

This article synthesises CT- and CBCT-based data on Keros classification and ethmoid roof asymmetry, emphasising global patterns, regional outliers and practical messages for skull-base and sinus surgeons.

This synthesis draws on a recent systematic review and meta-analysis of lateral nasal wall and anterior skull-base anatomy, combined with selected large CT and CBCT studies that specifically report Keros type distribution and/or ethmoid roof asymmetry.² Priority was given to radiological cohorts with ≥200 patients or ≥400 fossae, clear Keros definitions and side-specific reporting.

Key CT cohorts included Brazilian, Egyptian, Turkish, Indian and Pakistani series of adults or mixed-age populations undergoing nasal cavity and paranasal sinus imaging for routine clinical indications.^{3, 5} CBCT data from a 2022 study of ethmoid roof and skull-base variants were also considered to illustrate how newer volume imaging compares with multi-slice CT.³

For each study we extracted the relative frequencies of Keros types I–III (per fossa or per side), any reported subject-level Keros asymmetry, and measures of roof height or lateral lamella configuration asymmetry where available.^{4, 7} Because many primary reports used differing denominators and age ranges, we present descriptive percentages rounded to the nearest whole number rather than recalculating pooled effect sizes.

Complication data and broader skull-base risk considerations were summarised from large FESS series and skull-base imaging reviews, focusing on how Keros classification and roof asymmetry influence intraoperative decision-making rather than on formal risk modelling.^{10, 12}

Global Keros Distribution

Across modern CT and CBCT cohorts, Keros II is consistently the most frequent pattern, usually accounting for about two-thirds of fossae. In a large Brazilian CT study of 200 patients, Souza et al. reported Keros II in 73.3% of fossae, Keros I in 26.3% and Keros III in only 0.5%.⁷ Egyptian MDCT work similarly found Keros II to be the dominant pattern, with Keros I common and Keros III restricted to a very small minority of cases.⁶

A paediatric CT series from Turkey (780 fossae) demonstrated Keros I, II and III in 24.7%, 65.9% and 9.4% respectively, again confirming a strong Keros II predominance but with a higher proportion of deep fossae than in the Brazilian cohort.³ Asymmetry of olfactory fossa depth in this cohort was identified in 7.4% of patients, showing that even in children, side-specific differences are common.³

CBCT data assessing the ethmoid roof using Keros and related classifications in mixed-age populations have yielded similar overall proportions, with Keros II typically around 60–70% and Keros III remaining the least frequent but clinically critical category.³ Figure 1 summarises the distribution of Keros types from the large Turkish paediatric CT series as a representative modern dataset.

Regional Variation and Keros III

Kaplanoglu et al. analysed 1,000 fossae from 500 Turkish adults and found Keros I, II and III in 13.4%, 76.1% and 10.5% respectively, indicating a relatively high burden of deep fossae compared with Brazilian and Egyptian series.⁵ The same study reported Keros type asymmetry (different Keros types on each side) in around 6% of patients, underscoring the need to evaluate each side separately.⁵

In an Indian tertiary-care CT cohort of 256 patients, Naaz et al. observed that 82% of subjects had Keros II on both sides, whereas Keros III occurred in 5.1% of right and 3.5% of left fossae, with Keros type discordance between sides in 16.8% of patients.⁴ Asymmetry of lateral lamella height, using a broader definition, was present in 40.6% of patients, much higher than the limited Keros-type discordance alone would suggest.⁴

Egyptian MDCT work by Elwany and colleagues reported Keros I and II in approximately 42.5% and 56.8% of fossae respectively, with Keros III detected in a very small fraction of male patients and none of the women, placing overall Keros III prevalence near 1%.⁶ Brazilian CT data from Souza et al. showed an even lower Keros III prevalence of 0.5%, indicating a comparatively safer ethmoid roof profile in that cohort.⁷

By contrast, a recent Pakistani CT study from Karachi using combined Keros and Yenigun classifications found Keros II, III and I in 43%, 31% and 26% of fossae respectively, highlighting a substantially larger proportion of deep type III fossae than in many previous series.⁸ Such regional variation aligns with the broader systematic review by Zahedi et al., which showed significant geographical clustering of high-risk anterior skull-base configurations.²

Figure 2 maps the reported prevalence of Keros III across key CT cohorts to illustrate how the burden of deep fossae – and therefore potential skull-base risk – shifts between populations.

Ethmoid Roof Asymmetry

Beyond absolute Keros type, asymmetry of roof height and lateral lamella configuration is a major determinant of skull-base risk, because inadvertent dissection tends to follow the shallower side. In Souza’s Brazilian CT cohort, height asymmetry of the ethmoid roof was present in 12% of patients and contour asymmetry in 48.5%, with flattening of the roof on one side in nearly half of cases.⁷

Kaplanoglu et al. defined significant height asymmetry as a >1 mm difference in lateral lamella depth and found this in 80% of patients, although discordance in Keros type between sides was much less frequent at approximately 6%.⁵ This pattern suggests that surgeons should not rely solely on Keros type, but also on absolute depth and roof slope, when planning their superior dissection plane.

In the Indian cohort of Naaz et al., 40.6% of patients demonstrated lateral lamella height asymmetry, and nearly one in six had different Keros types on each side.⁴ Larger Indian CT series of 1,200 patients by Babu et al. likewise reported substantial inter-side variation in olfactory fossa depth, with discordant Keros types in roughly one fifth of cases.⁹

Paediatric Turkish data from Güven et al. add further nuance, showing Keros asymmetry in 7.4% of children despite an overall Keros II predominance, indicating that clinically relevant side differences are already established early in life.³ Figure 3 compares Keros III prevalence with roof asymmetry rates across three well-characterised CT cohorts to emphasise how deep fossae and asymmetry often cluster but do not perfectly overlap.

This synthesis reinforces three central messages for surgeons working around the ethmoid roof. First, Keros II is the predominant pattern worldwide in both CT and CBCT datasets, generally representing around two thirds of fossae, with type I common and type III consistently the least frequent but most hazardous variant.^{2, 7} This supports continued use of Keros classification as a simple, clinically meaningful descriptor of anterior skull-base depth.

Second, there is genuine regional variation in the burden of deep fossae. Brazilian and Egyptian cohorts show Keros III prevalences below 2%, whereas Turkish and Indian series report values around 9–11% and 3–5% respectively, and a recent Karachi cohort reports Keros III in about one third of fossae.^{4, 8} These differences likely reflect a combination of ethnic skull-base morphology, age structure, imaging selection and methodological choices, and they align with meta-analytic evidence that high-risk anterior skull-base configurations cluster geographically.²

Third, asymmetry and lateral lamella slope are at least as important as absolute Keros type. Studies using strict measurement criteria have reported lateral lamella height asymmetry in 40–80% of patients, even when Keros type is symmetrical.^{4, 5} This explains why major skull-base complications continue to occur despite routine preoperative CT: the surgeon may instinctively “follow” the shallower side or misjudge the upward angulation of the roof, particularly in Keros III fossae with long, oblique lateral lamellae.

Large FESS series and skull-base imaging reviews emphasise that meticulous preoperative scrutiny of coronal and sagittal reconstructions, with explicit documentation of Keros type on each side, the lowest roof point, and any sharp angulations or dehiscences, can markedly reduce the risk of CSF leak, orbital injury and intracranial haemorrhage.^{10, 12} Anterior ethmoidal artery position, supraorbital cells and other skull-base variants further modulate risk and should be interpreted in conjunction with Keros classification rather than in isolation.¹⁰

Finally, the available evidence underscores the limitations of extrapolating Keros distributions between populations. Centres serving patients from regions with higher Keros III prevalence – as suggested by Turkish, South Asian and Middle Eastern CT data – should maintain a lower threshold for image-guided navigation, staged dissection and skull-base backup when tackling complex inflammatory or neoplastic disease.

CT- and CBCT-based studies from multiple continents confirm that Keros II is the dominant ethmoid roof pattern, but the proportion of deep Keros III fossae and the frequency of roof asymmetry vary substantially between populations. High-quality series from Turkey, India and Pakistan suggest that some cohorts carry a markedly greater burden of deep fossae than others.^{4, 8}

For endoscopic sinus and anterior skull-base surgery, Keros classification remains a useful shorthand only when interpreted alongside side-specific roof height, lateral lamella slope and adjacent vascular and neural landmarks. Routine, systematic review of these features on preoperative CT – and clear communication of any high-risk patterns – is essential to minimise the risk of skull-base and orbital injury.