CBCT Assessment of Ethmoid Roof Variations through Keros, Gera, and TMS Classifications

Background This study is designed to assess the variations of the ethmoid roof through the Keros, Gera, and Thailand-Malaysia-Singapore (TMS) classifications by means of Cone-Beam Computed Tomography (CBCT). Materials and Methods 372 CBCT scans were reviewed. The depth of the olfactory fossa (OF) was defined by the height of the lateral lamella (LL). The degree of the angle formed by the LL and the continuation of the horizontal plane passing through the cribriform plate was calculated. The risk of injury to the skull base was assessed by TMS classification. Results The distribution of Keros types 1, 2, and 3 was 20.43%, 66.26%, and 13.31%, respectively. No significant difference was seen in the Keros classification among males and females, and asymmetry in the depth of the cribriform plate was seen in 33.3% of patients. 29.57% of the cases had type I, 61.42% had type II, and 9.01% had type III of Gera classification. 33.9% of the patients had asymmetry in the Gera classification. There was not significant difference in the Gera classification among males and females. 95.43%, 4.17%, and 0.40% of patients were of types 1, 2, and 3 of TMS classification, respectively. Conclusion The most prevalent variations of the ethmoid roof were Keros type II, Gera type II, and TMS type 1. Although the prevalence of the dangerous types (Keros type 3, Gera type III, and TMS type 3) was low, preoperative assessment is essential in reducing surgical complications. CBCT is beneficial in evaluating these variations due to its low radiation dose.


Introduction
Functional endoscopic sinus surgery (FESS) is highly performed as the current treatment for diseases of nose and paranasal sinuses (PNS). Although there is considerable reduction in the incidence of associated complications in FESS comparing with traditional approaches, it has complications such as cerebrospinal fluid leak, orbital hurt, and intracranial injury [1][2][3]. e olfactory fossa (OF) is a depression in the anterior cranial cavity. e cribriform plate (CP) of ethmoid forms its floor. is delicate bony plate separates the anterior cranial fossa from the nasal cavity. Lateral and medial limitations of OF are lateral lamella (LL) of cribriform plate and crista galli, respectively [4]. OF contains the olfactory bulbs and tracts. e lateral boundary of the CP of the ethmoid bone is called lamina lateralis (LL). It runs vertically and joins the fovea ethmoidalis inferomedially. e LL is the thinnest bone in the anterior skull base offering lowest resistor to perforation during surgical maneuvers. Hence, it is a high-risk site where eventually surgical complications occur [5,6]. If one of the lateral lamellae is inferiorly positioned, reporting should be considered, as this increases the risk of intraoperative injury. erefore, asymmetry in the OF depth or the height of LL is related with higher risk of intracranial penetration during surgeries such as FESS. (8-16 mm) [7]. e Keros type III is the most susceptible one, considering the major risk for iatrogenic injury of the LL of the CP [8].

Gera Classification.
It is the degree of the angle at which the lateral lamella joints on the cribriform plate show variations among patients. Also, the sloping shape of the skull causes limitations in the Keros classification in terms of defining the risk of intracranial penetration. erefore, the Gera classification was introduced to consider the sloping shape of ethmoid roof (ER) relative to CP by measuring the angle formed by the LL of CP and the continuation of the horizontal plane passing through CP. e risk of intracranial entry was divided into 3 classes: class I (>80 degrees, low risk), class II (45 to 80 degrees, medium risk), and class III (<45 degrees, high risk) [9,10].

TMS Classification.
e ailand-Malaysia-Singapore (TMS) Classification was proposed as a new practical radiological classification that completes the Keros and Gera classifications and assesses the anatomical risk of anterior skull base injury by means of the orbital floor as a reference. e distances from orbital floor to cribriform plate (OFL-CP) and from orbital floor to ethmoid roof (OFL-ER) were classified into type 1, type 2, and type 3 [11].
Radiographic analysis using multidetector computed tomography (MDCT) has been supposed to be the gold standard in the presurgical evaluation of the PNS, which provides a knowledge of the anatomic variations of the ethmoid roof in every patient. Cone beam computed tomography (CBCT) is widely applied in dentistry and otorhinolaryngology for analysis of PNS due to its low radiation exposure and high-resolution images. Variations of ethmoid roof have been assessed using the Keros, Gera, and TMS classifications on MDCT images. However, there are a few studies in the literature that have used CBCT images to assess those anatomic variations .
is study is clinically significant as it can show the ability of CBCT devices to show the anatomic variations of the ethmoid roof in the preoperative examinations of patients, while its radiation dose is far less than that of MDCT. So, this study is designed to assess the variations of ethmoid roof through the Keros, Gera, and TMS classifications using CBCT images. For choosing proper images, each of the 900 CBCT scans were opened on a computer viewer (E2042C, Korea) through Galileos software, and the panorama view was observed. en, all of the coronal sections in the PNS area were scrolled, and the inclusion and exclusion criteria were checked in panorama and coronal sections of each CBCT scan.

Materials and Methods
e inclusion criteria consisted of scans of patients with completely developed PNS, acquired for any reason (i.e., implant treatment planning, impacted teeth, orthodontic treatment planning, etc.) showing the crista galli of the ethmoidal bone and nasal fossa, with no gender consideration.
Patients with a history of paranasal sinus surgery, maxillofacial trauma, and pathological processes in the PNS were excluded from the study. Also, low-quality images, images containing only lower jaw, and images containing artifacts, which made visualization of anatomical structures difficult, were excluded from the evaluation. Finally, based on the above criteria, 372 CBCT scans were selected for the study.

Image Analysis.
All of the 372 CBCT scans were evaluated by one oral and maxillofacial radiologist with experience in CBCT imaging twice, with a month interval on an LG LED computer viewer (E2042C, Korea) by means of Galileos software. e following anatomical landmarks were identified on coronal sections and used for measurements:

Keros Classification.
e depth of the OF was defined by the height of the LL on the coronal sections and at the level of infraorbital foramina. e linear measurements (in millimeters) were done using the Galileos software's ruler tool for both lateral lamellae of the CP (right and left sides). Horizontal lines were drawn along the CP and at the ER. e vertical interval between these two horizontal lines was considered as the height of LL (Figures 1-3). OF depth was classified into three categories: [7].
(i) Type 1: has a depth of 1-3.9 mm (low risk) (ii) Type 2: has a depth of 4-7.9 mm (medium risk) (iii) Type 3: has a depth of 8-16 mm (high risk). e type of OF depth was determined for each side of a patient and the difference between the right and left side, considered as asymmetric LL.

2
International Journal of Otolaryngology

Gera Classification.
e degree of the angle formed by the LL and the continuation of the horizontal plane passing through the CP was calculated using the Galileos software's measure-angle tool for each side. According to Gera system, the angle is classified depending on its degree and on the hypothetical risk of iatrogenic injuries (Figures 4-6) [9].
(i) Class I (>80 degrees, low risk) (ii) Class II (45 to 80 degrees, medium risk) (iii) Class III (<45 degrees, high risk). e difference between the right and left sides was considered as asymmetry in the position of LL.

TMS Classification.
In each side, the orbital floor was defined at the point where the medial wall of the maxillary sinus roof was observed at its maximum height. e distances from OFL to CP (OFL-CP) and from OFL to ER (OFL-ER) were used in each side using the Galileos software's ruler tool. Horizontal lines were drawn along the CP and at the OFL and ER. e vertical interval between OFL and CP horizontal lines was considered as the OFL-CP distance, and the vertical distance between OFL and ER horizontal lines was considered as the OFL-ER distance ( Figure 7). e risk of skull base injury was classified into three groups as follows (Figures 8-10) [11]: (i) Type 1 (low risk) is when both OFL-CP and OFL-ER are 10 mm and above, or more than twice the depth of thru-cutting forceps when ESS is supposed to be safe. (ii) Type 2 (moderate risk) is when either OFL-CP or OFL-ER is less than 10 mm, or less than twice the depth of thru-cutting forceps when ESS should go forward cautiously. (iii) Type 3 (high risk) is when both OFL-CP and OFL-ER are less than 10 mm or less than the depth of thru-cutting forceps when ESS should proceed very cautiously.

Statistical Analysis.
Descriptive parametric data were offered as percentage. e Kappa coefficient was calculated for the agreement between the data obtained from the first   International Journal of Otolaryngology 3 and second observations in each classification. Wilcoxon signed ranks test was applied to assess the difference in scores of each classification on the right and left side. e U Mann-Whitney test was used to assess the difference of classification scores in gender. e correlations between TMS, Keros, and Gera classifications were estimated using Pearson's correlation coefficient (r). e statistical significance was set to P < 0.05.

Demographic Data.
CBCT scans of 372 subjects were contained in the study. Among 372 subjects, 195 were males (52.4%) and 177 were females (47.6%). e youngest subject was 14 years, and the eldest was of 96 years of age, with a mean age of 41 and a mode age of 19.

Keros Classification.
e Kappa coefficient for the agreement between the first and second observations of Keros classification was 0.978 (P-value � 0.00) in the left side and 0.984 (P-value � 0.00) in the right side. Due to the high agreement between the first and second observations, the data of the first observation was used in statistical analyzes.     Table 2 shows the distribution of keros scores according to the side. Table 1 presents the distribution of Keros scores according to their sides and sex. e U Mann-Whitney test showed no significant statistical difference in the distribution of the Keros scores among male and female. e Kappa coefficient for the agreement of Keros scores between the right and left side was 0.336 (P-value � 0.00). Table 3 shows that 66.7% of images had similar Keros scores, and 33.3% of the images had different Keros scores in the right and left sides. e Wilcoxon signed ranks test showed a significant difference between the Keros scores of the right and left side (P-value � 0.03).

Gera Classification.
e Kappa coefficient for the agreement between the first and second observations of Gera classification was 0.945 (P-value � 0.00) in the left side and 0.926 (P-value � 0.00) in the right side. Due to the high agreement between the first and second observations, the data of the first observation was used in statistical analyses. Table 2 shows the distribution of Gera scores according to the side. Table 1 presents the distribution of Gera scores according to their sides and sex. e U Mann-Whitney test showed no significant statistical difference in the distribution of the Gera scores among male and female. e Kappa coefficient for the agreement of Gera scores between the right and left side was 0.361 (P-value � 0.00). Table 4 shows that 66.1% of images had similar Gera scores, and 33.9% of the images had different Gera scores in the right and left sides. e Wilcoxon signed ranks test showed a significant difference between the Gera scores of the right and left sides (P-value � 0.011).

TMS Classification.
e Kappa coefficient for the agreement between the first and second observations of TMS classification was 0.967 (P-value � 0.00) in the left side and 0.973 (P-value � 0.00) in the right side. Due to the high agreement between the first and second observations, the data of the first observation was used in statistical analyzes. International Journal of Otolaryngology Table 2 shows the distribution of TMS scores according to the side. Table 1 presents the distribution of TMS scores according to their sides and sex. e U Mann-Whitney test showed a significant statistical difference between gender and TMS scores of the left side (P-value � 0.041).
e Kappa coefficient for the agreement of TMS scores between the right and left side was 0.571 (P-value � 0.00). Table 5 shows that 96.2% of images had similar TMS scores and 3.8% of the images had different TMS scores in the right and left sides. e Wilcoxon signed ranks test showed no significant difference between the TMS scores of the right and left sides (P-value � 0.467).

Comparison of Skull Base Classifications.
e correlation between Gera, TMS, and Keros classifications was assessed ( Table 6). Table 6 showed a weak negative correlation between Gera and Keros classifications in the right and left sides (P-value � 0.00) and a positive correlation between the right and left sides in each classification (P-value � 0.00).
ere was no significant correlation between age and Gera, Keros, and TMS classifications ( Table 6).

Discussion
FESS is a common method for the management of PNS disease. LL, which is the lateral boundary of the CP of the ethmoid bone, is considered a high risk site where most surgical complications occur. It may be prominent and protrude into the anterior ethmoid sinus. e LL may be harmed during operations, specifically when it protrudes into the sinus cavity. LL is susceptible to surgical trauma when the disease within the frontal recess along the superolateral wall of the middle turbinate is removed, particularly when using biting forceps.
us, a great attention to the LL of the CP plate is required during endoscopic ethmoid and frontal sinus surgeries to prevent potential complications [5,6].  [7]. In this study, the distribution of Keros types 1, 2, and 3 was 20.43%, 66.26%, and 13.31%, respectively. Table 7 summarized the incidence of Keros types in different populations and showed that most studies were done using MDCT, and only two studies have been performed using CBCT so far.
About the distribution of Keros types, type 2 was the most frequently viewed in this survey, which was similar to the majority of previous reports [1, 3, 4, 7, 8, 10, 11, 13, 14, 16-19, 22-24, 26-28, 30-32]. e most similar one to this study is the costa et al.'s study, which was done on Brazilian population using CBCT. ey have reported a frequency of 65.52 for Keros type 2 [13].
is study showed that Keros type 3 had the lowest distribution among the Iranian population, and the distribution of Keros type 1 was more than Keros type 3, while, in the articles of costa et al. and Guldner et al. using CBCT, with different sample size, the distribution of Keros type 3 was more than type 1, which might be due to the ethnical differences. Keros type 3 patients have been reported to be more susceptible to iatrogenic skull-base injuries [13,18,29].
is study showed no significant difference in the Keros classification among males and females, which is in accordance with some studies [9,13,14,24]. However, the distribution of Keros type between two genders is reported to be different in some other studies [1,3,13].
Asymmetry in the depth of the cribriform plate was seen in 33 [1,3,9,13,25,30]. No significant difference is reported in the distribution of Keros type between the left and right sides in some studies [12,13,24]. Considering the presence of asymmetry in many patients, preoperative assessments are required through MDCT or CBCT to avoid the risk of intracranial penetration during endoscopic sinus operation. Table 7 summarized the incidence of Keros types in different populations.

Gera Classification. Gera et al. introduced another classification due to the limitations of Keros classification in
2018. According to this classification, a more pronounced slop of the anterior skull base on the coronal plane or the Gera type III may predispose to iatrogenic skull base injuries [9]. By using Gera classification in this study, 29.57% of our CBCT cases had type I (low risk), 61.42% had type II (moderate risk), and 9.01% had type III (high risk). e distribution of Gera classification of this study was similar to the previous reports [9,11]. Table 8 summarized the incidence of Gera types in different populations and showed that all of the studies were done using CT, and there was no study using CBCT so far.
Preti et al. compared Keros classification with Gera classification on CT scans of 124 patients. ey concluded that Gera classification showed higher sensitivity, specificity values on preoperative CT scan risk evaluation and therefore suggested to use it in combination with keros classification to prevent injuries during FESS [10].
Asymmetry in the Gera classification was found in 33.9% of the patients of this study. None of previous studies have reported asymmetry between the two sides in Gera classification [9,11].
is study showed no significant difference in the Gera classification between two genders, which is in accordance with other studies [9,11]. e present study found a weak negative correlation between Gera and Keros classifications. Similarly, Gera et al. reported a negative correlation between them, while Abdullah et al. presented a weak positive correlation between them. Anyway, both studies have suggested to use both classifications for the anterior skull base injuries assessments [11].

TMS Classification.
Abdullah et al. introduced the TMS classification using CT scans in 2020 to determine the risk of skull base damage preoperatively and intraoperatively. ey concluded that the combination of TMS classification with Keros and Gera classifications provided a superior intracranial injury risk assessment before and during FESS [11]. International Journal of Otolaryngology In the present study, using CBCT, 95.43% had type 1 (low risk), 4.17% had type 2 (moderate risk), and 0.40% had type 3 (high risk) according to TMS classification. ere was a significant difference in the distribution of the TMS scores between two genders on the left side (P-value � 0.041), and the frequency of each TMS scores was higher in females. e right and left side asymmetry in the TMS scores was not significant.
TMS type 1 was the most current viewed in this study, which was close to Abdullah et al.'s findings. However, the distribution of TMS types 2 and 3 in the present study differed in that TMS type 3 was reported to be more frequent than type 2 in their study, which might be due to our greater sample size or ethnic differences. Moreover, the present study showed a significant difference in the distribution of the TMS scores between two genders on the left side, while such difference was not reported in Abdullah et al.'s study [11].
No significant correlation was observed between Gera and TMS classifications in this study, which was in accordance with Abdullah et al.'s findings. However, Abdullah et al.'s study reported a significant correlation between TMS

Conclusion
e most common type of OF in the study population was Keros type II. e most prevalent degree of the angle formed by the LL and the continuation of the horizontal plane passing through the CP was 45 to 80 degree (type II) according to Gera classification. e majority of our cases showed low risk for skull base injury (TMS type 1). Even though the prevalence of the high-risk types (Keros type 3, Gera type III, and TMS type 3) was low, preoperative assessment is essential to reduce surgical complications. Due to the low dose of radiation in CBCT, it is beneficial to evaluate the anterior skull base variations.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
All the authors have made significant contributions to this work, with all co-authors approving the final version of this article and agreeing with its submission for publication.  International Journal of Otolaryngology 9