The Application of Frontal and Maxillary Sinuses Indices for Computed Tomography-Based Sex Estimation of a Brazilian Population

. This investigation aimed to compare the sexing accuracy of frontal (FS) and maxillary (MS) sinus indices with multislice computed tomography (MSCT) in Brazilian adults. This cross-sectional study was conducted on 100 MSCT examinations (50 males and 50 females). The height, width, and diameter of the FSs and MSs were obtained with the RadiAnt software to measure four indexes (I): FSI-1 and MSI-1 (both width/ diameter); FSI-2 and MSI-2 (both height/diameter). Areas under receiver operating characteristic curves (AUROC) were determined. Sensitivity, specificity, likelihood ratio (LR), and accuracy values were also calculated. The highest area under the ROC curve for the FS was obtained by the FSI-1 (0.712 ± 0.052; p < 0.001), whereas for the MS it was found in the MSI-2 (0.431 ± 0.058; p = 0.232). FSI-1 showed greater specificity (64%) in predicting female sex as well as greater accuracy (64%) and LR (3.16). MSI-1 had the highest sensitivity value (68%) in estimating male sex and the second highest LR value (2.13). The FSI-1 developed in the present study exhibited higher accuracy compared to the other evaluated indices, which reinforces its importance for future studies with other populations.


Introduction
Sex estimation of human remains is a crucial forensic procedure during the identification process 1,2 . In this context, research has highlighted the use of teeth as an additional tool in the distinguishment between male and female individuals because of their ability to withstand destruction. However, it is noteworthy that sexual dimorphism (SD) can be observed in other structures of the human body, with the most prominent skeletal variations between the sexes being found in the bony pelvis and its adjacent bone structures, especially those that constitute the hip joint, such as the acetabulum and the head of the femur 3 .
The ability to classify the skull according to ethnicity is of great value when a skeleton or skull of unknown origin is found and no other methods of identification are possible due to decomposition 4 . In fact, the identification of human remains is one of the aspects of great relevance in the context of forensic anthropology, especially in investigation processes that require post-mortem identification 5,6 .
In the contemporary population, there is a relative scarcity of specific morphometric landmarks for sex estimation in skeletal remains from unknown individuals. This may reflect a historical evidence of insufficient and poor documentation of human skeletons 1,7 . In this context, the skull and its various landmarks, such as the supraorbital margin, nasal opening, and mastoid process are useful indicators of sex, supporting the sex estimation process with high levels of accuracy 8 .
There are several imaging modalities, ranging from conventional techniques, such as lateral cephalogram, to advanced technologies including multislice computed tomography (MSCT) and cone-beam computed tomography (CBCT) 9 . A current example that illustrates the relevance of modern imaging methods is the use of virtopsy, which has been gaining prominence in the field of forensic investigation.
This technique employs three-dimensional (3D) imaging methods routinely used in medicine to map the outer surface of the body where it is possible to record and document detail-rich images 10 .
Computed tomography (CT) facilitates the work of forensic anthropologists and pathologists by providing faster and more accurate identification 11 . The nasal cavity and the paranasal sinus (PS) are areas of significant structural variability and complexity that often require CT examination in preoperative evaluations and during the study of inflammatory nasosinusal diseases. This imaging exam has been increasingly used in recent decades to establish nasosinusal imaging parameters.
Moreover, this type of examination has yielded agreeable results with anatomical dissection studies 12 .
Several studies have indicated morphometric analysis of the maxillary sinus (MS) as an effective and faster method of sex identification 13,14 . The frontal sinus (FS) has also been proposed as an anatomical landmark, significantly improving sexing reliability. The term sinus index corresponds to the ratio between the height and width of the FS and it has been applied in lateral cephalometric radiographs to overcome limitations obtained from absolute cephalometric measurements 15 .
Several studies have investigated the potential of the frontal and maxillary sinuses during sex estimation. However, the available results suggest that the frontal sinus index has little potential, as successful identification indices have consistently been low.
In the present research, the indices (ratio between height and diameter, and width and diameter) of the frontal and maxillary sinuses were used as study variables. The aim of this study was to evaluate the accuracy of frontal and maxillary sinuses indices for sex estimation among adult individuals in the State of Ceará, Brazil.

Study design, context, and participants
This observational, cross-sectional, and retrospective study was carried out with MSCT images from the imaging service of a reference hospital in the State of Ceará, Brazil. Therefore, the recommendations for observational research proposed by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) initiative were followed 16

Variables
The sample comprised CT images of individuals of both sexes who suffered cranial and facial trauma in the last ten years and who also met the following eligibility criteria: a) aged between 18 and 39 years; b) maxillary dentition from the first D. Mendonça et al. premolar to the second molar. Tomographic images exhibiting signs of fracture, sinusopathies, or artifacts (such as dental implants and fracture fixation plates or screws), and images with inadequate quality for the evaluation of the sinus structures were excluded. CT images of individuals who had undergone more than one tomographic examination during the study period were removed to avoid duplicate data (in this case, the tomographic image with the earliest acquisition date was excluded).

Data sources and measurements
The MSCT images for the present study were obtained from a single tomography device (Somatom Emotion 6, Siemens, Forchheim, MedicalSolutions, Germany) under the following acquisition protocol: 1-mm table increment, 130 kVp, voltage setting ranging 80 to 140 mA, 2.0-mm image cross-sectional thickness, 180-mm field of view, and 0.6-second rotation time.
Data collection was performed by an observer previously trained to perform the analyses, which were repeated after 3 weeks. In this training stage, 20% of the study images were randomly chosen and used for the intra-examiner reliability analyses. For validation of measurements, from the five repetitions, Cronbach's Alpha was calculated to measure the internal consistency, the interclass correlation coefficient to verify the systematic error and the Hotelling's T-Squared multivariate analysis to calculate the random error (Table 1).
Linear evaluations were performed on the same computer (Dell Inc., model G3 3590, Intel® Core ™ i5-9300H processor CPU @ 2.40 GHz, 2400 Mhz, 4 cores, 8 logic processors -HD LED backlight display). The observer was allowed to adjust the screen brightness to perform the analyses. For the SP measurements, the axial, sagittal, and coronal sections were standardized with a slice thickness of up to 2.0 mm. The positioning of the sections followed a parallel plane to the hard palate in the coronal section overlapping with the median palatal suture in the axial plane. Height/Diameter (Figure 1). Table 2. Linear parameters evaluated in the frontal and maxillary sinuses in their widest section.

Definition View Parameter
Distance between the superior and inferior borders

Bias
To avoid selection bias, the sample size calculation aimed at estimating an adequate and equally divided number of individuals of both sexes.
After sample selection and standardization, MSCT scans with signs suggestive of pathological changes or trauma, which may be considered confounding factors, were excluded. As an additional effort to avoid measurement errors, imaging exams were evaluated by two trained observers who were blinded to the sex of the participants, and the reliability of these measurements was calculated.

Statistical methods
The Kolmogorov-Smirnov test was used to assess the normality of the data.
Measurements were expressed as mean and standard deviation values, and categorical data were expressed as absolute and relative frequencies. Bivariate analysis was performed with Student's t-test (linear measurements between males and females). The coefficient of variation was also calculated, and the homogeneity of variance regarding sex was assessed using Levene's test.
A receiver operating characteristics (ROC) curve was developed to identify SD-related cutoff points and to obtain the area under the curve (AUC), sensitivity, specificity, positive and negative predictive values, accuracy, and likelihood ratio. To establish an age-related subgroup analysis, sensitivity, specificity, positive and negative predictive values, accuracy, and likelihood ratio for individuals aged 30 years and older were calculated.

Reliability
The reliability with regard to the method was observed to be significant for the linear measurements relating to the sinuses, varying from satisfactory (r = 0.822) to highly satisfactory (r = 0.997). The paired t-test did not reveal any statistically significant difference between the initial set of measurements and the one repeated after 21 days. The ICC showed satisfactory values ranging from 0.896 to 0.998.

Reproducibility analysis
The validation analysis of the FS measurements exhibited high values of Cronbach's alpha (> 0.800) and interclass correlation index (ICC) (> 0.800), in addition to a significant correlation of Hotelling's T-squared (p < 0.001).

Cutoff points and ROC curves
The cutoff points based on the ROC curves ( Figure 2) to estimate SD are shown in Table 3. The FSI-1 presented a moderate predictive capacity (between 60 and 80%) with a significant likelihood ratio (3,16) and the highest accuracy values (64.0%) ( Table 4).

The measurements of the MS and FS indices were significant predictors of sexual dimorphism
The ROC curve-based ( Figure 2) cutoff points for the estimation of sexual dimorphism are shown in

Discussion
Sex estimation is one of the integral aspects of the personal identification of an unidentified corpse, thus requiring a very precise diagnosis 17 21 . Therefore, to reduce potential errors caused by these factors, 3D evaluations of tomographic exams must be performed. Sex estimation based on anthropometric methods has some limitations.
It is specific to the study population, which in turn depends on several factors such as genetics, maternal health culture, environmental quality conditions during prenatal development, and race 27 . Therefore, the sex estimation measurements presented in this study may be more appropriate for a specific group, and the reliability of these measurements and the degree of SD evidenced in this article might be questionable for other populations. Therefore, specific assessments are required for each population, with further studies with larger samples being necessary to better represent these groups.

L. P. Miranda & T. L. Oliveira
The methodology proposed for SD assessment in this study was based on indices of the frontal and maxillary sinuses. We used the ratio between linear measurements (indices) of MSCTs to assess the sinuses in our study. This study could help expand and prompt new research possibilities for the identification of humans, as the development of new indices might improve the selection of indicators for sex estimation.

Conclusion
In the present investigation, expressive mean values of the maxillary and frontal sinuses were found in both males and females. High values of reproducibility and precision were also observed.
The frontal and maxillary sinus indices constituted a valuable tool in the study of SD in MSCT images. The results of the present study are reliable and reproducible due to the standardization of the measurement procedures and the use of sagittal MSCT views.
These results indicate a potential role of FSI-1 as an auxiliary tool for sex estimation. We believe that this approach could be further investigated using other MSCT views or other imaging modalities such as CBCT focusing on SD.