Linear measurements of the mandible on panoramic radiograph for sex estimation in populations in Yogyakarta, Indonesia

Abstract

Background

Identification is one of the main aspects of forensics. Sex estimation is an essential part of identification because it can simplify the whole process. Mandible is the largest, strongest, and sexually dimorphic bone and is part of the orofacial structure, which has the second highest level of dimorphism after the pelvic bone. Sex estimation using the mandible can be performed by conducting various linear measurements on a panoramic radiograph, including right minimum ramus width (RMiRW), left minimum ramus width (LMiRW), right projective height of ramus (RPHR), left projective height of ramus (LPHR), right maximum ramus width (RMxRW), left maximum ramus width (LMxRW), right coronoid height (RCH), left coronoid height (LCH), symphysis height (SH), right mandibular corpus height (RMCH), and left mandibular corpus height (LMCH). This study aimed to analyze how linear measurements in panoramic radiographs differ between men and women and to estimate sex by utilizing these parameters.

Results

In this study, 195 panoramic radiographs from 95 men and 100 women aged 20–40 years obtained from our dental hospital were used as a training data. Meanwhile, 61 panoramic radiographs from 29 men and 32 women outside the training data with same characteristics were used as the testing data. The linear measurements of the mandible using panoramic radiographs were taken with EzDent-i Vatech software. Independent t-test showed significant difference (p < 0.05) in some linear measurements of the mandible between adult men and women. These significantly different linear measurements were then subjected to discriminant function analysis to produce sex estimation equations. The equation accuracy percentage ranged between 63.6 and 94.4% for the training sample and 59.5% and 85% for the testing sample.

Conclusion

The linear measurements of the mandible using diagnostically acceptable panoramic radiographs taken from a patient with standardized head positioning can serve as an alternative method for sex estimation. The accuracy of discriminant analysis for sex estimation varies depending on the parameter used in the estimation.

Background

Forensic odontology, a branch of dentistry, deals with the appropriate handling of dental evidence and findings for identification or legal issues (Krishan et al. 2015). The main aspect of forensics is identification, which assists investigators in determining an individual’s identity, such as sex, age, and specific population. Sex is one of the keys in identification. Given that men and women are the only two biological sexes in the world, the identification becomes easier when the sex is known because other possibilities can be eliminated; however, rare sex-related genetic disorders must be considered (Mello-Gentil and Souza-Mello 2022).

Sexually dimorphic bones are required in sex identification procedures. The mandible has been proven to be applicable in sex estimation procedures with high accuracy levels, just below the level of pelvic (Brogdon 2011). The mandible is a dense bone and resistant to impact; thus, it remains intact and usable for the identification of incomplete remains. Moreover, the mandible also provides sexually dimorphic characteristics due to differences in its growth duration, developmental phase, and muscle activity, which significantly differ between men and women (Toneva et al. 2021). The mandible in men has a larger size and tends to have a rougher shape compared with those in women (Brogdon 2011). The postpubertal period is chosen for measuring mandibular parameters for sex estimation, specifically from the age of 20 years when active mandibular growth has been achieved and thus sexual dimorphism can be observed due to hormone influences and mandibular growth stabilization until the age of 40 years before the aging process begins (Aurizanti et al. 2017).

The use of radiographs for identification has been widely accepted because it is a noninvasive, simple, and cost-effective method (Nagi et al. 2019; Zhang 2022). Panoramic radiography is an extraoral radiographic technique commonly employed because of its wide coverage of the oral and maxillofacial area, quick procedure, and low radiation exposure (Sairam et al. 2016). This approach provides a comprehensive 2D depiction of the oromaxillofacial region in a single image. It also offers morphological and morphometric information on bones during growth process and is accurate in measuring the vertical and horizontal dimensions of the mandible (Indira et al. 2012; Rad et al. 2020).Several previous studies have utilized multiple parameters in the mandible have been utilized to achieve a high level of accuracy in sex estimation (Prasetiowati et al. 2023). Nevertheless, the accuracy of these functions as sex predictors tends to decline when applied to samples that differ from the reference population. Hence, a sex estimation formula specific to the population must be created. This study analyzed various linear measurements of the mandible on panoramic radiographs that have been selected from a literature review and used them to construct a specialized sex estimation formula for the 20- to 40-year-old population in Yogyakarta, Indonesia, which has not been thoroughly explored.

Methods

Study design and ethical aspects

This analytical observational study had a cross-sectional design. Digital panoramic radiographs were collected retrospectively from the dental records in the Professor Soedomo Dental Hospital of Universitas Gadjah Mada, Yogyakarta, Indonesia. Ethical clearance was obtained from the Ethics Commission of the Faculty of Dentistry and Dental Hospital of Universitas Gadjah Mada (ref. no. 70/UN1/KEP/FKG-RSGM/EC/2023).

Sample characteristics

A total of 195 panoramic radiographs (95 men and 100 women) for training data and 61 samples (29 men and 32 women) for testing data were taken from patients aged 20–40 years who came to the university dental hospital from January 2021 to March 2024. These radiographs were taken based on the following inclusion criteria: (1) completely erupted teeth (excluding third molars); (2) the radiographic image covers the entire area that must be covered on a panoramic radiograph, namely the teeth, mandibular ramus and cervical vertebrae, nasal cavity and maxillary sinus, mandibular body, mandibular condyle, and hyoid bone; (3) minimal horizontal distortion and vertical distortion; (4) no artifacts, ghost images, or foreign objects (such as metal accessories, orthodontic appliances, glasses, and dentures) that could interfere with the anatomical image of the radiograph; (5) the hard palate and floor of the nasal cavity are clearly visible (not superimposed on the apices of the maxillary teeth); (6) the maxillary and mandibular anterior teeth are clearly visible; and (7) the radiograph has a right “R” and/or left “L” marker. The exclusion criteria were as follows: (1) radiographic images indicate errors in patient positioning during imaging, namely the patient’s neck is too extended and produce a ghost image of cervical vertebrae on the anterior teeth, head position is not on the focal trough and caused the mesiodistal width of anterior teeth to decrease if it too far forward and the image of the anterior teeth to widen if it too far back, the Frankfort horizontal plane (FHP) is not parallel to the floor either due to the patient’s head being too bent down and caused the occlusal plane to appear like the letter “V” or being too upturned and caused the occlusal plane to be inverted (reversed smile line), the patient’s sagittal plane is not perpendicular to the floor or turns to one side and caused the part closer to the receptor to become narrow and the part away from it to widen, the patient’s tongue is not on the palate and produced a radiolucent image covering the apex of maxillary anterior teeth, the patient moves during imaging and caused a deformity in the mandible, and the patient’s mouth is open and caused a distortion of anterior teeth; (2) tooth loss in the mandible (partial/fully edentulous); (3) pathological or traumatic lesions in the mandible; and (4) radiographic features indicating temporomandibular joint disorders, including condyle hyperplasia (unilateral/bilateral), condyle hypoplasia (unilateral/bilateral), coronoid process hyperplasia, ankylosis, tumors, juvenile arthrosis, bifid condyle, osteoarthritis, and rheumatoid arthritis.

Variable measurements

Measurements were conducted using the EzDent-i Vatech software (Gyeonggi-do, South Korea). The measurement data included minimum ramus width (MiRW), projective height of ramus (PHR), maximum ramus width (MxRW), coronoid height (CH), symphysis height (SH), and mandibular corpus height (MCH).

MiRW is the smallest anterior–posterior diameter of the ramus. PHR is the point between the highest part of the mandibular condyle and the lowest border of the bone. MxRW is the distance between the most anterior point on the mandibular ramus and a line connecting the most posterior point on the condyle and the angle of jaw (Sairam et al. 2016). CH is the measured distance between coronation and most protruding portion of the inferior border of the ramus. SH is the distance from infradentale between the central incisors to menton (Fekonja and Čretnik 2022; Rad et al. 2020). MCH is the distance from the alveolar process to the inferior border of the mandible, perpendicular to the base at the mental foramen (Sambhana et al. 2016).

Except SH, all the variable measurements were obtained on the right and left sides, with the assumption that all the radiographs were bilaterally symmetrical as described in Fig. 1. These measurements were collected and tabulated using special codes to ensure patient data confidentiality.

Fig. 1

Linear measurements on panoramic radiograph. A Minimum ramus width (MiRW), B projective height of ramus (PHR), C maximum ramus width (MxRW), D coronoid height (CH), E symphysis height (SH), and F mandibular corpus height (MCH)

To ensure that the size of panoramic radiographs used in this study was the same, standardization of measurement was done by using the radiographs produced from the same X-ray unit. Radiographs selection was also done by strictly following the inclusion and exclusion criteria so that the radiographs used as samples are guaranteed not carried out distortion. In addition, radiograph measurements used the same software and computer unit by all measurers. The measurement tools in the software had been calibrated so that the measurement results are accurate.

Statistical analysis

The initial step was to assess interobserver reliability by measuring 10% of the sample (20 men and 20 women). Within 2 weeks after the initial measurements were taken, intraobserver reliability was tested by remeasuring 30% of the sample (40 men and 40 women). Kolmogorov–Smirnov normality test was conducted to determine the distribution of the data as a prerequisite for selecting inferential statistical methods. Independent t-test was then performed to assess the intergroup differences in measurements and identify the significantly different variables between men and women, potentially serving as components in the discriminant function analysis equation. Given that the sample distribution prerequisites (normality) were satisfied, a discriminant function analysis equation was then established by utilizing the variables that significantly differed between men and women. The obtained equation was then tested for accuracy in estimating the sex of the same sample. The results were expressed in percentage. Afterward, the sex estimation equation was tested on a new sample, and the estimated sex was compared with the actual sex. The percentage of correct sex estimation results was then calculated.

Results

The reliability test was conducted using the intraclass correlation coefficient (ICC) two-way random effects model for consistency, and the results are listed in Table 1. Meanwhile, the results of the normality test are presented in Table 2.

Table 1 ICC score for intraobserver and interobserver tests

Table 2 Normality test with Kolmogorov–Smirnov

According to Table 2, the normality test results indicate that all variables have a significance level (p > 0.05), suggesting that the data are normally distributed. Further analysis was then performed using parametric methods. The outcomes of the independent t-test to determine the differences in measurements between men and women are outlined in Table 3.

Table 3 Results of independent t-test among all linear measurements

As shown in Table 3, the measurement results are significantly different (p < 0.05) between men and women. The mean ± SD of all linear measurements is greater in men than in women.

Based on the results in Table 3, discriminant function analysis equations were established by utilizing the variables that exhibited significant differences between men and women. The assumption was that the sample distribution prerequisites (normality test) are satisfied. Ten equations were developed as listed in Tables 4, 5, 6, 7, 8, 9, 10, 11, 12, and 13.

Table 4 Discriminant function analysis by using 11 linear measurements

Table 5 Discriminant function analysis by using linear measurements after stepwise method

Table 6 Discriminant function analysis by using linear measurements in the right mandibular side

Table 7 Discriminant function analysis by using linear measurements in the left mandibular side

Table 8 Discriminant function analysis by using MiRW

Table 9 Discriminant function analysis by using PHR

Table 10 Discriminant function analysis by using MxRW

Table 11 Discriminant function analysis by using CH

Table 12 Discriminant function analysis by using SH

Table 13 Discriminant function analysis by using MCH

The 10 equations obtained from some combinations of variables were then tested for accuracy in estimating the sex of the training sample. The results were expressed in percentages as shown in Table 14.

Table 14 Percentage of accuracy for the discriminant functions with the training sample

The sex estimation equation produced by discriminant function analysis was then tested on a new sample, and the estimated sex was compared with the actual sex. The percentage of correct sex estimation results with testing sample is shown in Table 15.

Table 15 Percentage of accuracy for the discriminant functions with the testing sample

Discussion

The methods for identifying bodies found in a disaster can be divided into comparative and reconstructive. The former requires the availability of complete antemortem data, which can be used for comparison with postmortem data. The latter focuses on deducing the identity of the body based on available postmortem data (Sassouni, 1963 sit. Sairam et al. 2016). The reconstructive method is useful for identifying fatalities in large-scale open disasters that occur over a large area, often in the absence of adequate antemortem data (Interpol 2018).

Sex estimation is the first step in body identification because other estimates such as age, specific population, and stature rely on the sex estimation results (Mello-Gentil and Souza-Mello 2022). Sex can be identified using sexually dimorphic bones. The mandible is a sexually dimorphic bone whose sex estimation accuracy is second only to the pelvis (Brogdon 2011). Mandibular characteristics indicate sexual dimorphism that is related to growth duration, developmental phases, and muscle activity, which differ in men and women. The mandible is also often used in identifying the bodies of disaster victims because its strength and resistance to pressure allow it to survive in extreme conditions (Toneva et al. 2021).

The mandible can be used to estimate sex either through morphological methods to visually inspect mandibular characteristics or morphometric methods using quantitative linear measurements (Nagare et al. 2018; Verma et al. 2020). Sex estimation using morphological characteristics is subjective, and many anomalies can be misinterpreted by inexperienced examiners. Other reliable methods for estimating sex are morphometric measurements and statistical analysis, which have been widely used on human skeletal elements (Wankhede et al. 2015). The morphometric method is preferred over qualitative sex estimation because of its high level of objectivity, accuracy, and reproducibility. Errors in intraobserver and interobserver tests for the morphometric method are also low (Bertsatos et al. 2019). Dental radiograph can be used as a morphometric tool for sex estimation using linear measurements as shown in Fig. 1. Dental radiographs, especially panoramic radiographs, are a reliable method for identifying corpses. They can be used as a primary identifier because they provide photographic records of the teeth, jaw, and surrounding tissues, which are body parts that are highly resistant to decomposition and fragmentation during a disaster (Interpol 2018; Mello-Gentil and Souza-Mello 2022). Comparative methods for antemortem and postmortem dental radiographs can provide certainty of identity at the individual level (Kumar et al. 2015; Viner and Robson 2017). Dental radiographs are also one of the postmortem data that can be employed in reconstructive methods to identify bodies in an open mass disaster (Carabott et al. 2014).

The results of the reliability test are shown in Table 1. ICC values > 0.9 indicate very good reliability, and ICC values > 0.7 indicate good reliability. These findings reveal that the measurements are reliable and have a high level of reproducibility (Koo and Li 2016). The use of software in this research also influences the reliability. The same software was employed by Ningtyas et al. (2023), who also reported good reliability. The confidence interval used in this study was 0.05. Table 2 shows the results of the normality test where the significance of all data for men and women is p > 0.05.

Table 3 shows that all variables are significantly different (p < 0.05) between men and women, with the mean value for men being greater than that for women. These results are largely similar to those of previous studies on various populations (Astuti et al. 2022; Fekonja and Čretnik 2022; Indira et al. 2012; Kurniawan et al. 2023; Maloth et al. 2017; Mehta et al. 2020; Rad et al. 2020; Saini et al. 2011; Sairam et al. 2016; Sambhana et al. 2016). The greater value in men is due to the gender differences in the vertical growth of the mandible: the growth and development in men are approximately 2 years longer than that in women (Proffit et al. 2019).

All the variables have been proven to differ significantly between men and women and therefore can be used in establishing sex estimation equations. Discriminant function analysis was employed for creating equations because the data distribution prerequisite has been met. This method has been widely used to estimate sex in certain populations with good accuracy, making it useful in forensic cases (Bertsatos et al. 2019; Sambhana et al. 2016).

Discriminant function analysis produced several equation models as shown in Tables 4, 5, 6, 7, 8, 9, 10, 11, 12, and 13, including the equations with all the variables (Table 4), equations with all the variables initially used and then eliminated using stepwise analysis (Table 5), equations with five variables on the right side (Table 6), equations with five variables on the left side (Table 7), and equations for each type of linear measurement that combines both sides (Tables 8, 9, 10, 11, 12, 13). As shown in Table 14, various accuracies are obtained in the range of 63.6–94.4%. A high percentage of over 90% is achieved for the equations that use all the variables. The equation using five measurement variables obtained for each side has better accuracy than the equation using one variable on both sides. The accuracy of the equation can increase if two or more variables are used cumulatively for sex estimation (Mehta et al. 2020). All the equations in Table 14 have achieved accuracies above 80%, except for the equations using MiRW and MxRW with percentages of 63.6% and 65.1%, respectively. A sex estimation equation is acceptable and considered meaningful if it produces an accuracy of > 80% (Ekizoglu et al. 2021; Mello-Gentil and Souza-Mello 2022).

Previous studies have combined one or more of the variables in this research to form sex estimation equations through discriminant function analysis. Maloth et al. (2017) used five variables on the mandibular ramus, including PHR and MiRW, and achieved an accuracy of 74% in an Indian population. Sairam et al. (2016) conducted research with an Indian population using the parameters on the mandibular ramus to form equations and achieved an accuracy of 77–79.5%. Fekonja and Čretnik (2022) showed that MSH and MCH were significantly different between the gender. Sambhana et al. (2016) also obtained a similar accuracy value for MCH at 64%. Mehta et al. (2020) used discriminant function analysis for six mandibular linear measurements, namely, MxRW, MiRW, PRH, condylar height, and gonial angle, and obtained an overall accuracy of 75% for each variable. MxRW and MiRW have the highest accuracies in estimating sex (77.2% and 76.7%, respectively). Indira et al. (2012) conducted similar research by utilizing five linear measurements of the mandible, namely, MxRW, MiRW, CH, PRH, and condylar height, resulting in an equation accuracy of 76%. Maloth et al. (2017) reported a 74% accuracy in estimating sex for the linear measurements of the mandible, including MxRW and CH.

The equation for estimating sex is specific to certain populations (Toneva et al. 2021). An equation that gives accurate results in one population may give inaccurate results for another population. Differences in the accuracy of equations could be due to the variations in sexual dimorphism patterns in skull bones among different populations. These variations in dimorphism patterns can be reflected in the percentage of male and female dimorphism in the same population in bone thickness between populations (Techataweewan et al. 2021). The level of accuracy can differ between one population and another due to variations in the skeletal morphology of people in different geographical conditions. Thus, the mathematical equation for estimating the sex of a population is specific and cannot be applied for other populations because it can cause inaccurate results (Astuti et al. 2022). Therefore, this study used six variables, namely, MiRW, PHR, MxRW, CH, SH, and MCH, which have been proven to be dimorphic, to form a sex-specific estimation equation for the Indonesian population. The findings contribute to the forensic identification of victims in the Indonesian region.

According to Mello-Gentil and Souza-Mello (2022), the accuracy that is considered acceptable and valuable for sex estimation equations is > 80.0%. Table 14 shows that when the proposed equation was used to estimate the same sample used in forming the equation itself (training sample), the accuracy exceeds this minimum value. Differences in accuracy can be caused by several factors, namely the type of linear measurement of the mandible, variations in characteristics within a population, and number of parameters used (Astuti et al. 2022).

As shown in Table 14, the percentage of similarities in the estimates for men is lower than for women. These differences indicate that the variance in parameter sizes in men is higher than that in women (Kongkasuriyachai et al. 2022). The percentage of test results that is generalized to the population is the average percentage of estimated results for men and women. When applied to populations, the equation is expected to provide confirmation and information about the sex of individuals that is previously unknown (Qaq et al. 2019).

The formation of sex estimation equations always ends with validation and generalization to other populations. Equations formed using discriminant function analysis tend to provide a low accuracy when applied to new samples. Discriminant function analysis uses the principle of discriminant comparison of the cut-off point in separating men or women so that the resulting sex estimation appears to be dichotomous (separating two opposing groups) (Bartholdy et al. 2020). This process is important because even in the same population, the percentage of sexual dimorphism patterns can differ between men and women and consequently affect the validity of the sex estimation equation (Techataweewan et al. 2021). Ningtyas et al. (2023) proved that even for a testing sample with the same criteria and population as the training sample, the percentage of correct results from the gender estimation equation tends to decrease. As shown in Table 15, the percentage of similarities in providing correct results differ for male and female gender estimates. The percentage accuracy of the equation in the testing samples tends to be lower than in the training samples.

The limitations of this study are as follows. First, the following factors cannot be controlled because the radiograph samples were taken retrospectively: nutritional, socioeconomic, and hormonal (endocrine growth factors). These factors influence bone development and dimorphism (Indira et al. 2012; Mehta et al. 2020; Sambhana et al. 2016). In addition, the limited number of samples and the age distribution of the patients are less varied. Further study with a wide age range and a large sample size is needed.

Conclusions

The accuracy of sex estimation tends to be high when using all the linear measurements on panoramic radiographs, with accuracies of 93.8% for the training sample and 73.5% for the testing sample. However, if measurements cannot be made on both contralateral sides in cases where a panoramic radiograph is used to image an incomplete mandible, then sex estimation can still be carried out using the measurements on one side with accuracies of approximately 88.7–89.7% for the training sample and 76.5–80.0% for the testing sample. Several mandibular linear measurements can be used for sex estimation, but their accuracy for the training sample is lower than that of sex estimation using several linear measurements on panoramic radiographs. This research was carried out using only panoramic radiographs with good quality. Therefore, the position of the patient’s head is an important factor that must be considered to obtain linear measurements and accurate sex estimation results using panoramic radiographs.