Gender Determination Using Digital Lateral Cephalograms: A Discriminant Function Analysis

Introduction

Several studies demonstrated that human remains identification is one of the most crucial aims in the forensic medicine and fingerprint, dental, anthropological, genetic or radiological examination, cheiloscopy, and so forth are used to construct the human biological profile (1-8). According to a study in the field, as the soft tissues of the corps disappear, the human remains could be identified only by skeletal examination (5). Thus, the most important factor helping to determine the gender of the remains is the presence of prominences in bones (9). Two techniques to help human identification has been described by using data assembled by skeletal examination; the first method, which is not completely verified by scientists, is the use of morphological and visual techniques and the second method is the measurement techniques. However, the more accurate results would be obtained when the two methods are combined (10, 11). Based on other studies, metric examinations are quantitative and current researchers are using this method more and more with direct measurements and biostatic advances (12, 13). However, several factors could provide dissimilar results in different populations; for instance, different environmental factors in different geographical areas and bio-differences (2, 9). Therefore, skeletal sex studies should be conducted in different ethnic and racial populations (14).

In addition, talus, humerus, femur, scapula, and methacarpal bones can be utilized in anthropology and forensic identification  (15). So far, studies of pelvic bone have been widely used for gender indentification. However, weakness and fragility, complex shape, and detachment of its components makes its difficult to find an intact pelvic bone; consequently, using the skull would be more practical in this respect (12, 16-18).

Also, some studies have demonstrated that teeth, especially mandibular canines, due to  strong anatomical structures,  can be very helpful in forensics (19-22)

Studies on the sex determination by scalp or mandibular bones

It is widely accepted that the mandibular bone is one of the strongest and most diverse bones after the skull and pelvic bone ; therefore, it could be very helpful in identifying human remains (2). However, most previous studies have focused on the cranial features and a few have used craniomandibular parameters (23).

Using the skull radiographs

According to some studies, radiographic identification is a great tool in this area. Therefore, pre-mortem radiographs taken clinically and post-mortem radiographs are required for identification(1-12, 24). Skull radiographs can be helpful in differentiating gender in 80 to 100% of cases  (2, 23). As a result, radiography is a simple and cost-effective way to determine sex in comparison to other methods such as histology and biochemistry (25). In addition, lateral and anterior posterior cephalometric radiographs of the skull are very helpful due to the large number of points for comparison (2, 23).

Poongodi et al. and Bhardwaj et al. studied sex determination and achieved good results using panoramic images (25, 26).  Moreover, Uthman AT et al. determined the sex of unknown corpses using spiral CT images of the frontal sinuses and other skull measurements (1).

Studies done by lateral cephalograms and posterolateral radiographs

Naikmasur VG et al. conducted a study on the Tibetan immigrants and the South Indian population. They concluded that among the 11 selected cephalometric landmarks, Bizygomatic width, Ramus height, and face depth were the most important factors in sexual dimorphism (2). Moreover, Patil KR et al.'s study aimed to obtain a discriminant function for sex determination and estimation of the height and size of individuals using lateral cranial cephalometric radiographs in a homogeneous population in India. The researchers found that the discriminant function obtained from the 10 cephalometric variables correctly led to the sex determination in 99% of cases (23).

In another study, Naikmasur et al. (2010) studied lateral cephalometric and PA images and concluded that among the 11 selected cephalometric landmarks, Bizygomatic width, Ramus height, and facial depth had the most important role in sexual dimorphism (2). Finally, Patil and Mody investigated 150 lateral cephalograms and concluded that differential analysis using 10 cephalometric variables had 99% validity in sex determination (23).

Materials and Methods

This study evaluated digital lateral cephalograms of patients referred to one of the private maxillofacial Radiology Clinics (Mashhad, Iran) for their therapeutic purposes. Therefore, a total of 202 lateral cephalograms was selected for the study. 101 radiographic samples were taken from females and 101 from males aged between 18 to 50 years. Since the purpose of this study was to determine gender in forensic medicine, and in most cases it is not possible to estimate the age of a skeleton or body that does not have a recognizable gender age grouping was not performed.

Radiographs with clear cephalometric landmarks and the head in the NHP status when taking the image were selected for the study. Exclusion criteria were any visible skeletal abnormalities in the radiographic images. These patients also had no symptoms such as skeletal abnormalities, previous orthognathic surgery, and trauma in their lateral cephalometric radiographs. Finally, samples with bilateral frontal sinus aplasia were excluded.

It should be mentioned that this study did not consider skeletal classification because the indices associated with mandible and maxilla were used separately. Therefore, all patients were participated in the study. All digital lateral cephalograms were made by Planmeca EC 2003 in one clinic.

Several craniomandibular variables were measured as suggested by Iscan and Steyn (27). This objective procedure was performed by a student under the supervision of an orthodontist and a radiologist with onyxceph®2.6 software that is a dental imaging software with a precision of one-tenth of a mm. OnyxCeph is an image processing software that examines the visual purposes of VTO treatment (visualized treatment objective) (Figure 1). There are some pre-designed analyses in this software, which are practical especially in orthodontics and surgery. Since our study required analysis other than those available in the software, it was necessary to prepare the analysis specific to our study. There is a scaled ruler in the cephalogram image that is used for calibration. At first we entered 10 mm on the cephalogram ruler and then we selected 10 mm in the calibration option in the software menu.

Linear indices in millimeters (mm) and angles in degrees (°) were shown at the right bottom of the screen. Variable measurements were performed by a calibrated operator, and the values were entered in the Microsoft Excel after completing the measurement of all 202 cephalometers. The age and sex of radiographs were clear.

Therefore, 14 cephalometric points were identified on the radiographs and consequently 11 longitudinal and 1 angular indices were obtained (Tables I) and (Tables II). Finally, the line between the two images was taken into account in cases where the binary image was created. (Figure 2)

PASW18 statistical software was used to perform data discrimination, mean comparison of linear indices, and discriminant function analysis. The significance level was also considered to be 5%.

Table I: Landmarks and their definitions

Table II: Linear and Longitudinal Variables and Their Definitions

Figure 1: Overview of Onyx Ceph software.

Figure 2: Craniomandibular variables measured in this study.

 Results

The mean and standard deviation of the patients' age was 24.03 ±5.20 and the minimum and maximum age was 18 and 50 years, respectively. (Table III)

The mean and standard deviation of the patients' age was 24.03 ±5.20 (were 22.98±4.12 for males and 24.68± 5.70 for females ) and the minimum and maximum age was 18 and 50 years, respectively. (Table III)

Table IV lists the results of tracing based on the gender, which included 11 linear variables and one angular variable. Moreover, independent t-test was used to compare the presence or absence of significant differences between the mean of the variables in each group (p <0.001) (Table IV).

According to the Table IV, it could be concluded that the mean of all studied variables except for SNAr was higher in the male than female and Cond-Gn and N-Me were the most determinant linear variables.

Differential analysis equations by incorporating all independent variables into the model

Differential analysis equation

According to Table V, coefficients and constant values of the differential analysis equation were used to obtain the discriminant score.

The equation is as follows.

Table V: Coefficients and constant values of the differential analysis equation to obtain the discriminant score.

D= -19.626+0.07(N-Me)+0.106(Ar-Go)+0.066(Me-Go)+0.161(FsWd)

If the value obtained from the equation was greater than zero , it was more likely to be male and if the value was lower than zero it was more likely to be female.Also, the closer the value was to +1.14, the more likely it was to be male and the closer the value was to -1.14, the more likely it was to be female.

The other concept in differential analysis is centroid. Centroids are the mean discriminant values for each group. The centroid table is used to base the cutting point for classifying the samples. If the two groups have equal numbers, the best sectioning point is the mean of the two centroid values. (Table VI)

According to the above table, the sectioning point value is zero. This means that if the discriminant score of a person is positive, that person is likely to be male, and if negative, she is more likely to be female. In 85.1% of cases the male gender was detected correctly and in 89.1% of cases the female gender was correctly identified. (Table VII)

Table III: Descriptive findings of the age of the patients by sex

Table IV: Descriptive statistical findings and t-value of cephalometric variables by sex.

Table V: Coefficients and constant values of the differential analysis equation to obtain the discriminant score.

Unstandardized coefficients

Table VI: Centroid values associated with either sex

Unstandardized canonical discriminant functions evaluated at group means

Table VII: Predictive findings of the group membership based on discriminant analysis.