Sex Determination From The Sterrnum- A Computed Tomography Based Forensic Approach

2.1 Sex determination

Sex determination is a primary step in forensic identification, as it greatly narrows the range of possible identities and plays a major role in subsequent biological profiling in the field of forensic anthropology, and helps in medico-legal investigations. Edwards et al.,¹¹ and Zech et al.,¹² pointed out the difficulty associated with sex determination in mass disasters, explosions, fire accidents, mutilation, and in cases of high degrees of decomposition. Although DNA examination is one of the most precise methods for determining sex, Battan et al.⁶ emphasized that “degraded samples, contamination, cost, and technical limitations often limit the use of DNA analysis”; in such situations, forensic anthropology is an important field that can be used to determine sex from skeletons.

2.2 Sex determination from bones

Traditionally, the pelvic bones and the skull are the most reliable, as these bones show the highest degree of sexual dimorphism, as noted by Franklin et al.² and Zech et al.¹² A study on the South Indian population conducted by Nagesh et al.¹³ showed that the acetabulum-pubis index accurately identified 83% males and 81% females. This study also highlighted that skeletal indices vary among populations and cannot be generalized. Likewise, a study conducted in the Greek population showed that sex classification varied from 79.7% to 95.4% depending on the index employed.¹⁴ In a Turkish population, Gülhan¹⁵ showed that 3D pelvic indices obtained from CT scans are as reliable as conventional osteometry. A CT-based study conducted on the North Indian population showed strong sexual dimorphism in the coxal, ischiopubic, and sacral indices.¹⁶ Though the pelvic gridle is considered to be the most dependable skeletal part for sex determination because of its strong sexual dimorphism, its practicality in forensic studies is often limited because the pelvis is often broken, incomplete, or absent in cases of trauma, mass disasters, and advanced decomposition.¹⁷ Thus, the assessment of other skeletal remains is required to ensure accurate sex determination in the absence of a pelvis.

Skull- the commonly used skeletal remains for sex determination, followed by the pelvis. A skull-based study conducted in the South Indian population found that biauricular breadth was the most reliable parameter among the 26 chosen parameters, with an accuracy of 85.7%.¹⁸ In another study on the isolated mandible conducted by Sharma et al in the Indian population showed that even though diagonal length & minimum ramus breadth were statistically significant, overall sex determination was only 60%.¹⁹ In a study on the Turkish population, using 3D CT, Dereli et al.²⁰ stated that the glabella was found to be the most influential trait with 92% accuracy, but the female skull showed lower classification accuracy compared with the male skull. Likewise, Zaafrane et al.²¹ established craniometric standards using CT scans in a modern Tunisian population, achieving up to 90% accuracy with multivariate logistic regression models, underscoring the importance of population-specific equations. However, despite the enhanced accuracy and reproducibility associated with the use of CT-based methods, sex determination from the skull is limited by its dependence on multiple landmarks and statistical analyses, which make it less accurate than the pelvis, especially when it comes to fragmented specimens.^20,21

Rahmani et al.²² measured the patella dimensions by MRI in an Iranian population and reached a maximum accuracy of 85.7% using transverse and craniocaudal measurements. Liu et al.²³ assessed sphenoid sinus morphometry by MRI in Chinese adults and found an accuracy of 63.3%, indicating fair discriminatory ability and underlining the ability of MRI to image deep structures without superimposition of bones. More recently, Cavlak et al.²⁴ used convolutional neural networks on MRI slices of the patella and achieved a maximum accuracy of 88.88%, reflecting the increasing use of artificial intelligence in forensic anthropology.

While MRI has the advantage of not using radiation and the ability to study living individuals, some drawbacks are mentioned: reliance on the machine settings, high cost, soft tissue artifacts, lack of availability in forensic practice, and only moderate accuracy in certain areas.^23,24

In a metric study of the long bones in a contemporary Japanese population, Sakaue²⁵ analyzed 47 variables for the humerus, ulna, radius, femur, and tibia using discriminant function analysis and obtained sex discrimination accuracies of 91–95%. The analysis also showed that epiphyseal breadth measurements, especially the distal articular width of the humerus and the proximal epiphyseal breadth of the tibia, were more effective than bone lengths. The presence of sexual dimorphism was observed to be more evident on joint surfaces, whereas measurements of shafts and lengths were relatively less reliable. However, the author stressed the importance of population specificity, and different regressions must be used on different populations. Six femoral measurements were assessed by Mall et al.²⁶ in a modern German population, with classification accuracy of up to 91.7% for a combination of midshaft diameter and head circumference, with transverse head diameter being the single best predictor. Likewise, Steyn & Işcan²⁷ studied the femoral and tibial dimensions of South African whites and concluded that the distal epiphyseal breadths were the most sexually dimorphic features, with accuracies of 86–91%.

Studies have shown consistently that the use of breadth and epiphyseal measurements is a better discriminant than the use of length alone. However, it was emphasized by all authors that discriminant functions cannot be generalized over different ethnic groups and that population specificity is required.²⁶ Amer et al.²⁸ analyzed Egyptian subjects by combining the parameters from MRI of hand and foot bones and obtained a cross-validated accuracy of 90.2% by multivariate discriminant analysis, indicating that the combination of parameters improves the accuracy. Although long bones offer a high degree of accuracy when complete, their recovery may be restricted in the case of fragmented specimens or epiphyseal damage.²⁵

2.3 Sex determination from the sternum

However, to overcome the limitations of the traditional skeletal components, there have been several studies on flat and irregular bones, like the scapula,^3,29,30 clavicle,³⁰ vertebrae,³¹ ribs,³² and sternum.⁴ These bones are often retained even in adverse forensic situations. Of these, the sternum, an isolated bone, is one of the important bones in the determination of sex because of its robust nature and centrality to the thoracic cage.³³

Another aspect of the sternum is that, compared to other bones, it is more likely to be available even when the bones are fragmented.³⁴ The presence of sexual dimorphism in the sternum was proven by Uysal Ramadan et al.⁴ who established that the sternum is effective in determining sex even in the absence of other bones. The effectiveness of the sternum was further proven by Changani et al.⁹ in the Gujarati population. A study conducted by Koçak et al.³² demonstrated that the superior-inferior height of the sternal end of the rib was the most accurate feature in Turkish, with accuracies of 86.5–88.6%. A study established that there are sex differences in the sternum, reflecting its efficacy in providing the necessary information for forensic analysis.⁹

As forensic radiology is increasingly used in forensic science, computed tomography has been recognized as a non-invasive, precise, and repeatable technique in the skeletal analysis of individuals.²⁹ CT-based sternal analysis has shown high accuracy in sex determination, with discriminant function models showing classification accuracy rates above 80% and as high as 95.5% when multiple variables are combined, thus emphasizing its ruggedness for forensic applications.³⁵ Zhang et al.¹ have confirmed that it is a reliable technique for sex determination without the need to physically touch the bone. Rodríguez et al.⁵ have highlighted the importance of avoiding observer bias by utilising computed tomography in forensic anthropology in their study. Ali et al.³⁶ studied an Egyptian population and demonstrated the importance of MDCT in estimating sternal parameters, which are essential for logistic regression-based sex-estimation models.

While MRI provides high contrast for soft tissues, it is technically demanding for bone morphometry due to the low proton density and poor signal of cortical bone, thus making it less ideal than CT for detailed analysis of bony morphology.³⁷ Moreover, CT scanning enables accurate reconstruction of complex bone geometry that aids in landmark identification and measurement, which is important in sex determination, where minute dimensional differences are employed.

Thus, when compared with the conventional morphological, manual osteometry, and MRI-based methods, the CT scan method stands out as a better technique for sex determination because of its non-destructive, three-dimensional, and reproducible nature, and the ability to be applied to both dead and living individuals, hence justifying its choice as the chief methodological approach in this study.

2.4 Sternal parameters across populations

Different studies have used various sternal parameters for sex determination. A study by Zhang et al.¹ in the western Chinese population measured the lengths of the manubrium, the width of the manubrium, the sternal body length, the combined sternum length, and the sternal index. In their study, significant sexual dimorphism existed in all the measured lengths, and a classification accuracy of about 80% using multivariate CT-based analysis. Combined sternum length and dimensions of corpus sterni were used by Franklin et al.² in the Australian population, resulting in an accuracy above 80%.

Earlier, in the Turkish population, Uysal Ramadan et al.⁴ found that the parameters of mesosternal length and combined length were highly reliable. Another study conducted in the Turkish population by Koşar et al.⁸ found that the combined length and total length of the sternum alone were sufficient to predict sex with close to 85% accuracy. More parameters were introduced by Ghorbanlou et al.,³⁸ such as the sternum area, which were found to have an accuracy rate above 90% in the Iranian population.

A study conducted on the African population by Banyeh et al.⁷ found that most linear parameters of the sternum were significantly larger in males, whereas the sternal index was smaller for males. However, it was emphasized that the model results should be generalized with caution, considering the variability among populations.

Indian studies have considerably enhanced regional forensic standards, although the sample population remains restricted. On one hand, a study by Changani et al.,⁹ undertaken among the Gujarati population, found that the best predictor for sex determination was the sum of the lengths of the manubrium and sternum, while the sternal index showed limited predictive value. Meanwhile, the sexual dimorphism of the sternum was again re-established by Chowdhuri et al.³⁹ among the Eastern Indian population, which revealed moderate accuracy in determining sex based on sternal morphometric parameters. More recently, Sravan et al.¹⁰ established that the highest predictive potential was offered by the measurement of the mesosternal length among the Central Indian population. For all these populations, it has, however, been evident that the sternal index has low forensic accuracy.

Studies conducted by Changani et al.⁹ among the Gujarat population, Ali et al.³⁶ among the Egyptian population, and Banyeh et al.⁷ among the Ghanaian population have concluded that females exhibit higher sternal index values. In contrast, all other parameters were consistently higher in males, which led to a considerable overlap between the sternal index value for both males and females, making it an unreliable parameter for sex determination.

From the existing literature, it is evident that morphometric standards differ greatly from population to population. Zhang et al.,¹ Franklin et al.,² Rodríguez et al.,⁵ Banyeh et al.,⁷ Koşar et al.,⁸ and Ghorbanlou et al.³⁸ stress this important fact: do not apply these equations to other populations, as genetics, environment, nutritional, has great impact on the shape of the human skeleton. This fact is particularly relevant given that the country of our interest, India, has rich geographical and ethnic diversity.

The parameters chosen in the current study- Manubrium Width (MW), Combined Length (CL), and individual Sternebra widths (S1W to S4W)- have been well-established by prior research for sex determination.^1,9,10 The values for ML, MW, MSL, and CL have constantly exhibited significant dimorphism between the sexes worldwide, as well as within Indian individuals.^1,4,8,10 The inclusion of individual Sternebra widths has been justified by previous investigations indicating their contributory role when combined with other parameters, even though these parameters have been reported to yield lower accuracy.^7,38

4.1 Study design

This retrospective study included 146 sternal CT images (73 males and 73 females) from the Karnataka population, obtained from a medical institution in Karnataka.

4.2 Participants

The research included participants aged 25 to 80 years whose sternal CT images had sufficient image quality for morphometric measurements. Participants below the age of 25 years, above 80 years, CT scans of the sternum with trauma, deformities, or poor image resolution were excluded. Ethical clearance (IEC No. IEC2:483/2025) was obtained from the Kasturba Medical College and Hospital Ethics Committee. As per the Institution Ethics Committee (IEC), it is mandated to get consent from the Medical Superintendent of Kasturba Medical College to access the CT images, and written consent with participant ID was obtained to access the participant’s CT images. CT imaging was obtained using a 128-slice Incisive CT scan machine (Philips Health Care Suzhou Co., Ltd., China). Imaging was obtained with a thickness of 3 to 5 mm, while the reconstructions were made in the coronal orientation using the B-reconstruction filter. Measurement of the sternum was done with InstaRISPACS version 5.0.0.

4.3 Data collection

The following measurements were taken into consideration for morphometry and recorded in millimeters: