Sex determination using image processing reveals elliptical features of the female pelvic inlet

Abstract

The pelvic inlet is one of the critical constraints within the relatively narrow birth canal for the passage of the fetal head. This study tested the hypothesis that the shape of the pelvic inlet differs between females and males, with the expectation that the female pelvic inlet is more elliptical than its male counterpart. Abdomino-pelvic computed tomography images from 600 cases, equally distributed by sex and age groups, were analyzed. Using 3D reconstruction, image processing, ellipse fitting, and a genetic algorithm, we evaluated whether the proposed measures of elliptical similarity in pelvic inlet shape varied by sex. Additionally, discriminant analysis was employed to assess the accuracy of sex determination based on these measures. The results demonstrated significant differences between the two sex groups for all proposed measures. Discriminant analysis achieved a maximum sex determination accuracy of 79.7%. These findings support the hypothesis that the female pelvic inlet has evolved to be more elliptical compared to the male pelvic inlet. If validated by future studies on diverse populations, this hypothesis may gain broader scientific acceptance.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-03120-6.

Introduction

The pelvic inlet, as the entry point to the birth canal, is one of the most significant anatomical constraints faced by the fetus during childbirth. Due to the limited relaxation or separation of pelvic joints during delivery, the pelvic inlet forms a rigid, bone-structured barrier¹. To accommodate the passage of the fetus, particularly its widest part—the skull—the female pelvic inlet exhibits distinct structural adaptations compared to the male pelvic inlet, which does not serve this function. These sex-based differences have intrigued researchers since the eighteenth century and have been the focus of numerous studies^2–8.

The long-standing classification of male and female pelvic inlets based on their sagittal and transverse diameters has been widely accepted and is prominently featured in textbooks on Anatomy⁹, Forensic Medicine¹⁰, Obstetrics¹¹, and Forensic Anthropology¹². According to this classification, the male pelvic inlet is described as narrower and heart-shaped (the Android type), whereas the female pelvic inlet is wider and oval-shaped (the Gynecoid type). However, a recent study has challenged the accuracy of this classical description². In our previous research, which served as a preliminary study for the current investigation, these findings were corroborated. Nonetheless, this does not imply that the pelvic inlet lacks sexually dimorphic features; on the contrary, the data from our prior study supports the presence of such differences¹.

The present study was conducted to test a main hypothesis: the female pelvic inlet has evolved to adopt a shape as elliptical as possible to facilitate the passage of the fetal head, which is widely accepted to approximate an elliptical form^13–15. To evaluate this main hypothesis, four subordinate hypotheses were formulated:

1. An ideal ellipse that most closely matches the pelvic inlet will be identified, and the degree of alignment between the pelvic inlet and this ideal ellipse will be scored. It is hypothesized that female pelvic inlets will achieve higher ellipse fitting scores compared to male pelvic inlets.

2. The largest possible ellipse that can fit within the pelvic inlet will be identified using a genetic algorithm approach, and the area of this maximum ellipse will be calculated. It is hypothesized that the area of this maximum ellipse and its ratio to the pelvic inlet area will be greater in females than in males.

3. The smallest possible ellipse encompassing the pelvic inlet will be identified, again using a genetic algorithm approach. It is theorized that the excess area within this minimum enclosing ellipse but outside the pelvic inlet will be smaller in pelvic inlets that closely resemble an ellipse and larger in those that do not. Consequently, it is hypothesized that in female pelvic inlets, which are presumed to more closely resemble an ellipse, the excess area and its ratio to the pelvic inlet area will be lower compared to male cases.

4. An ellipse will be constructed based on the anterior half of the pelvic inlet, which more closely resembles an ellipse. It is hypothesized that the excess area within this constructed ellipse but outside the posterior half of the pelvic inlet will be smaller in females, whose pelvic inlets are hypothesized to more closely resemble an ellipse. Additionally, it is hypothesized that the distance between the most posterior point of this ellipse and the promontory will also be smaller in females.

To test these hypotheses, this study analyzed Abdomino-Pelvic CT images from 600 adult Turkish cases aged 21 and older, comprising 300 females and 300 males, equally distributed by age. Confirmation of these hypotheses would provide significant evidence supporting the notion that the female pelvic inlet tends to adopt a more elliptical shape. If validated, these findings would substantially contribute to the literature in fields such as evolutionary biology, anatomy, forensic anthropology, and obstetrics.

Material and methods

This study utilized Abdomino-Pelvic CT images from 600 living individuals with known sex and age. The sample consists of patients who presented to the Radiology Department of the Dursun Odabaş Medical Center at Van Yuzuncu Yil University for various reasons between January 1, 2020, and January 1, 2022.

Radiological images

Abdomino-Pelvic Computed Tomography (CT) examinations were performed using a 16-slice Multislice Computed Tomography machine (Somatom Sensation 16; Siemens Medical Solutions, Erlangen, Germany). The CT settings were adjusted as follows: kV/Effective mAs/Rotation time (sec) values of 120/120/0.75; gantry rotation time of 420 ms; physical detector collimation of 16 × 0.6 mm; section thickness of 0.75 mm; final section collimation of 32 × 0.63 mm; feed/rotation of 6 mm; Kernel of U90u; increment of 0.5 mm; and resolution of 512 × 512 pixels. CT images were acquired in the supine position.

The study employed 3D reconstruction and a multiplanar approach, and since the pelvis forms a rigid, unified structure, minor postural variations were not expected to affect the results. The thin section thickness was critical for ensuring measurement precision. Therefore, the slice thickness specified above was used as the maximum threshold, and CT images with thicker slices were excluded from the study.

Case selection

The study aimed to create a sample of adults aged 21 and older, equally distributed by age and sex. Additionally, to ensure that the discriminant analysis could be applied¹⁶ and the results generalized to the entire Turkish population, the sample size was designed to be as large as possible. Randomization was critical for guaranteeing the generalizability of the results.

To achieve this, CT images acquired between January 1, 2020, and January 1, 2022, were listed in reverse chronological order based on their acquisition date. Images meeting the study’s inclusion criteria were sequentially selected in reverse order until the required number of cases for each sex and age group was reached. This approach ensured randomization. As a result, a sample of 600 cases that met these criteria was successfully created. Detailed information about the case selection process is provided in the Supplementary Information.

3D reconstruction and extraction of pelvic inlet boundary coordinates

In this stage, the bony pelvic inlet was isolated and visualized through 3D reconstruction of Abdomino-Pelvic CT images using Radiant DICOM Viewer® software (Fig. 1). The coordinates of the closed contour defining the boundaries of each case’s pelvic inlet were then extracted using ImageJ® software¹⁷ and saved in a scaled format as a CSV file. A detailed explanation of this process is provided in the Supplementary Information.

Fig. 1.

(1) Bone structures obtained through 3D reconstruction from Abdomino-Pelvic Computed Tomography images. (2) Virtual dissection of the bone structures above the lumbo-sacral joint line using a digital knife. (3) Virtual dissection of the bone structures located postero-inferior to the pelvic inlet using a digital knife. (4) Tracing the boundaries of the exposed pelvic inlet with a graphic tablet to obtain boundary coordinates data.

Creating average male and female pelvic inlet images

To support the formulation of hypotheses, an overlay and pixel density mapping technique was applied to visualize the average appearance of male and female pelvic inlets (Fig. 2). This procedure is described in detail in the Supplementary Information.

Fig. 2.

Graphical representation of the average male and female pelvic inlets obtained using the Pixel Density Mapping method. The color gradient from blue to yellow indicates an increasing number of points per unit area. The images provide an idea of how an average female and male pelvic inlet appear.

Calculating the optimum ellipse that fits the pelvic inlet coordinates

At this stage, the best-fitting ellipse for the pelvic inlet boundary coordinates of each case was determined using a least squares ellipse fitting method¹⁸ to test the first subordinate hypothesis. The degree of alignment between the pelvic inlet and the ideal ellipse was then scored using a similar least squares method, with this score defined as the Ellipse Similarity Score (Fig. 3A). Detailed explanations and calculation methods for this procedure are provided in the Supplementary Information.

Fig. 3.

(A) Ideal ellipse shapes and Ellipse Similarity Scores (ESS) obtained through the ellipse fitting method for four different pelvic inlets. The top two images represent pelvic inlet views from two different female cases, where the pelvic inlets have been evaluated as closely resembling an ellipse. The bottom two images belong to two different male cases, where the pelvic inlets appear to resemble an ellipse to a lesser extent. Notably, the ESS value increases as the resemblance to an ellipse decreases. (B) The largest fitted ellipses and Largest Ellipse Ratio (LER) values obtained through the genetic algorithm method for four different pelvic inlets. The top two images depict the pelvic inlet and ellipse overlays from two different female cases, while the bottom two images show the corresponding overlays from two different male cases. Notably, the LER value decreases as the morphologic resemblance to an ellipse diminishes. (C) Smallest enclosing ellipses and Smallest Enclosing Ellipse Ratio (SEER) values obtained through the genetic algorithm method for four different pelvic inlets. The top two images display the pelvic inlet and ellipse overlays from two different female cases, while the bottom two images show the corresponding overlays from two different male cases. It is noteworthy that the SEER value increases as the resemblance to an ellipse decreases. (D) Fitted ellipses, Excess Areas, Pelvic Inlet Recess Depth (PIRD) values, and Excess Area Ratio (EAR) values for four different pelvic inlets. The top two images represent the pelvic inlet and anteriorly fitted ellipse overlays from two different female cases, while the bottom two images show those from two different male cases. Notably, as the resemblance of the pelvic inlet to an ellipse decreases, the PIRD and EAR values tend to increase.

Finding the largest ellipse that can fit inside the pelvic inlet

To test the second subordinate hypothesis, the parameters of the largest ellipse that could fit within each case’s pelvic inlet were determined. Since this is an optimization problem, a genetic algorithm¹⁹ was employed for this purpose. The area of the resultant largest ellipse, expressed as a percentage of the pelvic inlet area, was termed the Largest Ellipse Ratio and recorded for each case (Fig. 3B). Detailed explanations and calculation methods for this section are provided in the Supplementary Information.

Finding the smallest ellipse that encloses the pelvic inlet

A method similar to that used for identifying the largest ellipse was applied to determine the smallest ellipse enclosing the pelvic inlet. For pelvic inlets that more closely resemble an ellipse, the excess area within this enclosing ellipse but outside the pelvic inlet is expected to be smaller. After determining the smallest enclosing ellipse for each case, the area of this ellipse was divided by the pelvic inlet area, multiplied by 100, and defined as the Smallest Enclosing Ellipse Ratio (Fig. 3C). This parameter was hypothesized to serve as an additional measure of how closely a pelvic inlet resembles an ellipse. Detailed explanations and calculation methods for this parameter are provided in the Supplementary Information.

Calculating the excess area ratio and the recess depth ratio

Finally, parameters termed “Excess Area” and “Pelvic Inlet Recess Depth” were calculated using an ellipse formed based on the anterior section of the pelvic inlet. First, the Transverse Diameter (TD) and Antero-Posterior Diameter (APD) were measured. An ellipse was then constructed passing through the two endpoints of the TD line and the anterior-most point of the APD line. The area within the posterior half of this ellipse, but outside the pelvic inlet, was defined as the Excess Area. It was hypothesized that this excess area would be smaller in females, as the posterior pelvic inlet more closely resembles a semi-ellipse.

Additionally, because the TD in males is positioned more posteriorly and the promontory is more anteriorly pronounced, the distance between the most posterior point of this ellipse and the promontory was hypothesized to be longer in males. This distance was defined as the Pelvic Inlet Recess Depth (Fig. 3D). At this stage, the Excess Area Ratio, calculated by dividing the Excess Area by the pelvic inlet area and multiplying by 100, and the Recess Depth Ratio, obtained by dividing the Pelvic Inlet Recess Depth by the APD and multiplying by 100, were computed.

Furthermore, the pelvic inlet area (AREA) was calculated at this stage to evaluate sexual dimorphism. Detailed explanations regarding the Excess Area Ratio and Recess Depth Ratio are provided in the Supplementary Information.

Ellipse Similarity Score, Transverse Diameter, Antero-Posterior Diameter, Pelvic Inlet Recess Depth, and Pelvic Inlet Perimeter were measured and reported in millimeters, while Pelvic Inlet Area was measured and expressed in square millimeters. Largest Ellipse Ratio, Smallest Enclosing Ellipse Ratio, Excess Area Ratio, and Recess Depth Ratio were unitless ratios.

Statistical analysis

Descriptive statistics were calculated for the Ellipse Similarity Score, Largest Ellipse Ratio, Smallest Enclosing Ellipse Ratio, APD, TD, Pelvic Inlet Recess Depth, Excess Area Ratio, Recess Depth Ratio, and AREA parameters. The distribution of pelvic inlets based on classical classification and sex was also analyzed using APD and TD measurements. Additionally, the relationship between these parameters and age was examined, including variations across age groups and correlations with age. A comparison was conducted to determine whether the distribution of these parameters differed significantly between the two sex groups. The degree of sexual dimorphism for each parameter was assessed by calculating the accuracy of sex determination using univariate discriminant analysis.

Finally, the extent to which sex determination could be achieved by combining all parameters was evaluated using discriminant function analysis. Statistical significance was set at 0.05. Statistical analyses were performed using SPSS version 22.0, while ellipse image processing and related calculations were carried out in MATLAB R2023a.

Inter-observer and intra-observer error analysis

In this study, non-automated stages of the analysis that could introduce observer-dependent variability or inconsistency included the extraction of the 3D pelvic inlet image, its isolation, and the manual extraction of the pelvic inlet surface boundary coordinates. To assess Inter-Observer error, the coordinates for each case were independently extracted by two observers (Y.E. and E.K.). For Intra-Observer error, the coordinates of 30 randomly selected cases were re-extracted 15 days after the initial extraction.

Consistency was evaluated by calculating the Relative Technical Error of Measurement (rTEM) and the Intraclass Correlation Coefficient (ICC) for the resulting parameters. An rTEM below 1.5% was considered acceptable for Intra-Observer error, and below 2.0% for Inter-Observer error. An ICC value of 0.75 or higher was interpreted as indicative of consistent results²⁰.

When the Intra-Observer error analysis was conducted, rTEM values were below 1.5% (ranging from 1.105 to 1.422). Similarly, during the Inter-Observer error analysis, the values remained under 2% (ranging from 1.353 to 1.907). Moreover, the calculated ICC values exceeded 0.75 (ranging from 0.810 to 0.921). These results indicate consistency in the findings both among observers and within individual observers, confirming the reliability of the measurements.

Ethical considerations

This study was approved by the Van Yüzüncü Yıl University Non-Interventional Clinical Research Ethics Committee (Decision Number: 2022/04–03; Date: April 15, 2022). It involved the use of Computed Tomography (CT) images from human subjects; however, no personally identifiable information was included in the manuscript.

Given the retrospective nature of the study and the absence of personal data, obtaining informed consent from participants was deemed not applicable and was therefore not required. An informed consent waiver was granted by the Van Yüzüncü Yıl University Non-Interventional Clinical Research Ethics Committee.

All procedures adhered to the ethical standards of the responsible committee on human experimentation and the principles outlined in the Helsinki Declaration. Detailed information regarding all stages of the study was provided during the ethical committee application, with a specific commitment to ensuring the confidentiality of personal data and sharing study data in anonymized format.

Results

When examining the average pelvic inlet density maps for females and males, it can be observed that the average male pelvic inlet has a more angular and narrow structure. Additionally, the widest transverse section of the male pelvis is positioned more posteriorly compared to that of the female pelvic inlet. In contrast, the average female pelvic inlet exhibits more rounded contours and a shape that provides a wider internal space (Fig. 2).

Descriptive statistics for all nine parameters are presented in Table 1. Comparison of the parameter distributions between the two sex groups revealed significant differences for all parameters. The Ellipse Similarity Score, Smallest Enclosing Ellipse Ratio, TD, Pelvic Inlet Recess Depth, Excess Area Ratio, and Recess Depth Ratio parameters showed significant positive correlations with age, whereas the Largest Ellipse Ratio, APD, and AREA parameters demonstrated significant negative correlations with age.

Table 1.

Descriptive statistics of all parameters for the two sex groups.

		MEAN	95%CI	STD. DEVIATION	MIN	MAX	P*
ESS†	Male	3.87	3.78–3.97	0.81	1.56	6.62	0.000**
	Female	3.34	3.24–3.45	0.93	1.10	6.70
LER‡	Male	78.3	77.9–78-8	4.10	66.1	91.1	0.000**
	Female	81.8	81.3–82.4	4.73	68.6	92.6
SEER†	Male	126.9	126.3–127.5	5.37	113.9	149.3	0.000**
	Female	122.4	121.7–123.0	5.76	108.9	145.0
TD†	Male	127.9	126.9–128.9	8.75	106.2	170.9	0.000**
	Female	135.6	134.4–136.3	8.43	113.6	167.8
APD‡	Male	105.9	104.8–107.1	10.2	76.4	135.7	0.000**
	Female	113.9	112.8–115.2	10.6	88.8	140.9
PIRD†	Male	38.9	37.9–39.8	8.2	7.29	70.24	0.000**
	Female	31.5	30.4–32.6	9.7	6.1	77.6
EAR†	Male	34.3	32.9–35.6	11.9	4.9	119.8	0.000**
	Female	24.4	23.3–25.5	9.8	4.7	61.6
RDR†	Male	37.2	36.2–38.2	9.0	6.3	69.7	0.000**
	Female	28.1	27.0–29.2	9.4	4.9	62.7
AREA‡	Male	10,742.5	10,602.4–10,882.6	1233.4	7251.3	17,528.4	0.000**
	Female	12,288.4	12,132.6–12,444.1	1370.9	9306.1	17,654.7
PERI	Male	445.9	440.5–451.4	48.1	342.5	681.3	0.000**
	Female	480.3	470.2–490.4	89.1	384.6	1276.8

^†The relevant parameters exhibit a significant and positive correlation with age.

^‡The relevant parameters exhibit a significant and negative correlation with age.

*The P-value represents the comparison of the distribution of each parameter across sex groups.

**The distributions of the relevant parameter showed significant difference between the two sex groups.

ESS, Ellipse similarity score; LER, Largest ellipse ratio; SEER, Smallest enclosing ellipse ratio; TD, Transverse diameter; APD, Antero-posterior diameter; PIRD, Pelvic inlet recess depth; EAR, Excess area ratio; RDR, Recess depth ratio; AREA, Pelvic inlet area; PERI, Pelvic inlet perimeter.

ESS, TD, APD, PIRD, and PERI were measured in millimeters; AREA in square millimeters; and LER, SEER, EAR, and RDR were unitless ratios.

According to the classical pelvic inlet classification, the most common type of pelvic inlet in both sexes was the android pelvis, observed in 59.7% of males and 58.3% of females. No significant difference was found in the distribution of cases between the two sex groups based on this classification (Pearson Chi-Square: 0.630, P = 0.890) (Table 2).

Table 2.

Comparison of the distribution of cases according to classical pelvic inlet classification in sex groups.

	Android	Anthropoid	Gynecoid	Platypelloid	Total
Male	179 (59.7%)	14 (4.7%)	34 (11.3%)	73 (24.3%)	300
Female	175 (58.3%)	18 (6.0%)	36 (12.0%)	71 (23.7%)	300
Total	354 (59.0%)	32 (5.3%)	70 (11.7%)	144 (24.0%)	600

Pearson Chi-Square: 0.630.

P: 0.890*

*Pelvic inlet classification does not differ significantly between sex groups.

In the univariate discriminant analysis, the parameter with the highest sexual dimorphism and the best sex determination accuracy was AREA, achieving an accuracy of 72.8%. This was followed by the Recess Depth Ratio with 69.8% accuracy and the Excess Area Ratio with 69.3% accuracy. These results indicate that the proposed parameters exhibit a moderate level of sexual dimorphism. The univariate discriminant analysis results for all parameters are summarized in Table 3.

Table 3.

Results of sex determination using univariate discriminant analysis for each parameter.

Estimated sex	Real sex		Accuracy (%)
	Male	Female
Ellipse similarity score (ESS)
Male	183	117	61.0
Female	105	195	65.0
Total			63.0
Largest ellipse ratio (LER)
Male	193	107	64.3
Female	103	197	65.7
Total			65.0
Smallest enclosing ellipse ratio (SEER)
Male	193	107	64.3
Female	93	207	69.0
Total			66.7
Transverse diameter (TD)
Male	202	98	67.3
Female	106	194	64.7
Total			66.0
Antero-posterior diameter (APD)
Male	201	99	67.0
Female	106	194	64.7
Total			65.8
Pelvic inlet recess depth (PIRD)
Male	194	106	64.7
Female	92	208	69.3
Total			67.0
Excess area ratio (EAR)
Male	193	107	64.3
Female	77	223	74.3
Total			69.3
Recess depth ratio (RDR)
Male	207	93	69.0
Female	88	212	70.7
Total			69.8
Pelvic inlet area (AREA)
Male	225	75	75.0
Female	88	212	70.7
Total			72.8

Stepwise Discriminant Analysis was conducted to determine the maximum sex estimation accuracy achievable through the combined use of the proposed parameters. The analysis concluded that the best combination involved AREA, Excess Area Ratio, Pelvic Inlet Recess Depth, APD, Ellipse Similarity Score, and PERI. This collective analysis achieved a sex determination accuracy rate of 79.7% (Table 4).

Table 4.

Results of sex determination using stepwise discriminant analysis.

Classification function
D(x)  = − 0.717 X AREA + 0.546 X EAR + 0.498 X PIRD−0.465 X APD + 0.368 X ESS − 0.290 X PERI
Summary of classification (Leave-one-out cross-validated results)
		Estimated sex			Accuracy
Real sex		Male	Female
Male		247	53	Male	82.3%
Female		69	231	Female	77.0%
				Total	79.7%

AREA, Pelvic inlet area; EAR, Excess area ratio; PIRD, Pelvic inlet recess depth; APD, Antero-posterior diameter; ESS, Ellipse similarity score; PERI, pelvic inlet perimeter.

The Ellipse Similarity Score, Largest Ellipse Ratio, Smallest Enclosing Ellipse Ratio, Pelvic Inlet Recess Depth, Excess Area Ratio, and Recess Depth Ratio parameters, proposed to test the main and subordinate hypotheses, were found to exhibit significant sexual dimorphism and allow for sex determination to a certain extent. These findings support the main hypothesis.

Additionally, significant correlations with age were observed for the variables. An evaluation by separate age groups revealed that sexual dimorphism was present for nearly all variables across most age groups. Based on these findings, it was concluded that the study’s results are not attributable to a Type-1 error caused by age as a confounding factor. Details of this analysis are provided in the Supplementary Information (Supplementary Information-Table).

Power analysis

To test the generalizability of the study’s results, a statistical power calculation formula was used based on the difference in means between two groups for normally distributed continuous variables²¹. The formula was as follows:

where; Φ represents the cumulative distribution function (CDF) of the standard normal distribution; z₁ − α/2 refers to the critical z-value corresponding to a two-tailed test with a significance level of α; ∆ denotes the effect size, which is the absolute difference between the means of the two groups (μ1–μ2); and represent the variances of the two groups, respectively; and n₁ and n₂ indicate the sample sizes of the two groups, which are 300 for both.

When this formula was applied to the variables Ellipse Similarity Score, Largest Ellipse Ratio, Smallest Enclosing Ellipse Ratio, Transverse Diameter, Antero-Posterior Diameter, Pelvic Inlet Recess Depth, Excess Area Ratio, Recess Depth Ratio, Pelvic Inlet Area, and Pelvic Inlet Perimeter, the statistical power for all variables was approximately 1.00. These results indicate that the study’s sample size was sufficient to detect differences between males and females for each analyzed variable.

Additionally, as the study sample was entirely randomly selected, the results are representative of the characteristics of an Eastern Turkish population. The leave-one-out cross-validation conducted during the discriminant function analysis further demonstrated that sexually dimorphic traits persist, albeit with a moderate accuracy rate of 79.7%, highlighting the strength and generalizability of the findings.

Discussion

The classical pelvic inlet classification, traditionally perpetuated in textbooks as fact—suggesting that the female pelvic inlet is predominantly gynecoid and the male inlet predominantly android—has been refuted by previous studies and the findings of the current study. However, this does not imply the absence of shape differences between male and female pelvic inlets. Rather than relying solely on simple assessments based on APD and TD, this study aimed to utilize more advanced and complex analyses.

Our study employed techniques such as 3D reconstruction, image processing, and genetic algorithms, revealing that the female pelvic inlet tends to exhibit a more elliptical shape compared to the male pelvic inlet. Based on the literature reviewed in this study, no prior research has examined the human pelvic inlet using such methods, nor has the elliptical nature of the female pelvic inlet been demonstrated with such an objective approach. In this sense, the findings of this study are considered to be of significant importance.

The human birth canal and pelvic inlet have undergone various evolutionary processes and have acquired their present form. Bipedal locomotion, which evolved approximately 4 to 5 million years ago, led to some major changes in the hominid pelvis²². Subsequently, about 2 million years ago, early Homo experienced a significant increase in brain size²³. This development, coupled with the adaptations of the human pelvis to bipedalism, made the birthing process increasingly difficult as the pelvis struggled to accommodate fetuses with progressively larger skulls. This situation limited the options for resolving these birth-related challenges within the constraints of bipedalism²². While a simple expansion of the human birth canal could have provided a solution to this problem, it is believed that a wider birth canal might have posed a disadvantage in terms of locomotion^22,24,25. Additionally, the increased likelihood of pathologies such as prolapse and incontinence, is thought to have limited the extent of this expansion^22,26,27. Unlike in some species like bats²⁸, the separation of the pubic symphysis during childbirth has likely not evolved in humans, possibly due to similar risks. Consequently, pubic and sacroiliac relaxation during childbirth has remained very limited²⁹.

Another related phenomenon involves male sexual selection. It is widely recognized that certain features of the female body, which signal health, fertility, and consequently the ability to give birth, are perceived as attractive by men. A low waist-to-hip ratio and rounded hips—classically regarded as attractive—may have acquired their appeal because of their association with reproductive capability³⁰. A finding that may be considered supportive of this hypothesis was reported in a study on cephalopelvic disproportion, which can be defined as a mismatch between the size of the fetal head and the maternal pelvis that prevents safe vaginal delivery. In that study, an inter-trochanteric diameter of less than 30 cm in women was identified as one of the significant risk factors for cephalopelvic disproportion³¹. This finding may support the notion that hip features perceived as attractive by men could serve as indicators related to fertility and ease of childbirth.

Because the female pelvis has been shaped and differentiated from the male pelvis through the evolutionary processes detailed above, pelvis has become the most sexually dimorphic part of the human skeleton. For this reason, the pelvic bones have consistently been a focal point of interest in sex determination studies. Consequently, sex determination accuracies achieved from these bones are generally higher compared to other skeletal elements ^32,33. In recent years, a substantial number of studies have been conducted on this subject, reporting accuracies ranging from 60.9 to 100% ^32–52. These studies have analyzed individual regions such as the greater sciatic notch ^48–52, sacrum ^33,47, ilium ^44,45, pubis ⁴⁶, acetabulum ^34,35, hip joint ³⁵, and subpubic angle ⁵², as well as the pelvis or coxa as a whole or in various combinations ^32,36–43.

Compared to the results of these studies and the generally accepted accuracy rate of 95% in forensic anthropology applications, the 79.7% accuracy achieved in our study is not particularly high and can be interpreted as a moderate-level outcome. Among these studies, Steyn and İşcan reported an accuracy of 77.2% using measurements of the pelvic inlet transverse diameter, anteroposterior diameter, and bi-iliac diameter in a multivariate discriminant analysis ³². In another study that performed measurements comparable to those in our study, the univariate sex determination accuracies of traditional pelvic inlet measurements ranged from 54.2 to 68.3% ⁵³. Considering that the univariate sex determination accuracies achieved in our study ranged from 63.0 to 72.8%, it can be stated that the results are comparable and, perhaps, indicate a certain degree of improvement. However, differences between populations should also be taken into account.

Over the last two centuries, the variations in pelvic shape that have intrigued researchers were initially thought to be based on “racial affiliation,” influenced by the prevailing intellectual currents of the time ². For instance, Turner determined the types of pelvic shapes across different “races” in his study ³. However, as this anthropological approach was later recognized as flawed, it was determined that the shape of the pelvic inlet varies depending on numerous factors, including nutrition, physical activity, and body measurements². Initially proposed by Prof. Zaaijer of Leyden and originally used to denote “racial” differences, the brim index and its variations were later applied as a metric for distinguishing sex differences ^2,3. Subsequent studies using various types of brim indices have suggested that the pelvic inlet is typically android in males and gynecoid in females ^2,54,55. This information has been widely accepted as accurate in numerous textbooks across various fields and has been perpetuated over the years ^9–12,56,57.

Although this assertion has been refuted by recent analyses ^1,2, including our study, new dimorphic features of the pelvic inlet were identified in our research. We proposed a hypothesis that had not been previously tested in the literature, and our analyses demonstrated that this hypothesis holds true for the cases in our sample. Specifically, we found that the female pelvic inlet more closely resembles an ellipse compared to the male pelvic inlet. When examining the shape of the male pelvic inlet, particularly through the density map technique (Fig. 2), it can be inferred that the male pelvic inlet is more angular and resembles a "Calf Cut Diamond" shape. However, to assess the generalizability of our findings to other populations, it would be beneficial to replicate this study in different populations. Although beyond the scope of this study, future research could investigate whether the male pelvic inlet aligns more closely with the "Calf Cut Diamond" shape than the female pelvic inlet using similar methods. Additionally, it is anticipated that similar results may be observed if this study is replicated in animals with relatively narrow birth canals compared to the fetus, such as macaques ²².

Although the sex determination findings of our study may not yet reach a level suitable for practical application in forensic anthropology, we believe the novelty of our research lies in its contribution to the literature on the evolution of the female pelvic inlet. Whether the findings of this study show any correlation with sexual attraction could be explored in future research. Investigating the potential relationship between waist-to-hip ratio and the parameters proposed in our study may also offer valuable insights. Additionally, examining whether these parameters about the degree of ellipticity of the female pelvic inlet are associated with ease of childbirth or with cephalopelvic disproportion would serve to further test our hypothesis that the female pelvis has evolved to resemble an ellipse to facilitate parturition. In this way, the evolutionary biology implications of our findings could be more concretely established. If supported by positive results, the parameters we propose may also hold clinical value and potential applications in pelvimetry.

Limitations

The hypotheses in this study are partially based on the assumption that the cross-section of the fetal head is elliptical, which is not entirely accurate. This assumption could be criticized as incorrect, potentially invalidating one of the foundations of the hypotheses. However, even under this criticism, the results of the study are still considered significant. Furthermore, while the elliptical nature of the female pelvis could not be mathematically expressed in previous years due to the lack of integrated computer technologies, subjective observations have long suggested that the female pelvis is more ellipsoid. Although part of a classification deemed ineffective and invalid due to its simplicity, relying solely on APD and TD, the Gynecoid-type pelvic inlet can be interpreted as reflecting the subjective observation that the female pelvic inlet resembles an ellipse.

Another limitation of the study is the applicability of its results in field forensic anthropology practices. As outlined, conducting the complex mathematical analyses required for these evaluations may be challenging in field settings due to limitations in resources, time, and equipment for capturing precise photographs. To address this issue, there are plans to transform the study’s findings into a mobile application. This approach would enable practical estimations by allowing the necessary calculations to be performed quickly using simple, non-scaled photographs captured with a mobile device.

Finally, another challenge is the potential incompleteness of pelvic skeletal remains. In many cases, the pelvis may be fragmented or have missing parts, limiting the applicability of our study’s findings. This issue can be addressed by manually reconstructing the pelvic ring using various assembly methods when all parts are present. In cases where some pelvic bones are missing, reconstruction could be attempted based on the assumption that the pelvic inlet is approximately symmetrical in the left–right direction. In cases where the remains are too fragmented to allow for the above analyses, the partial ellipse fitting method could be employed. For example, if only a portion of the unilateral terminal line is preserved, the corresponding segment of the ellipse can be determined using a partial ellipse fitting method. From this, an Ellipse Similarity Score could also be calculated. However, adapting this study to fragmented remains falls within the scope of a new and separate study.

Conclusion

In conclusion, this study has demonstrated, through image processing on our sample, a previously unknown or mathematically unestablished characteristic of the female pelvic inlet: it resembles an ellipse more closely than the male pelvic inlet. Replicating the proposed method on other populations and confirming the results would provide greater clarity on whether the human female pelvic inlet consistently resembles an ellipse. Additionally, future studies could explore whether the male pelvic inlet aligns with a "diamond shape." We believe this study contributes to literature by enhancing our understanding of the human birth canal and providing a foundation for future research.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: Y.E., Methodology: Y.E., E.E., E.K., U.D.; Formal Analysis: Y.E., M.A.; Investigation: Y.E., E.K., U.D.; Resources: Y.E.; Software: E.E.; Data Curation: Y.E., E.E., M.A., Y.H.; Writing—Original Draft: Y.E., M.A.; Writing—Review & Editing: M.A., Y.H.; Visualization: Y.E., E.E., E.K., U.D.; Supervision: M.A., Y.H., Y.E.; Project Administration: Y.E.

Data availability

All data generated and analyzed during this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.