|Year : 2020 | Volume
| Issue : 4 | Page : 126-134
Stature estimation from body dimensions in Han population of Southern China
Lu-Yao Xu, Fan-Zhang Lei, Jie-Xuan Lin, Ling Song, Xin-Yu Li, Qi Wang
Department of Forensic Pathology, School of Forensic Medicine, Southern Medical University, Guangzhou, China
|Date of Submission||15-Jul-2020|
|Date of Acceptance||08-Dec-2020|
|Date of Web Publication||05-Jan-2021|
Department of Forensic Pathology, School of Forensic Medicine, Southern Medical University, No. 1023, South Shatai Road, Baiyun District, Guangzhou, Guangdong
Source of Support: None, Conflict of Interest: None
Stature estimation is widely used for individual identification in forensic field. Previous studies have proposed several regression equations derived from a single population for this purpose. However, this may not be suitable for other populations because of different hereditary and environmental conditions. In this study, stature estimation equations for southern China Han population have been provided. The study was conducted on a sample population of 121 men and women aged 18–25 years. A total of 19 parameters, including stature, head, torso, and parts of upper limbs and lower limbs, were measured according to standard anthropometric procedures. Herein, the anterior superior spine–malleolus medialis line showed the highest correlation coefficient (r = 0.817) and was the most reliable predictor (R2 = 0.667) in men, while the best predictor for women was total leg length (R2 = 0.746) with the highest correlation coefficient (r = 0.863). The regression analysis results via multiple predictors showed a high accuracy in stature estimation. Moreover, the analysis of multiple regression predictors showed that the dimensions of lower limbs were more reliable for stature estimation compared to head, torso, and upper limb measurements. This study provided equations of stature estimation for southern China Han population which can be useful in cases of dismembered body.
Keywords: Body dimensions, forensic anthropology, regression analysis, southern China Han population, stature estimation
|How to cite this article:|
Xu LY, Lei FZ, Lin JX, Song L, Li XY, Wang Q. Stature estimation from body dimensions in Han population of Southern China. J Forensic Sci Med 2020;6:126-34
|How to cite this URL:|
Xu LY, Lei FZ, Lin JX, Song L, Li XY, Wang Q. Stature estimation from body dimensions in Han population of Southern China. J Forensic Sci Med [serial online] 2020 [cited 2022 Aug 16];6:126-34. Available from: https://www.jfsmonline.com/text.asp?2020/6/4/126/306180
| Introduction|| |
Forensic anthropology is a branch of medical science that applies anthropological theories and methods to solve individual identification problems related to the skeleton in judicial matters. In forensic anthropology, sex, age, and stature determination are the primary tasks for establishing the biological profile of an individual, which possibly leads to positive personal identification.,, In forensics, it is a common practice to divide corpses in order to escape prosecution. Analogously in catastrophe and traffic accidents, the human body is often severely damaged with tissue separation, making it difficult to identify individuals. In addition to the cases of dismembered bodies, catastrophes and accidents, unknown corpses, highly decayed corpses, and other cases, there is a need for quick and accurate confirmation. Consequently, in such circumstances, stature estimation will be very useful in confirming the process of final identification and development of anthropometrical databases.
Besides, previous studies have recommended regression equations for estimating stature among different human populations at different anatomical dimensions. Moreover, most studies have concluded that lower limb offers a better estimate of stature.,, Nevertheless, some studies have demonstrated that the upper extremity dimensions are also related to the stature.,, Several studies have derived regression formulae from the measurements of foot and footprint dimensions.,,,, Furthermore, hand length, hand breadth, and even more detailed measurements used to calculate stature have also been reported. However, with advanced medical imaging, many research methods have been upgraded from anthropometry to more accurate X-ray for bone measurements, hence reducing experimental errors.
Meanwhile, many researchers have realized that a formula derived from one population may not be suitable for other populations as body dimensions show ethnic variation due to hereditary and environmental conditions. Hence, it is essential to create new regression formulae to help estimate the stature of a specific population. The aim of this study is to assess the relationship of human body dimensions to stature and develop new regression equations to obtain a more reliable stature estimation for Han population living in southern China. This is achieved by relating stature and the following other 18 dimensions including, maximum head length (MHL), head circumference (HC), glabella-external occipital protuberance line (GPL), head height (HH), upper limb length (ULL), upper arm length (UAL), forearm length (FAL), shoulder breadth (SB), anterior superior spine–malleolus medialis line, total leg length (TLL), thigh length (TL), shank length (SL), foot length (FL), foot breadth (FB), hand length (HL), back of HL (BHL), hand breadth (HB), and middle finger length (MFL).
| Materials and Methods|| |
The study comprised a sample Han population of 121 students from southern China (59 men and 62 women) aged 18–25 years. Students with any deformity, diseases, injury, fracture, or allergic to evaluating instruments were excluded from the study.
Measurements were taken by five trained operators according to standard procedures to reduce interobserver error. In addition, in order to reduce errors, we used standard anthropometrical instruments for all participants' measurements, and these measurements were taken around the same time in the evening, when most volunteers were available for the experiment. As it was confirmed that there were no statistically bilateral differences,,,,, limb dimensions were taken on the right side for each individual. Dimensions, involving lower limbs and stature, were taken on barefoot. All dimensions including stature were taken twice in centimeters to the nearest millimeter.
In the case of instruments, stature and dimensions related to upper and lower limbs were measured by anthropometric straight angle bar gauge (ZAPM-E, Beijing Dingyong Huatai Technology Co. LTD, China). Hand and foot measurements were recorded by spreading caliper for anthropometry (ZAPM-E, Beijing Dingyong Huatai Technology Co. LTD, China). Dimensions involving head were taken by a tape and spreading caliper for anthropometry (ZAPM-E, Beijing Dingyong Huatai Technology Co. LTD, China).
- Stature (S) is the distance between the vertex and the floor in the anatomical position, with the head at the Frankfurt plane orientation
- MHL is the maximum linear distance between the glabella and the posterior cranial point at the median sagittal plane
- HC is the circumference from the point between the glabella and the posterior occipital point
- GPL is the linear distance between the glabella and the occipital eminence at the median sagittal plane. This indicator is an innovation on MHL
- HH is the projection distance of the line from the overhead point to the submental point at the coronal plane
- ULL is the linear distance from the acromion to the tip of the middle finger
- UAL is the direct distance from the acromion to the radial point
- FAL is the direct distance from the radial point to the styloid point of the radius
- SB is the crow flies distance between the left and right acromion points
- TLL is the height of the anterior superior iliac spine minus the height of the medial malleolus point, multiplied by 96%
- Anterior superior spine–malleolus medialis line (AML) is the direct distance between the anterior superior iliac spine and the medial malleolus. This indicator is an innovation on TLL
- TL is the straight-line distance from the anterior superior iliac spine to the tibial point
- SL is the straight-line distance from the tibial point to the medial malleolus point
- FL is the maximum crow flies distance from the pternion to the acropodian
- FB is the distance between the points of the anterior epiphysis (distal) of the 1st metatarsal, the most prominent inner side of the foot (metatarsal tibiale), and the joint of the 5th metatarsal anterior epiphysis, the most prominent outer side (metatarsal fibulare)
- HL is the projected distance between the interstyloid point and the tip of the third finger,
- BHL is the distance from the midpoint of the line between the radial styloid point and the ulnar styloid point at the back of the hand to the proximal phalanx of the middle finger
- HB is the distance between the most prominent point outside the 2nd metacarpal lower epiphysis (metacarpal radiale) to the most prominent inside point of the 5th metacarpal lower epiphysis (metacarpal ulnare),
- MFL is the direct distance between the tip of the middle finger and the proximal phalanx of the middle finger
All the dimensions are illustrated in Supplementary Material 1 (body dimensions) and Supplementary Material 2 (head dimensions).
To identify significant differences among sexes, the independent t-test was performed. In order to verify whether gender has an impact on the relation between stature and body dimensions with significant correlations with stature, an analysis of covariance was done to test using the general linear model. Karl Pearson's correlation analysis was conducted to determine the association between stature and dimensions as well as the degree of correlation. Meanwhile, correlation coefficient (r), R2, and standard error of estimate (SEE) were calculated. Finally, linear regression, multiple regression, and step-wise regression analysis were completed to determine stature equations from various dimensions. The data obtained were statistically analyzed using IBM SPSS Statistics Version 23.0 for Windows (Armonk, NY: IBM Corp) and GraphPad Prism version 8.0.1 for Windows (San Diego, California USA). P < 0.05 was considered statistically significant.
| Results|| |
The results of descriptive statistical analysis for all the 19 dimensional measurements including stature and the sexual dimorphism t-test outcomes are shown in [Table 1]. Moreover, frequency histograms and Q-Q plots were drawn to show the normality distributions of all the 19 dimensional measurements, which are presented in Supplementary Materials 3-6.
|Table 1: Descriptive statistics: Stature and human body dimensions (cm) in men and women|
Click here to view
Notably, the stature of women showed a relatively concentrated distribution of low level average height (mean = 160.09, standard deviation = 5.34). The mean values of all dimensions in men were virtually higher than those in women. Statistically significant differences were found (P < 0.01) in the independent t-test. At the same time, BHL displayed the smallest difference between men and women (t = 2.882), while FL manifested the largest difference (t = 9.810), as in previous studies. The results of the analysis of covariance confirmed the influence of sex on the relationship between stature and ULL, UAL, AML, SL, and FL (P < 0.05). Meanwhile, the analysis of covariance demonstrated the latent nonsignificant influence of sex on the relationship between stature and TLL, FB, and MFL (P > 0.05). Notably, in Levene's test of equality of error variances, other body dimensions showed significant values which were <0.05, indicating that the variance across groups was not significantly different. Therefore, considering the results of independent t-test and the consistency of experiment, the results of the regression analysis were presented separately by sex. All the above-mentioned results of analysis of covariance and Levene's test of equality of error variances are shown in Supplementary Material 7.
[Table 2] shows the results of the correlation coefficient in both sexes between stature and all the other 18 body dimensions [Table 2]. In men, 12 out of the other 18 body dimensions (ULL, UAL, FAL, AML, TLL, TL, SL, FL, FB, HL, BHL, and MFL) were presented with variable degree of significant correlations with stature. However, in women, significant correlations were found in 15 out of theother 18 body dimensions (HC, HH, ULL, UAL, FAL, SB, AML, TLL, TL, SL, FL, FB, HL, BHL, and MF). In men, AML showed the highest correlation coefficient with stature (r = 0.817), whereas MHL demonstrated the lowest (r = −0.007). However, in women, TLL was presented with the highest correlation coefficient (r = 0.863) and MHL remained the lowest (r = 0.126). All the above-mentioned results are presented in Supplementary Materials 8-41.
Owing to the mentioned relations at different significant levels between stature and the other 18 body dimensions, we performed a series of stature estimations using the following different regression models: linear regression, step-wise regression, and multiple regression. Moreover, according to the sexual dimorphism which was verified by previous researches and our independent t-test, it is important to consider gender when calculating regression models.
Linear regression analysis
To carry out stature estimation with a single predictor, a linear regression analysis was conducted and its reliability to estimate stature was evaluated via SEE and R2 [Table 3]. [Table 3] illustrates the linear regression equations, SEE, and R2 values for stature estimation based on different body dimension measurements. Predictors with a higher correlation coefficient showed higher R2 and lower SEE values. Moreover, SEE ranged from ±4.000 to ±6.611 in men, and AML was found to be the predictor with the highest coefficient of determination (R2 = 0.667) and lowest SEE of ±4.000. Similarly, in women, SEE ranged from ±2.717 to ±5.468 and TLL resulted in the most reliable predictor with the highest coefficient of determination (R2 = 0.746) and lowest SEE of ±2.717. In addition, in males, SB showed the lowest R2 of 0.086 and MFL showed the highest SEE of 6.611. In females, FAL showed the lowest R2 of 0.068 and SB showed the highest SEE of 5.468.
|Table 3: Linear regression equations for stature estimation (cm) from human body dimensions|
Click here to view
Multiple regression analysis
For further investigations, combined dimensional measurements were used to estimate stature by conducting multiple regression analysis. The results of multiple regression analysis are presented in [Table 4]. In general, the SEE of multiple regression equations in both sexes are lower than those of linear regression equations when using different dimensional combinations as predictors, with the overall female model SEEs being relatively lower than those in males. In the present study, in women, the most ideal predictor was the combination of HH, AML, TLL, FAL, BHL, and FL, which showed the highest coefficient of determination (R2 = 0.832) and the lowest SEE (±2.304). In men, the predictor combination of HH, AML, TL, FAL, BHL, and FL showed the highest coefficient of determination (R2 = 0.804) and the lowest SEE (±3.213).
|Table 4: Multiple regression equations for stature estimation (cm) from human body dimensions|
Click here to view
Step-wise regression analysis
Considering the interference of multi-collinearity in multiple regression, a step-wise regression was carried out to prevent the above-mentioned potential influence (variance inflation factor <10). The list of the step-wise regression equations with relevant R2 and SEE is shown in [Table 5]. Due to the sexual dimorphism demonstrated before, the number of calculated equations varied with gender. Among the 23 step-wise regression equations for men, the combination of HH, AML, TLL, BHL, and FL was the best predictor with the highest R2 value of 0.795 and the lowest SEE of ±3.259. However, in the 26 step-wise regression equations for women, the combination of HH, AML, TLL, BHL, and FL also manifested the highest R2 value of 0.832 and the lowest SEE of ±2.286.
|Table 5: Step-wise regression models for stature estimation (cm) from human body dimensions|
Click here to view
| Discussion|| |
In this study, statistical sex differences were found in t-test, consistent with that of previous studies.,,,,, Therefore, it is necessary to consider sex differences when calculating regression models. In our studies, as for men, AML was the best predictor [Figure 1], while the most reliable predictor for women was TLL [Figure 2]. SB and MFL were found to be less reliable in men, whereas SB and FAL were less reliable in women. In general, the lower extremity dimensions were better than upper extremity dimensions, while the head dimensions were the worst as previously observed. For women, upper limb body dimensions, FAL, and BHL displayed unique relatively low correlation coefficients.,,, This inconsistency is probably due to the local environmental effect, the Han population's genetic heterogeneity, as well as dense distribution of a relatively low level of height, a fact which needs further investigation.
|Figure 1: Correlation analysis between AML and stature in men. R2 = 0.6672, P < 0.0001. AML: Anterior superior spine–malleolus medialis line|
Click here to view
|Figure 2: Correlation analysis between total leg length and stature in women. R2 = 0.7456, P < 0.0001. TLL: Total leg length|
Click here to view
As for multiple regression [Table 4], in women, the best predictor is the combination of HH, AML, TLL, FAL, BHL, and FL, with relatively better coefficient of determination (R2 = 0.832) and SEE (±2.304). As for men, the predictor combination of HH, AML, TL, FAL, BHL, and FL showed the highest coefficient of determination (R2 = 0.804) and the lowest SEE (±3.213). With relatively low multi-collinearity, step-wise regression has a promising performance. The combination of AML, TLL, BHL, and FL was the best predictor for men, whereas the combination of AML, FL, HH, and TLL manifested the highest R2 and the lowest SEE in women. These combinations that comprised the lower extremities often indicate better reliability., According to the increasing R2 values, the results of multiple regression and step-wise regression showed comparatively higher reliability for stature estimation than the linear regression due to the use of multiple indicators. In addition, although some dimensions such as HH (in men) and BHL (in women) had a weak correlation with stature, the result would be better if those participants were factored into the equations with increasing R2 and decreasing SEE. This improvement may relate to certain latent additive effect in stature estimation when adding parameters to the equations, but this hypothesis needs further verification.
Undeniably, despite our corroborative study, there are some limitations. First, instead of performing in the morning, we had to make all the measurements in the evening when the bulk of our volunteers were available for the experiment. In practice, anthropologists have already illustrated the intra-individual diurnal variation in stature, which is caused by the loss of fluid from the inter-vertebral discs and the vertebral column shortening from postural changes, which may substantially affect the reliability of height data. According to Krishan and Vij, there is a very rapid decrease in stature within the first 2 h of the day and further loss continues throughout the day in small amounts. Due to the nonsignificant stature loss in the evening, measurements taking at night might have made fewer errors than that in the day time. As a matter of fact, compared with measuring in the early hours of the morning, measurements in the evening are susceptible to daily physical activities and postural changes, leaving a limitation to our stature estimation.
Second, the bilateral differences in some dimensions such as ULL, UAL, FAL, AML, TLL, TL, SL, HL, HB, BHL, MFL, FL, and FB have not been studied in this research due to resource constraints. Therefore, we cannot verified whether there are any bilateral differences in those mentioned dimensions and their latent impacts when estimating stature. Previous studies have revealed controversy when there are bilateral differences in some dimensions., There are no statistically significant bilateral differences in limb lengths, as is evident from some studies. However, other researches carried out demonstrated statistical bilateral differences mostly in hand and FB, mainly for more intense physical activity of one side over the other.,
Accordingly, the significant bilateral differences in various dimensions still remain virtually unknown in a number of factors such as race, age, occupation, and physical condition. If we want to carry out more suitable equations for the Han population in southern China, it would be optimal to analyze both sides of all dimensions. Moreover, the age of volunteers in this study ranged from 18 to 25 years, which is representative of only young adults but not the whole Han population in southern China. Hence, it would be better to extend the range of volunteers' age in future studies.
| Conclusion|| |
The study was conducted at Southern Medical University in Guangdong Province, which involved a sample population of 121 students from southern China (59 men and 62 women) aged between 18 and 25 years. Significant differences were found among genders in our samples, hence the need to establish regression equations by gender. The most reliable predictors were AML in men and TLL in women, whereas the least reliable indicator with significant correlation with stature were MFL in men and FAL in women. Moreover, equations with multiple predictors proved better for estimating stature than those with a single predictor. Overall, the present study provides population-specific equations in three different models, which are reliable for stature estimation and individual identification in the southern China Han population.
The patient consent was obtained and the study was approved by the ethics committee of Southern Medical University, Guangzhou, China.
We sincerely appreciate all the participants and volunteers from Southern Medical University, whose cooperation and dedication have been an indispensable support to the study. This contribution is the result of the National Students' Innovative and Entrepreneurship Training Program (No. 20181212112X).
Financial support and sponsorship
This study was supported by the National Natural Science Foundation of China (Grant No. 81871526) and the National Students' Innovative and Entrepreneurship Training Program (No. 20181212112X).
Conflicts of interest
There are no conflicts of interest.
| References|| |
Uhrova P, Benus R, Masnicova S, Obertova Z, Kramarova D, Kyselicova K, et al
. Estimation of stature using hand and foot dimensions in Slovak adults. Leg Med (Tokyo) 2015;17:92-7.
de Boer HH, Blau S, Delabarde T, Hackman L. The role of forensic anthropology in disaster victim identification (DVI): Recent developments and future prospects. Forensic Sci Res 2019;4:303-15.
Ubleaker DH, Khosrowshahi H. Estimation of age in forensic anthropology: Historical perspective and recent methodological advances. Forensic Sci Res 2019;4:1-9.
Mahakizadeh S, Moghani-Ghoroghi F, Moshkdanian Gh, Mokhtari T, Hassanzadeh G. The determination of correlation between stature and upper limb and hand measurements in Iranian adults. Forensic Sci Int 2016;260:27-30.
Zhang K, Zhan MJ, Cui JH, Luo YZ, Qiu LR, Deng LP, et al
. Estimation of stature from radiographically determined lower limb bone length in modern Chinese. J Forensic Leg Med 2019; doi.org/10.1016/j.jflm.2019.02.012.
Menéndez Garmendia A, Sánchez-Mejorada G, Gómez-Valdés JA. Stature estimation formulae for Mexican contemporary population: A sample based study of long bones. J Forensic Leg Med 2018;54:87-90.
Ahmed AA. Estimation of stature from the upper limb measurements of Sudanese adults. Forensic Sci Int 2013;228:178.e171-7.
Howley D, Howley P, Oxenham MF. Estimation of sex and stature using anthropometry of the upper extremity in an Australian population. Forensic Sci Int 2018;287:220.e210-21.
Akhlaghi M, Hajibeygi M, Zamani N, Moradi B. Estimation of stature from upper limb anthropometry in Iranian population. J Forensic Leg Med 2012;19:280-4.
Kim W, Kim YM, Yun MH. Estimation of stature from hand and foot dimensions in a Korean population. J Forensic Leg Med 2018;55:87-92.
Uhrova P, Benus R, Masnicova S. Stature estimation from various foot dimensions among Slovak population. J Forensic Sci 2013;58:448-51.
Caplova, Z, Svabova P, Fuchsova M, Masnicova S, Nescakova E, Bodorikova S, et al
. Estimation of stature and body weight in Slovak adults using static footprints: A preliminary study. Leg Med (Tokyo) 2018;34:7-16.
Kanchan T, Krishan K, Sharma A, Menezes RG. A study of correlation of hand and foot dimensions for personal identification in mass disasters. Forensic Sci Int 2010;199:112.e111-6.
Torimitsu S, Makino Y, Saitoh H, Sakuma A, Ishii N, Yajima D, et al
. Stature estimation from skull measurements using multidetector computed tomographic images: A Japanese forensic sample. Leg Med (Tokyo) 2016;18:75-80.
Ramezani M, Shokri V, Ghanbari A, Salehi Z, Niknami KA. Stature estimation in Iranian population from x-ray measurements of femur and tibia bones. J For Radiol Imag 2019;19:100343.
Zhan MJ, Cui JH, Zhang K, Chen YJ, Deng ZH. Estimation of stature and sex from skull measurements by multidetector computed tomography in Chinese. Leg Med (Tokyo) 2019;41:101625.
Dedouit F, Guglielmi G, Olier A, Savall F, Nasuto M, Thanassoulas T, et al
. Analysis of size and shape differences between ancient and present-day Italian crania using metrics and geometric morphometrics based on multislice computed tomography. Forensic Sci Res 2017;2:85-92.
Li CZ, Wu W, Zhu B, Liu XF, Huang P, Wang ZY, et al
. Multiple regression analysis of the craniofacial region of Chinese Han people using linear and angular measurements based on MRI. Forensic Sci Res 2017;2:34-9.
Hishmat AM, Michiue T, Sogawa N, Oritani S, Ishikawa T, Fawzy IA, et al
. Virtual CT morphometry of lower limb long bones for estimation of the sex and stature using postmortem Japanese adult data in forensic identification. Int J Legal Med 2015;129:1173-82.
Xi HJ, Chen Z. Anthropometry Method. Beijing: Science Press; 2010. p. 147-75.
Krishan K, Sharma A. Estimation of stature from dimensions of hands and feet in a North Indian population. J Forensic Leg Med 2006;14:327-32.
Ishak NI, Hemy N, Franklin D. Estimation of stature from hand and handprint dimensions in a Western Australian population. Forensic Sci Int 2012;216:199.e191-7.
Giurazza F, Vescovo RD, Schena E, Battisti S, Cazzato RL, Grasso FR, et al
. Determination of stature from skeletal and skull measurements by CT scan evaluation. Forensic Sci Int 2012;222:398.e391-9.
Krishan K, Kanchan T, DiMaggio JA. A study of limb asymmetry and its effect on estimation of stature in forensic case work. Forensic Sci Int 2010;200:181.e181-5.
Fairbank J. Height measurements and stretching. Lancet 1998;351(9110):1212.
Krishan K, Vij K. Diurnal variation of stature in three adults and one child. Anthropologist 2017;9:113-7.
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5]