Introduction

One of the goals of forensic analysis is the identification of skeletonized or decomposed human remains. Together with sex, age, and ancestry, stature is a biological parameter that characterizes individuals and can be evaluated from the skeleton even when many years have passed after death [1–4]. These parameters expedite the analysis of remains to narrow down the pool of cadavers needing to be matched and allow for more definitive markers, such as DNA, to be used for the confirmation of final identification [5].

There are two methods to estimate stature from human remains. One is the anatomical method, which requires a whole skeleton to calculate the combination of several bone lengths [6]. Although this method is held to be the most accurate method for stature estimation, it is often unavailable because a complete and well-preserved skeleton is very uncommon after serious events such as natural disasters, plane crashes, and terrorist attacks. The other is the mathematical method, which employs sex- and population-specific regression formula derived from correlations between stature and the measurements of some skeletal elements. Many previous studies used extremity measurements for stature estimation because comparatively accurate estimates can be produced from long bone regression equations [6–9]. However, if corpses are on the ground surface, the extremities are often unavailable, having been chewed and removed from the cadaver by animals [10]. In addition, intact long bones may not be available in some mass disaster, burn, and skeletal cases [11]. Thus, there is still a need to assess the correlation of bones other than long bones with stature.

Several studies have examined the relationship between stature and skeletal elements, such as the size of the cranium [12], scapula [13], sternum [14–16], vertebrae [17–20], and sacrum [10, 11, 21, 22]. However, despite the fact that the pelvis has been used all over the world for the determination of sex [23] and age estimation [24–31], few studies have investigated the correlation between measurements of the pelvic bones and stature. Although Giroux and Wescott [11] examined the utility of pelvic bone measurements for stature estimation in dry bones of American Black and White males and females, our search of the literature revealed that the correlation between stature and pelvic measurements using three-dimensional (3D) multidetector computed tomographic (MDCT) images in a Japanese population has not been investigated. We previously reported that the length of the sacrum and coccyx using MDCT can predict stature with precision [21]. In addition, the coxa and femoral head are often preserved in skeletonized, mutilated, and burned human remains that have an intact sacrum, and measurements of these bones are commonly collected by anthropologists [32]. Thus, we hypothesized that pelvic measurements using MDCT images may also be useful in stature estimation.

Recently, postmortem computed tomography (PMCT) has been routinely performed using MDCT in forensic departments and institutes equipped with computed tomography (CT) [33]. 3D images of the pelvis can be easily generated from MDCT data. In particular, forensic investigators can assess and evaluate bones on CT images without the removal of surrounding tissues when a subject of measurement is not a skeletonized bone sample. Therefore, if CT for the evaluation of pelvic bones is available, it may result in the reduction of the time and cost of forensic investigations.

The aim of this retrospective study was to examine the feasibility of stature estimation based on measurements of the pelvic bones of Japanese cadavers using 3D CT images.

Materials and methods

The experimental protocol was approved by the ethics committee of our university and did not require approval of the subjects’ next of kin.

The present study reviewed the data of 210 cadavers (108 males and 102 females) of known age and sex who had undergone PMCT with subsequent forensic autopsy at the department of legal medicine at our university between May 2011 and November 2013. Cases were excluded if the history highlighted conditions or events that could have affected the pelvis; cases of pelvic fracture, burning, buttock injuries, and acquired or congenital abnormalities were excluded from this study.

During autopsy, the cadaver was laid on its back in full extension and a measuring tape was used against the board to measure the cadaver stature (CS) as the maximum length between the skull vertex and the heel in millimeters after the rigor mortis of the cadaver was fully resolved. To determine the intraobserver and interobserver errors of the measurements, stature was remeasured by both the original researcher and another pathologist not participating in the study.

PMCT was performed with a 16-row detector CT system (ECLOS; Hitachi Medical Corporation, Tokyo, Japan). The scanning protocol was as follows: collimation of 1.25 mm, reconstruction interval of 1.25 mm, tube voltage of 120 kVp, tube current of 200 mAs, and rotation time of 1 per second. A hard filter was used. Image data were processed on a workstation (SYNAPSE VINCENT; Fujifilm Medical, Tokyo, Japan) to obtain orthogonal multiplanar reconstruction images and volume-rendered images.

A 3D reconstructed image was used for assessment. The linear distances from the anterosuperior margin of the left and right anterior superior iliac spines (ASIS) to the posterior margin of the left and right ischial spines were defined as the LSS and RSS, respectively (Figs. 1a, c). The linear distances from the anterosuperior margin of the left and right ASIS to the anteroinferior margin of the left and right ischial tuberosities were defined as the LST and RST, respectively (Figs. 1b, c). These measurements were performed using electronic cursors and were recorded to the nearest 0.1 mm. Repeat measurements of 20 images selected randomly from the subjects were performed by both the original and another researcher to evaluate intraobserver and interobserver errors.

Fig. 1
figure 1

CT scan imaging with volume rendering reconstruction; a linear distances from the anterosuperior margin of the left and right anterior superior iliac spines to the posterior margin of the left and right ischial spines (LSS and RSS) in the coronal plane viewed from posterior side of the body; b linear distances from the anterosuperior margin of the left and right anterior superior iliac spines to the anteroinferior margin of the left and right ischial tuberosities (LST and RST) in the coronal plane viewed from anterior side of the body; c LSS and LST in the sagittal plane viewed from left side of the body

Full size image

Statistical analyses were performed on a personal computer using Statistical Package for the Social Sciences (SPSS) version 21.0 computer software (IBM Corp., Armonk, NY) and Excel software (Microsoft 2010). The mean, standard deviation, and the range of the four parameters (LSS, LST, RSS, and RST) were calculated for all subjects as well as for the male and female subjects separately. To evaluate intraobserver and interobserver errors, the technical error of measurement (TEM), the relative technical error of measurement (rTEM), and the coefficient of reliability (R) were calculated. The acceptance ranges of rTEM (%) using beginner anthropometrist levels for intraobserver error was less than 1.5 % and for interobserver error, less than 2.0 % [34]. R values more than 0.95 were considered as sufficiently precise [35]. The correlation between the CS and each parameter (LSS, LST, RSS, and RST) was evaluated using Pearson product–moment correlation coefficients (r). P values less than 0.05 were considered statistically significant. Simple linear regression analysis was performed for the four parameters to derive regression formula for stature estimation. Multiple regression analysis for stature estimation was also performed using the multiple parameters to determine the most accurate formula. The coefficient of determination (r 2) and the standard error of estimation (SEE) were calculated to assess the significance of regression.

Results

Descriptive statistics of age, stature, and the four parameters are presented in Table 1. The calculated TEM, rTEM, and R are tabulated in Table 2. All the rTEMs for intraobserver error (0.261–0.497 %) were less than 1.5 % and for interobserver error (0.283–0.543 %) less than 2 %. All the R values (0.952–0.993) were more than 0.95.

Table 1 Descriptive statistics of cadaver age, stature, and the four parameters (LSS, LST, RSS, and RST)
Full size table
Table 2 Technical error of measurements (TEM), relative technical error of measurements (rTEM), and coefficient of reliability (R) (n = 20)
Full size table

Table 3 illustrates the correlation coefficients between stature and each parameter among all subjects, among male cadavers, and among female cadavers, respectively. All measurements presented statistically significant correlation with stature (P < 0.001) (Fig. 2). The correlation coefficients for measurements with comparatively large values (LST and RST) were greater than those with small values (LSS and RSS). The highest correlation coefficient was found for LST (all subjects, r = 0.922; male subjects, r = 0.881; female subjects, r = 0.770), and the lowest correlation coefficient for RSS (all subjects, r = 0.873; male subjects, r = 0.758; female subjects, r = 0.705). The correlation coefficients of each of the four parameters were higher among male subjects than among female subjects.

Table 3 Correlation between stature and each variable among all subjects, among male subjects, and among female subjects
Full size table
Fig. 2
figure 2

Scatter plots comparing a LSS and cadaver stature (CS), b LST and CS, c RSS and CS, and d LST and CS. Closed circles represent male subjects and open circles represent female subjects

Full size image

Table 4 shows the simple linear regression equations, the SEEs, and the coefficients of determination derived for stature estimation in all subjects, male subjects, and female subjects from each of the four parameters. The SEEs were in the range of 4.40–5.56 cm for all subjects, 3.72–5.14 cm for male subjects, and 5.02–5.57 cm for female subjects, respectively. The SEEs calculated from the simple linear regression equations for male subjects were lower than those for female subjects.

Table 4 Linear regression equation for stature estimation (in cm) from each variable among all subjects, among male subjects, and among female subjects
Full size table

Table 5 shows the multiple regression equations formulated using multiple variables. The SEEs calculated from multiple regression equations were lower (4.34–4.56 cm for all subjects; 3.69–3.89 cm for male subjects; 4.93–5.15 cm for female subjects) than those calculated from simple linear regression equations.

Table 5 Multiple regression equations for stature estimation (in cm) from each variable among all subjects, among male subjects, and among female subjects
Full size table

Discussion

The main findings of this study were the significantly positive correlations between stature and the four parameters independent of sex, indicating that measurements of the pelvic bones on 3D CT images may be a reliable predictor of stature in Japanese cadavers. In particular, the LST is the best single-element estimator in both sexes. Studies on the correlations between stature and the pelvic bones are summarized in Table 6. Compared with previous studies, the correlation coefficients in this study were the largest and the SEEs were the smallest. Giroux and Wescott [11] reported that hip height exhibited stronger correlation with stature than sacral height. The present study also demonstrated similar results in comparison to our previous study [21]. Giroux and Wescott [11] reported that interobserver error and differences are the most serious problems in the measurements of cadaver stature. In our study, the TEM, rTEM, and R values were calculated and the intraobserver and interobserver errors for measurements of stature and pelvic bones were found to be very small. Taken together, these findings suggest that measurements of the pelvic bones, especially the ilium and ischium, may be appropriate and reliable for stature estimation if CT scanning with 3D reconstruction is available.

Table 6 Summary of studies on the correlation between stature and measurement of pelvic bone
Full size table

Note that Sakuma et al. [36] performed the measurements of the skull with calipers, and then also performed the measurements of the 3D CT reconstructed images in our department, revealing that there was no significant difference between the actual measurements and the measurements on the 3D CT images. Thus, the measurements of the pelvic bones on the 3D CT images are also considered similar to real measurements on dry bones performed with calipers.

To date, few researchers have applied CT scanning to the field of anthropometry. Giurazza et al. [37] evaluated the correlation between stature and the measurements of femur and skull through CT scans and derived regression equations for stature estimation in the Caucasian population, concluding that a CT scan is a valuable technique to measure bone length. Giurazza et al. [38] assessed the correlation between scapula size and stature by CT scan evaluation and reported that CT is one of the most accurate radiographic techniques for the study of bone morphometry. We previously evaluated the relationship between stature and the length of the sacrum and coccyx using MDCT and derived regression equations for stature estimation in the modern Japanese population, and the calculated SEEs were lower than those calculated from the real measurements of dry bones, probably because of the difficulty in measuring dry bones with precision [21]. In addition, MDCT enables forensic investigators to perform measurements on bone images instead of fresh bones that require preparations such as removing and cleansing the skeleton if surrounding tissues remain [21]. Moreover, digital CT images can be preserved almost permanently in little or no physical space, enabling repeated measurements and potentially decreasing intraobserver and interobserver error [10].

In this study, the multiple regression equations provided better estimates (lower SEEs and higher r 2) than the simple linear regression equations in all cases. In accordance with the previous studies using pelvic bones, these results suggest that the multiple regression equations for the estimation of stature are more reliable than the single linear regression equations [11, 22, 39]. Hasegawa et al. [40] estimated stature from the long bones of the limbs of Japanese living subjects and showed lower SEEs using the lower limbs (2.63–3.03 cm) than those in our study (3.64–4.88 cm). However, the SEEs calculated using the humerus were in the range of 3.75–4.45 cm, which were similar to those in our study, again suggesting that measuring the size of the pelvic bones with 3D CT may be a useful tool for stature estimation in bodies with no lower limbs.

In this study, we did not apply any correction to the recorded cadaver stature measurements. It is well known that cadaver stature is different from living stature because of the physiologic changes that occur in the human body after death [37]. Therefore, the possibility cannot be excluded that the formula derived in this study may not be expected to accurately estimate stature. In addition, standing living stature may also not be used in identification. Willey and Falsetti [41] reported that forensic stature, which is the stature recorded on driver’s licenses, medical records, and other documents, is often greater than the stature directly measured from a corpse. However, in some cases, measured stature is greater than the stature on the driver’s license probably because the individual failed to update the driver’s license information after growth [41, 42]. Thus, if the information about stature that is used in the identification process is derived from personal documents, it tends to further confound stature estimation.

In conclusion, our measurements of the pelvic bones performed on 3D CT images significantly correlated with stature in all cases, indicating that they may be a reliable stature predictor in the Japanese population and can be used as a tool for stature estimation, especially for corpse identification, when long bones are unavailable or not intact. However, owing to the physiologic changes that occur in the human body after death, the possibility cannot be excluded that the formula developed in this study may not be used to accurately estimate stature. In particular, the LST had the best correlation with stature among Japanese cadavers regardless of sex. Similar studies in other age groups and in different populations are warranted.