Stature estimation from handprint measurements: an application to the medicolegal investigation

Background The estimation of the stature of human beings is a major part of medicolegal investigation when only body parts are found. The study aimed to estimate the stature from different handprint measurements in a Bangladeshi population using statistical considerations. A sample of 200 young Bangladeshi adults (100 men and 100 women) with no physical disabilities participated in this study. Stature and seven anthropometric measurements were measured using standard anthropometric measurements. The bilateral asymmetry was tested using the independent t test. The Pearson’s correlation coefficient (R) between the stature and different handprint measurements was calculated. Consequently, the simple and multiple linear regression models were developed to estimate the stature from the handprint measurements. Results The bilateral asymmetry was statistically not significant (p > 0.05) in right and left handprints. Sex difference significantly (p < 0.05) influences the relationship between stature and handprint measurements. A positive and strong coefficient of correlation (R) value presents between stature and the handprint measurements. The right 2nd digit length in men (R = + 0.709, R2 = 0.502, SEE = ± 44.141 mm) and the right handprint length in women (R = + 0.552, R2 = 0.305, SEE = ± 49.074 mm) were the most reliable estimator of stature. However, when data were combined for both sexes, the right handprint length was identified as the most reliable estimator of stature with higher values of R (+ 0.777) and R2 (0.603), and a lower value of SEE (± 55.520). Multiple regression equation showed greater reliability than linear regression equations in stature estimation from handprint measurements in Bangladeshi population. Conclusions It was concluded that the estimation of stature from handprint measurements is possible and reliable. The findings of this study are very useful from the forensic and medicolegal point of view and can use to estimate the stature in Bangladeshi population.


Background
The identification of the criminal or the perpetrator is the primary goal of a case solver to deal with a crime case. This process can be hindered by lack of evidence found at the crime scene. However, it would be a great support for the law enforcement agency if a biological profile of the suspects can be estimated and developed from the scarce evidence. It would help them in narrowing down the pool of suspects.
Although numerous studies have been conducted for estimating the stature from footprints notwithstanding less work has performed on the handprints. In the past studies, it has been found that the handprint dimensions can be successfully used for the estimation of the stature (Ishak et al. 2012b;Krishan et al. 2015;Moorthy and Yin 2016;Paulis 2015;Salama 2013;Zulkifly et al. 2018). Those studies revealed a strong positive correlation between handprint dimensions and stature for both hands and sexes. Thus, simple (Ahemad and Purkait 2011;Ishak et al. 2012b;Paulis 2015;Zulkifly et al. 2018) and multiple linear regression models (Ishak et al. 2012b;Paulis 2015;Zulkifly et al. 2018) were successfully used by previous researchers for estimating the stature. The study of Ahemad and Purkait (2011) on 503 Indian men revealed the strong positive correlation (p < 0.05) between handprint dimensions and stature. Among the seventeen hand dimensions studied, hand length was the most reliable parameter for estimating the stature. Ishak et al. (2012b) analyzed seven handprint dimensions in a Western Australian population and found a strong positive correlation (p < 0.01) between handprint dimensions and stature in both sexes. Moreover, Paulis (2015) conducted a study on the Egyptian population and used simple and multiple linear regression for estimation stature from handprint dimensions. The study revealed that hand length was the most reliable single factor for estimation stature from handprint dimensions. Similarly, Zulkifly et al. (2018) found a strong positive correlation (p < 0.05) between handprint dimensions and stature on Iban subjects in both sexes. Zulkifly et al. (2018) reported the handprint length as the most reliable estimator of stature; the correlation coefficient between stature and handprint length was ranged between + 0.59 and + 0.68.
Anthropometric measurements vary from one population to another (Asadujjaman et al. 2017(Asadujjaman et al. , 2020b. The degree of access to nutrition and health services may have an effect on the stature of the different populations around the world (Perkins et al. 2016). Moreover, as genetics can play a vital role in human growth, therefore the previous studies based on a particular population cannot be used for estimating stature from handprints for other populations.
In Bangladesh, to date, to the best of our knowledge, no study has conducted to estimate the stature from handprint measurements in this population. Therefore, this study aimed to investigate the possibility of establishing the relationship between stature and handprints, and finally to derive standard linear and multiple regression formula for the Bangladeshi population to estimate the stature from handprint measurements.

Materials
In this study, data were obtained from 200 volunteers having no physical disability (100 male and 100 female) aging from 18 to 30 years (23.62 ± 2.96 years in males and 23.58 ± 3.48 years in females) old. A separate questionnaire for collecting basic demographic information such as sex and age was provided while collecting the measurements of stature and handprints of each subject. Consent and permission for using their personal data for research work were collected from the volunteers in written form. The researchers are committed to protecting the privacy of personal information of the participants.

Handprint acquisition
A flat box with powdered soil was used for collecting the handprints of both hands of the subjects. The subject was asked to make hand impressions on the powdered soil and thus, a digital slide calliper was used for acquiring the measurement of handprints. While taking the hand impression on the powdered soil, it was ensured that the powdered soil was flat every time. The technique to measure the hand length from handprint is shown in Fig. 1.

Measurements
Stature and seven handprint measurements (handprint breadth, maximum handprint breadth, 1 st digit length, 2 nd digit length, 3 rd digit length, 4 th digit length, and 5 th digit length) were measured using standard measuring techniques. The anthropometric measurements were followed by the study of Moorthy and Yin (2016). Measurements of handprint were taken for both left and right hands. Different handprint measurements are illustrated in Fig. 2. All the studied handprint measurements were taken from the baseline shown in Fig. 2.
Stature is the person's natural height in an upright position. The subject was asked to stand in an erect position without any wear on foot and head to take the measurement. Therefore, the distance from the floor to the highest point of the head was taken as the stature. The maximum handprint breadth is the distance from the most lateral point of the thumb finger metacarpal head to the most medial point of the little finger metacarpal head with closing the fingers of the handprint. The 1 st 2 nd , 3 rd , 4 th , and 5 th digit lengths are the distance between the midpoints of the distal transverse crease of the wrist to the thumb, index, middle, ring, and little fingertip of the handprint, respectively. The handprint length is the 3 rd digit length of the handprint.
To avoid inter-observer error, all the measurements were taken by one observer. To limit measurement error, all the measurements of each subject was taken twice, and the mean value was taken. If two primary measurements have a difference of more than 4 mm, then both data were rejected, and again two new measurements were taken, and the mean was used. Intraobserver measurement errors for 30 subjects were then calculated. The measurement margin error, then calculated using the absolute technical error measurement (TEM), relative technical error measurement (rTEM), and coefficient of reliability (r) (Kim et al. 2018;Ulijaszek and Kerr 1999).

Statistical analysis
For statistical analysis of the data, SPSS statistical software (version 23.0) and Microsoft Excel 13 was used. Descriptive statistics such as mean and standard deviation were calculated. A paired sample t test was performed to analyze the bilateral asymmetry in handprint measurement. An analysis of covariance using the general linear model (GLM) was conducted to test whether sex has an effect on the stature and handprint measurements. To estimate the stature from handprint measurements, simple linear and multiple linear regression models were used. Pearson correlation coefficient (R) was used for establishing the relationship between stature and handprint measurements. R value indicates the strength and direction of the relationship between the  stature and handprint measurements. The coefficient of determination (R 2 ), standard error of estimation (SSE), and p values were also used to describe the prediction of stature from handprint measurements from regression equations. R 2 interprets the proportion of the variance in stature that is estimated from the handprint dimensions. The values of R 2 vary from 0 to 1. Higher values of R and R 2 means a greater reliability in predicting the stature from handprint measurements with a lower prediction error. The SEE predicts the deviation of estimated stature from the actual value. It represents the average distance that the observed stature measurement falls from the regression line. A low value of SEE means greater reliability in the stature estimation. P value shows the statistical significance in stature estimation from the handprint measurements. A p value less than 0.05 is statistically significant. Table 1 shows the assessment of the intra-observer error in the measurement variables used in this study. In measurement variables, rTEM was less than 5% and the r values were higher than 0.97 in measurement variables. According to Ulijaszek and Kerr (1999), intra-observer error was regarded as an acceptable range for the measurement variables.

Significance test
Descriptive statistics (mean, standard deviation) for handprint measurements and the results of paired sample t test to analyze the bilateral asymmetry in handprint dimensions for both sexes are presented in Table 2. The mean of all measurements of male subjects was larger than the female subjects. There was no statistically significant bilateral asymmetry in handprint dimensions (p > 0.05) in both sexes.
The mean stature in males and females was 1694.057 ± 62.236 mm and 1566.418 ± 58.554 mm, respectively. The results of the analysis of covariance using GLM revealed the influence of sex on the relationship between stature and handprint measurements (p < 0.05). Therefore, the regression models were developed separately by sex.

Stature estimation from handprint measurements using linear regression analysis
The linear regression equations with the R, R 2 , SSE, 95% prediction interval, and p value to estimate the stature from left-and right-handprint dimensions are presented in Table 3. The investigator or the police does not know whether the handprint was made by a man or a woman. Therefore, to apply the method in real cases where the sex of the subjects is not available from a handprint, the regression equations were developed combining both male and female data. There was a substantial amount of statistical significance (p < 0.001) in the correlation coefficients of all the derived regression equations of all handprint parameters of males and females. The values of R ranged from + 0.472 to + 0.709 in males, and + 0.333 to + 0.552 in females. Regression equations developed by combining both male and female data revealed a higher value of R ranged between + 0.587 and + 0.777. In males, the maximum value of R was found between stature and 2 nd digit length of right handprint (R = + 0.709). On the other hand, in females, the maximum value of R was between stature and the right handprint  length of right handprint (R = + 0.552). However, using the combined data, the right handprint length (R = + 0.777) was the most reliable estimator of the stature. The value of R 2 ranged from 0.223 to 0.502 in males, 0.111 to 0.305 in females, and 0.344 to 0.603 in combined data. The SEE in males was varied from ± 44.141 to ± 55.138 mm, in females from ± 49.074 to ± 55.495 mm, and in combined data from ± 55.490 to ± 71.360 mm. The 95% prediction interval ranged from ± 92.608 to ± 107.527 mm in males, ± 96.185 to ± 108.770 mm in females, and ± 108.819 to ± 118.272 mm in the combined sex.

Stature estimation from handprint measurements using multiple regression analysis
Stature estimation accuracy can be improved by formulating multiple regression equations (Ahemad and Purkait 2011;Ishak et al. 2012b;Paulis 2015). Multiple regression equations are presented in Table 4. All the handprint parameters were combined to formulate the

Discussion
In this study, the stature of male subjects (mean 1694.06 ± 62.24 mm) was larger than female subjects (mean 1566.42 ± 58.55 mm). Measurements are larger in males than females were also true for all handprint dimensions ( In our study, no bilateral asymmetry was observed. However, bilateral asymmetry was reported in some previous studies. Ahemad and Purkait (2011) found bilateral asymmetry in handprint width, distal segment of the thumb, and distal and middle segments of the little finger. Ishak et al. (2012b) found bilateral asymmetry in handprint breadth of female subjects. Zulkifly et al. (2018) reported bilateral asymmetry in handprint breadth, index fingers, thumb distal, and index distal phalanges for female subjects.
From the linear regression analysis (Table 3), it was seen that right handprint 2 nd digit length was the most precise single factor to estimate stature from handprint dimensions in males with the lowest value of SSE (± 44.141 mm) and the highest value of R (+ 0.709) and R 2 (0.502) in males. In contrast, in females, the right handprint 3 rd digit length was reported the highest value R (+ 0.552) and R 2 (0.305) with the lowest value of SSE (± 49.074 mm). Therefore, the right handprint 3 rd digit length or the right handprint length was the most precise single factor to estimate the stature from handprint dimensions in females. Using the combined data set, the right handprint length was the most reliable single parameter to estimate the stature (R = + 0.777, R 2 = 0.603, SEE = ± 55.490 mm). Figures 3, 4, and 5 show the best fit curves to estimate stature using the most reliable single parameter in males, females, and combined data, respectively. A comparative study of the values of the R and R 2 between stature and different handprint measurements found among various populations is shown in Tables 5 and 6, respectively. Similar to this present study, all the previous studies (Ahemad and Purkait 2011;Ishak et al. 2012b;Paulis 2015;Zulkifly et al. 2018) found a positive correlation between the handprint measurements and the stature. Handprint length (i.e., the 3 rd digit length in this study) was the only common handprint measurement between this study and previous studies. Therefore, the comparison was presented with respect to hand length. In the study of Paulis (2015) done on the Egyptian population, right hand length of males (R = 0.519, R 2 = 0.270, SSE = ± 45.40 mm) and females (R = 0.298, R 2 = 0.089, SSE = ± 53.80 mm) were more dependable. In a study of central Indian male population by Ahemad and Purkait (2011), hand length (R = 0.558, R 2 = 0.312, SSE = ± 46.35 mm) was the most reliable parameter for estimating stature. In another study of the Western Australian population, Ishak et al. (2012b) reported that the right handprint length in males (R = 0.640, R 2 = 0.410, SSE = ± 54.20 mm) and left handprint length of females (R = 0.650, R 2 = 0.420, SSE = ± 54.60 mm) were the most reliable parameters. Similarly, the right handprint length in both males (R = 0.680, R 2 = 0.460, SSE = ± 55.50 mm) and females (R = 0.640, R 2 = 0.410, SSE = ± 46.70 mm) were also noted by Zulkifly et al. (2018) as most dependable parameters for estimating the human height. Different handprint dimensions were used by different researchers as the independent variable of the linear regression equations for estimating the stature (Table 7). Paulis (2015) developed regression equations using handprint breadth, handprint length, and phalangeal length, whereas the SSE was between ± 45.4 to ± 58.9 mm in males, and ± 53.8 to ± 66.8 mm in females. Moreover, Ahemad and Purkait (2011) used handprint length,  The SSE values were ranged from ± 54.2 to ± 75.5 mm in males, and ± 46.7 to ± 59.9 mm in females. In this present study, the regression equations were derived using handprint breadth, maximum handprint breadth, 1st digit length, 2nd digit length, 3rd digit length, 4th digit length, and 5th digit length. In comparison, our study revealed lower values of SSE (± Fig. 5 Best fit curve to estimate stature from combined handprint 3 rd digit length  The multiple regression equations were developed for both hands and for both sexes, to improve the accuracy of estimation (Table 4). When all the parameters were considered for stature estimation, the accuracy was increased for both sexes and both hands as the values of R have increased and the values of SSE have decreased in every case. The values of SSE ranged from ± 43.937 to ± 44.031 mm in males, and ± 47.830 to ± 48.265 mm in females, which was lower than the simple linear regression models. The largest value of R (+ 0.729) and the lowest value of SSE (± 43.937 mm) was found in the right hand of males. In contrast, the lowest value of R (+ 0.601) and the highest value of SSE (± 48.26 mm) was found in the left handprint of males. Therefore, the right handprint dimensions in males were more reliable, and the left handprint dimensions in females were less reliable for the estimation of stature. However, using combined sex data, multiple regression models showed greater reliability in stature estimation. The right handprint measurements showed the highest reliability with a value of R = + 0.819, R 2 = 0.671, and SEE = ± 51.166 mm in combined sex. These findings are similar to the past research where the estimation accuracy was improved by using multiple regression analysis (Ishak et al. 2012b;Paulis 2015;Zulkifly et al. 2018).

Application of handprint in forensic practice
When there are no suspects comparing, estimating stature from the handprints found at a crime scene can help the law enforcement agencies to narrow down the pool   , Egyptian (Paulis 2015), Western Australian (Ishak et al. 2012b), and Iban (Zulkifly et al. 2018) population. Handprints have been used by law enforcement authorities in many cases to identify the criminals even when they already have several suspects (Handprint left at "massacre" scene 2002; Suspect linked to bloody handprint in 1986 killing pleads not guilty 2019; Forensic expert testifies about bloody hand prints found in apartment as Berry murder trial continues 2019). Handprints may be found in various forms such as on a flat surface, or on soil or mud. In developing countries (e.g., Bangladesh, Pakistan, India), clash or fight between two parties in the open field is common. In case of clash or fight between two parties, handprint may find on soil or mud. For instance, in 2015, 10 injured in Brahmanbaria clash (2015) where people fight in the open field. Even in this Covid 19 period, in Brahmanbaria, a clash between villagers happened in an open ground (Villagers keep fighting amid coronavirus 2020). In such cases, where when people fight in the field, the injured even can die owing to fighting, where handprint may be found in the soil and that may be used to identify the criminals or victims.
Various method of handprint acquisition was used by previous researchers. Sharma and Kapoor (2001) narrated a method of estimating stature from fingertip length and fingerprint length using inked impressions.
Further, Jasuja and Singh (2004) collected handprint measurements by taking an inked impression of hands and estimated stature from handprint length and breadth. Moreover, Ahemad and Purkait (2011) collected handprints by taking an inked impression and only studied men subjects only. Zulkifly et al. (2018) also used ink impression of hands for collecting handprint dimensions. Furthermore, Ishak et al. (2012b) collected scanned images of hands and from the printed copies of these images hand and finger dimensions were manually measured. Beside this, Paulis (2015) automated the handprint collection process and used a computer to measure the handprint dimensions from scanned images of hands.
In this study, the handprint dimensions were collected by taking hand impressions on powdered soil, which was a new method and not examined in the literature. However, handprint may be found on a flat surface. Therefore, the difference between the handprint on soil and the flat surface has been examined in this study. Figure 6 shows the hand impression on the soil and the flat surface of the same participant at the same hand pressure. Ink impression was used to take the handprint on the flat surface (Fig. 6b). The difference between handprint measurements on soil and flat surface is presented in Table 8. It was seen that the handprint impression on soil was larger than the handprints on the flat surface. When the hand impression was taken on the flat surface, full hand area did not touch the flat surface. On the other hand, at the same pressure when hand impression was taken on soil, hand penetrates a little bit into the Fig. 6 a Handprint on soil. b Handprint on flat surface soil owing to the softness. Therefore, handprint dimensions were larger on soil than the flat surface. The ratio of handprint measurements on soil and flat surface is useful to identify the individuals, whether the hand impression is found on the soil or the flat surface.

Conclusion
This study reported the application of handprint measurements in stature estimation in Bangladeshi adults. The stature can be reliably estimated from the handprint dimensions. The 2 nd digit length of the right handprint in males, and the handprint length in females were the most reliable single parameter for the estimation of stature. Using combined sex data, the handprint length was the most reliable single parameter to estimate the stature. The estimation accuracy was increased in the case of multiple regression analysis. This study uses the hand impression on soil which is not reported in previous studies. Therefore, this study examines the measurement difference between the handprint on soil and flat surface. Consequently, this study is useful to find the stature of unknown individuals, whether the hand impression is found on a flat surface or soil. For forensic, medicolegal, and crime investigation purposes, the new standard to estimate the stature has a great impact in Bangladesh. The pool of suspects can be narrowed down form the handprints found at a crime scene and thus can be cross-matched with other pieces of evidence. The age of subjects taken in this study of the Bangladeshi population ranged between 18 and 30 years. In the future, new studies can be done on subjects for other different age ranged people of Bangladesh.