
A Cross-sectional Study on Stature Estimation from Arm Lengths among North Indian Population using Machine Learning
HC01-HC05
Correspondence
Dr. Arunima Dutta,
Assistant Professor, Department of Forensic Science, SGT University, Chandu Budhera, Gurugram-122505, Haryana, India.
E-mail: arunima_fosc@sgtuniversity.org
Introduction: During forensic investigations, cases are often encountered where the deceased bodies are in a partially or completely decomposed, charred, or skeletonised condition. In such scenarios, the examination of skeletal remains becomes imperative to establish the identity of an individual. Anthropometric measurements assist in identifying these parameters with accuracy.
Aim: To assess the relationship between upper limb dimensions and stature in North Indian adults using regression formulae and a decision forest-based model for stature estimation from these dimensions.
Materials and Methods: A cross-sectional study was conducted among 262 (M=120/F=142) students aged between 18 and 25 years using random sampling. This study was carried out in the Department of Forensic Science, Faculty of Applied and Basic Sciences, SGT University, Gurugram, Haryana, India from October 2023 to January 2024. The primary inclusion criteria stipulated that participants should be of North Indian origin (New Delhi NCR, Haryana and Punjab regions) and should not have suffered from any congenital or traumatic deformities of the upper and lower limbs. The stature, forearm length and Total Arm Length (TAL) were measured based on anthropometric points. Descriptive statistics, p-values, t-values and Pearson’s correlation coefficient were studied using Statistical Package of Social Sciences (SPSS) software version 21.0 and a decision forest model was designed on a cloud-based coding platform using Python programming language.
Results: The present study depicts a higher mean value of TAL (55.9±4.18) and stature (175.41±5.63) for males in comparison to females. All the measurements were statistically significant, with p-values <0.001. It also reveals a strong positive correlation between TAL and stature for both males (r-value=0.951) and females (r-value=0.975). The decision forest model achieved an accuracy of 0.951 and a Root Mean Square Error (RMSE) of 1.75.
Conclusion: The present study suggests that stature shows strong correlations with forearm length and TAL for both sexes. The decision forest model can classify the sexes with an accuracy of 77.5% using TAL. However, demographic variations must be considered when applying the regression formulae. Additionally, such anthropometric data should be updated regularly due to secular and temporal changes within the population.