Correlation of cadaveric stature with posterior curve length of sternum in Mewat Region of Haryana: A comparative analysis of SPSS with machine learning

Abstract

The aim of the present study was to develop a population specific regression formula for estimating cadaveric length from posterior  curve length of fresh sternum and compare the linear regression results of SPSS and machine learning. Cadaveric length and posterior  curve length (PCL) of the sternum were measured from 74 dead bodies (39 males and 35 females) aged between 18 and 95 years from  known corpses during medico-legal autopsies. Cadaveric stature and PCL was greater in males as compared to females (p<0.001).  Regression equations and correlation coefficients were derived for PCL by SPSS and machine learning, which were  Y=122.79+2.504(PCL) and Y= 122.98+2.49(PCL), respectively, with correlation coefficients of 0.609 and 0.606, respectively.  Individual regression equations were also formulated for males and females separately with significant correlation. The standard error  of estimate and R square model were also derived. Cross validating linear regression results of SPSS with machine learning showed  almost similar results. The study suggests that posterior curve length of sternum in relation with post-mortem or cadaveric stature  shows regional or geographic variation, a moderately positive correlation and relatively low reliability in estimating cadaveric stature,  and thus, has limited forensic value.