Multidimensional foot measurements-based sex estimation for Taiwanese population in foot length overlapping interval

Estimating the sex from unknown bodies or fragmented human remains is essential for forensic identification. For sex estimation from foot features, previous studies highly depend on the foot length, resulting in a single-dimensional threshold and thusly poor performance for data falling into the foot length overlapping interval. Moreover, available information about the sex estimation from foot dimensions among Taiwan population is lacking. Hence, this study proposes a novel method to achieve multidimensional foot measurements-based sex estimation for the human falling into the overlapping foot interval. In total, 780 men and 670 women with same foot lengths ranging from 23.0 to 25.5 cm were recruited. A three-dimensional foot scanner was used to measure 10 dimensions of both feet. Bilateral differences were detected in several foot dimensions under the same foot length categories. The support vector machine method provided better results in terms of accuracy (86.05%), recall (86.44%), specificity (85.59%), and precision (83.77%). This study indicates the importance of other variables including foot length, ball girth, instep girth, ball of foot length, and outside ball of foot length in sex estimation. The findings will provide helpful information for forensic practice for Taiwanese and extend the application range of machine learning methods.