Age estimation from the immature pars lateralis using decision tree analysis.

Age-at-death estimation is instrumental to the medicolegal investigation of unknown fetal and infant remains due to a large number of such remains received by medico-legal laboratories in South Africa. The maturation and growth of immature bones serves as an alternative to soft tissue and dental age estimation. The development of bones such as the pars lateralis serves as one of the possible ways to estimate age in immature individuals. The aim of this study was to test previously described osteomorphological descriptors of the pars lateralis associated with early prenatal, late prenatal and early postnatal development in relation to estimating age. Data from a sample of 99 human pars laterali originating from the Johannesburg Forensic Paediatric Collection (JFPC), University of the Witwatersrand was utilized in the construction of decision trees with the use of the WEKA (version 3.9.3) software for the purposes of forensic age estimation of skeletal remains. The study sample was subdivided into early prenatal (younger than 30 gestational weeks; n = 29), late prenatal (30–40 gestational weeks, n = 40) and early postnatal (birth to 7.5 months, n = 30) age groups. A pruned tree algorithm (J48) was used to construct five decision trees and included four morphological variables associated with development the pars lateralis. Overall, 85.9% of all individuals were correctly classified using the combination of all four variables. Features associated with specific areas on the bone displayed a range of classification success (jugular limb – 78.8%; shape of bone – 75.8%; medial border of foramen magnum – 64.7%; hypoglossal canal – 63.6%). This study highlights the supportive role of the pars lateralis for forensic age estimation. Furthermore, machine learning algorithms provide quantitative confirmation of proposed morphological age predictors applicable to biological profiling of immature remains.