Is the (hu)MANid Program Valid in a Pediatric Population.pdf

2020-05-30T04:32:03Z (GMT) by NATALIE R TOMÁS
This study aims to determine whether the (hu)MANid forensic identification program can accurately identify sex and ancestry of older children and adolescents, using measurements from Cone Beam Computer Tomography (CBCT) scans from the diverse pediatric population seen at the University of Illinois College of Dentistry (UIC COD). The (hu)MANid program was established in 2017 by Berg and Kenyhercz. It uses morphometroscopic analysis to classify adults mandibles based on ancestry, age, and sex.
One hundred and ten eligible CBCTs were used in our sample. Our sample included patients ages 8-17 with a CBCT taken between January 2008 and January 2019 and race/ethnicity recorded. Subjects were categorized to ancestry groups matching (hu)MANid ancestry designations. Patients that identified as non-Hispanic White/Caucasian were of White ancestry. Patients that identified as Hispanic were of Hispanic ancestry and lastly, non-hispanic Black/African- Americans were considered Black ancestry. 3D Slicer software (BHW, Boston, MA, USA) was used to measure morphoscopic and morphometric data and input into the (hu)MANid program to estimate the likelihood that an individual is a certain sex and ancestry.
Results show low overall prediction accuracy of the (hu)MANid program in our pediatric population. The program provides an estimated likelihood that an individual is M/F and White/Hispanic/Black, identifying a most-likely sex/ancestry assignment White (n= 29) subjects were most likely to be assigned to the correct group, followed by Black (n=32), and then Hispanic (n=45). The highest percent likelihood score an individual received for ancestry was slightly more likely to match that individual’s actual ancestry
1
(p<0.05) than either of the other two groups. Accuracy of ancestry estimation showed no appreciable changes when older adolescents (>14 y.o. girls, >16 y.o. boys) were examined separately.
The application struggled to predict both sex and ancestry for a given individual. For ancestry prediction only, Whites were identified at a rate consistent with prior literature. For sex prediction only, females reached this threshold. In contrast with our expectations, estimations did not appreciably improve when older adolescents (>14 y.o. girls, >16 y.o. boys) were examined separately. When predicting ancestry, (hu)MANid was slightly more likely to assign an individual to the correct group than to either incorrect group, but the frequency of correct assignment is not forensically valuable except for in the White sample. In our diverse pediatric sample, we were unable to validate the (hu)MANid application for ancestry or sex prediction in older children and adolescents.