Description of original award (Fiscal Year 2020, $557,607)
Over 650,000 children are victims of abuse and neglect each year. Infants are particularly vulnerable and represent the age group most likely to experience a recurrence of maltreatment if the abuse is not detected. Child abuse cases, however, can be some of the most challenging cases for prosecutors, law enforcement professionals, and child protection advocates. No one other than the accused is typically present to witness the event, and children, especially infants, are too young to communicate what events led to their injuries. To confound matters, falls are the leading cause of non-fatal injury in infants, and the most common history provided in cases of child abuse. Therefore, distinguishing a truthful history of a fall from a false one proves to be a difficult, but important, task.
Linear skull fractures are common in both accidental and abusive head trauma, but little is known about the mechanics of skull fracture in infants. Further, it is unknown how fall characteristics affect skull fracture patterns. Recently, we used fundamental fracture mechanics to develop a fracture-simulation framework. This framework is capable of predicting infant skull fracture patterns from head impact and, thus, enables exploration of the effect of fall characteristics, such as fall height and impact angle, on skull fracture length and orientation.
This proposal will begin to translate the aforementioned fracture-simulation framework into a forensic tool capable of estimating fall height and impact location from images of skull fracture. A two-phase approach will be used to provide intermediary milestones within the project. In Phase 1, we will evaluate multiple supervised machine learning algorithms and identify the best one to estimate fall characteristics from images of fracture patterns. Concurrently, we will characterize the natural variability of skull thickness across the infant skull in a large dataset of infants. Using uncertainty quantification methods, we will then measure the effect of skull thickness distribution on skull fracture patterns in single events. In Phase 2, we will integrate the effects of skull thickness distribution into the machine learning algorithm and will evaluate the new tool against real-world cases of infant skull fracture from well-witnessed accidental falls.
Successful completion of this project will provide the medical, legal, and forensic communities with data to understand skull fracture variability among infants and produce an image analysis tool capable of improving accuracy in identifying child abuse cases. Note: This project contains a research and/or development component, as defined in applicable law, and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14). CA/NCF