Researchers from the University of Oxford have created a cutting-edge AI-driven tool based on physics to support forensic and law enforcement traumatic brain injury (TBI) investigations. The results were published in Communications Engineering.
Image Credit: Teeradej/Shutterstock.com
TBI has serious, long-lasting neurological effects, making it a serious public health concern. Determining whether an impact might have caused a reported damage is essential for legal processes in forensic investigations, but there is not a standardized, quantitative method for doing so at the moment.
The latest study shows how machine learning methods informed by mechanistic simulations could produce evidence-based harm predictions. Based on recorded assault scenarios, this would assist law enforcement and forensic teams in making precise predictions about TBI consequences.
The study’s AI framework demonstrated exceptional prediction accuracy for TBI-related injuries after being trained on actual, anonymized police reports and forensic data:
- 94% accuracy for skull fractures
- 79% accuracy for loss of consciousness
- 79% accuracy for intracranial hemorrhage (bleeding within the skull)
The model demonstrated good sensitivity and specificity (a low rate of false positive and false negative outcomes) in every instance.
Using a broad computational mechanistic model, the framework simulates the effects of numerous impacts, including punches, slaps, and strikes against a flat surface, on different regions of the head and neck. This offers a simple forecast of the likelihood of tissue stress or deformation following an impact. Nevertheless, it cannot independently forecast any risk of harm. An upper AI layer combines this data with any other pertinent metadata, such as the victim's height and age, to provide a prognosis for a particular injury.
This research represents a significant step forward in forensic biomechanics. By leveraging AI and physics-based simulations, we can provide law enforcement with an unprecedented tool to assess TBIs objectively.
Antoine Jérusalem, Study Lead Researcher and Professor, Department of Engineering Science, University of Oxford
The researchers trained the overall framework using 53 anonymized genuine police records of assault instances. Each report included information regarding various parameters that could influence the severity of the blow (for example, the victim's or offender’s age, gender, and body type). This produced a model capable of combining mechanical biophysical data with forensic facts to forecast the risk of various injuries.
When the researchers determined which elements influenced the prediction value for each type of injury, the results were strikingly similar to medical findings. For example, when estimating the chance of skull fracture, the most relevant component was the maximum amount of force sustained by the scalp and skull during an impact. Similarly, brainstem stress measurements were the best predictor of loss of consciousness.
The research team cautions that the model is not meant to take the role of real forensic and clinical professionals in analyzing assault cases. Instead, the goal is to provide an objective assessment of the likelihood that a documented assault was the genuine cause of a reported injury. The model could be used to identify high-risk situations, improve risk assessments, and devise preventive techniques for lowering the frequency and severity of head injuries.
Understanding brain injuries using innovative technology to support a police investigation, previously reliant on limited information, will greatly enhance the interpretation required from a medical perspective to support prosecutions.
Ms Sonya Baylis, Senior Manager at the National Crime Agency
“Our framework will never be able to identify without doubt the culprit who caused an injury. All it can do is tell you whether the information provided to it is correlated with a certain outcome. Since the quality of the output depends on the quality of the information fed into the model, having detailed witness statements is still crucial,” Jérusalem added.
Dr Michael Jones, Researcher at Cardiff University and Forensics Consultant, added, “An “Achilles heel” of forensic medicine is the assessment of whether a witnessed or inferred mechanism of injury, often the force, matches the observed injuries. With the application of machine learning, each additional case contributes to the overall understanding of the association between the mechanism of cause, primary injury, pathophysiology and outcome.”
A multidisciplinary team of engineers, forensic specialists, and medical professionals from the University of Oxford, Thames Valley Police, the National Crime Agency, Cardiff University, Lurtis Ltd., the John Radcliffe Hospital, and other institutions collaborated on the study.
Journal Reference:
Wei, Y. et. al. (2025) A mechanics-informed machine learning framework for traumatic brain injury prediction in police and forensic investigations. Communications Engineering. doi.org/10.1038/s44172-025-00352-2