AI System Rapidly Shortlists Potential Matches in Forensic Radiograph Analysis

An interdisciplinary team from Michigan State University (MSU), including faculty and doctoral students from the Department of Anthropology and Computer Science and Engineering, has developed a groundbreaking AI system to help forensic anthropologists identify individuals more quickly and accurately. This system, co-authored in a study published in *IEEE Access*, utilizes deep neural networks to analyze chest radiographs, a common forensic tool, by pinpointing different regions of interest (ROIs) that aid in person identification. The team, comprising Dr. Carolyn Isaac, Dr. Todd Fenton, Dr. Joseph Hefner, and doctoral student Alexis VanBaarle, utilized over 5,000 chest radiographs to train and test their AI model. These radiographs, which capture the unique anatomy of the chest, are crucial in forensic investigations, especially in mass fatality situations where thousands of individuals may need to be identified. Dr. Isaac emphasized the significant time-saving potential of this AI system. "In mass fatality situations, the system can assist by short-listing potential matches for a practitioner to visually assess," she explained. "It can analyze more than 1,800 radiographs in just 17 seconds, a task that would typically take a human practitioner 30 to 60 hours." This dramatic reduction in time not only speeds up the identification process but also helps in reducing practitioner bias, ensuring a more objective and accurate assessment. The AI approach is particularly innovative because it focuses on evaluating different ROIs within radiographs, a feature that has not been previously explored in this context. "This research shows how AI can be used to enhance forensic casework by making tasks more efficient," Isaac said. The deep neural networks compare a target radiograph to thousands of others to find the most likely matches, a task that forensic anthropologists have traditionally performed manually and which can be labor-intensive and prone to errors. Dr. Arun Ross and Redwan Sony from the iPROBE Lab in Computer Science and Engineering contributed to the development of the AI system. Isaac highlighted the value of the collaboration, noting that the unique perspectives of computer science and forensic anthropology were essential in creating a practical and effective tool. "I love when we are brainstorming on the project and get to see the unique perspectives of computer science versus the domain experts in forensic anthropology," she said. The potential applications of this AI system extend beyond mass fatality situations. It can also be integrated into unidentified or missing person databases to propose potential matches for consideration, thereby aiding in the resolution of cold cases and missing persons investigations. By automating the initial stages of the identification process, practitioners can focus on more detailed and critical aspects of their work, leading to more efficient and accurate outcomes. In forensic casework, chest radiographs are used to identify individuals by focusing on anatomical features such as the shape and size of the ribs, the sternum, and the lungs. The AI system is designed to recognize these features and match them with known radiographs, significantly streamlining the process. This technology could revolutionize how forensic teams handle large-scale identification tasks, making their work less daunting and more effective. Industry insiders and experts in the field of forensic science have praised this development, noting its potential to transform the way forensic anthropologists operate. The collaboration between MSU's Anthropology and Computer Science and Engineering departments exemplifies the power of interdisciplinary research in solving complex problems. The MSU Forensic Anthropology Lab, known for its cutting-edge research and contributions to the field, continues to lead the way in advancing forensic science through innovative technologies.