Application of deep learning for detecting implants in computed tomography scout images with multi-institution and multi-vendor for personal identification.

The identification of deceased individuals is essential in forensic investigations, particularly when primary identification methods such as odontology, fingerprint, or DNA analysis are unavailable. In such cases, implanted medical devices may serve as supplementary identifiers for positive identification. This study proposes deep learning-based methods for the automatic detection of metallic implants in scout images acquired from computed tomography (CT). Using a multi-institutional and multi-vendor dataset, two object detection models, RetinaNet and Faster R-CNN, were trained and evaluated to ensure generalizability across diverse imaging conditions. The proposed models achieved improved performance in detecting various types of implants, significantly reducing false positives and improving classification consistency. These findings emphasize the potential of CT scout image analysis as a practical tool for supporting forensic identification through the efficient detection of implanted medical devices.