False-negative probability in the SEM/EDS automated discovery of iGSR particles: A Bayesian approach

• Published in: Journal of forensic sciences 2023-09, Vol.68 (5), p.1792-1799
• Main Authors: Onetto, Martín A, Carignano, Edgardo, Pregliasco, Rodolfo G
• Format: Article
• Language: English
• Subjects: Automation; Bayesian analysis; Electron microscopes; Forensic science; Organic matter; Pixels; Scanning electron microscopy
• ISSN: 0022-1198; 1556-4029
• DOI: 10.1111/1556-4029.15323

The automated search software integrated with a scanning electron microscope (SEM/EDS) has been the standard tool for detecting inorganic gunshot residues (iGSR) for several decades. The detection of these particles depends on various factors such as collection, preservation, contamination with organic matter, and the method for sample analysis. This article focuses on the influence of equipment resolution setup on the backscattered electron images of the sample. The pixel size of these images plays a crucial role in determining the detectability of iGSR particles, especially those with sizes close to the pixel size. In this study, we calculated the probability of missing all characteristic iGSR particles in a sample using an SEM/EDS automated search and how it depends on the image pixel resolution setup. We developed and validated an iGSR particle detection model that links particle size with equipment registers and applied it to 320 samples analyzed by a forensic science laboratory. Our results show that the probability of missing all characteristic iGSR particles due to their size is below 5% for pixel sizes below 0.32 μm . These findings indicate that pixel sizes as large as twice the one commonly used in laboratory casework, that is, 0.16 μm , are effective for initial sample scanning, yielding good detection rates of characteristic particles that could exponentially reduce laboratory workload.