False negative probability in iGSR detection: a Bayesian approach

TL;DR

This paper presents a Bayesian model to estimate the probability of missing inorganic gunshot residue particles in SEM/EDS detection, showing that larger pixel sizes still maintain high detection rates, thus optimizing forensic analysis efficiency.

Contribution

It introduces a novel Bayesian approach to quantify detection probabilities in iGSR analysis and validates a particle detection model relating size to equipment settings.

Findings

Probability of missing all particles is below 5% for pixel sizes under 0.32 μm²

Larger pixel sizes up to double the standard still maintain effective detection rates

Using bigger pixels can significantly reduce laboratory workload without compromising detection quality

Abstract

This article discusses the detection of inorganic gunshot residue (iGSR) particles through scanning electron microscopy with energy dispersive X-ray spectrometer (SEM/EDS) in the discovery step. We calculated the probability that all characteristic inorganic gunshot residue particles (iGSR) go undetected and their dependence on the image pixel resolution setup. We built and validated a iGSR particle detection model that relates particle size with equipment registers, and we applied it to 1,174 samples analyzed by a forensic science laboratory. Our results indicate that the probability of missing all characteristic iGSR particles is below 5% for pixel sizes below $0.32 \mu m^{2}$. These values indicate that pixel sizes as great as the double in area that is commonly used in laboratory casework, $0.16 \mu m^{2}$, are effective for an initial scanning of a sample as it yields good rates of detection of characteristic particles, which might exponentially reduce laboratory workload.