Combined interpretation of objective firearm evidence comparison algorithms using Bayesian networks.

Traditionally, firearm and toolmark examiners manually evaluate the similarity of features on two bullets using comparison microscopy. Advances in microscopy have made it possible to collect 3D topographic data, and several automated comparison algorithms have been introduced for the comparison of bullet striae using these data. In this study, open-source approaches for cross-correlation, congruent matching profile segments, consecutive matching striations, and a random forest model were evaluated. A statistical characterization of these automated approaches was performed using four datasets of consecutively manufactured firearms to provide a challenging comparison scenario. Each automated approach was applied to all samples in a pairwise fashion, and classification performance was compared. Based on these findings, a Bayesian network was empirically learned and constructed to leverage the strengths of each individual approach, model the relationship between the automated results, and combine them into a posterior probability for the given comparison. The network was evaluated similarly to the automated approaches, and the results were compared. The developed Bayesian network classified 99.6% of the samples correctly, and the resultant probability distributions were significantly separated more so than the automated approaches when used in isolation.

Characterization of fracture match associations with automated image processing.

During the examination of trace evidence, often a realignment along the edges of known and questioned items are made to determine if a physical fit is present and if those objects were once one continuous piece or object. Duct tape is an evidence type in which the evaluation of physical fits is often conducted and is regarded as conclusive evidence of an association between the items. The examination and conclusion of a physical fit between edges relies heavily on examiner discretion to identify distinctive features across the edges since there are no statistical approaches or objective methodologies for the comparison. This study developed an automated image processing and comparison method to quantify tape end matches using cross-correlation scores and an empirical approach to the assessment. Characterization of 150 hand torn duct tape end pair physical fits were also conducted where matching and non-matching sample distributions were created. This study also evaluated partial duct tape edges and the influence this has on a comparison. Given the strength associated with a physical fit and the presence of stretching or deformation along the fractured edge, an understanding of the value these samples have is paramount. Furthermore, random match probabilities were calculated based on the correlation scores from the inter-comparisons to model the weight of evidence or strength of association between the edges. Finally, the study demonstrated that not every true match holds the same association strength through score distributions, but the approach is able to distinguish matching and non-matching samples at edge widths greater than 27 %.

The use of error rates to predict and characterize the efficacy of presumptive field tests.

From a forensic perspective, a presumptive test, one which indicates the presence or absence of a certain target material such as blood, is an invaluable tool. Among these tests, there are different specificities, sensitivities, and shelf lives. The accuracy of a test is an algebraic combination of the specificity and sensitivity of the test. Each test has limitations as given by its false positive and false negative rates. The aim of this study was to illustrate how the false positive and false negative rates are to be properly determined using a simulation study for the phenolphthalein test. New presumptive tests must be properly evaluated/validated through testing of commonly encountered household items and other potentially probative items usually found at crime scenes, however, the makeup of test sets must appropriately capture all error rates. In order to correctly use these results when the test is applied to an unknown sample recovered at a crime scene, the error rates cannot be applied directly to estimate whether or not the sample is actually the analyte of interest. In a validation study, the forensic scientist calculates the false positive rate as the p(Positive Reaction|Blood), whereas at the scene, the crime scene investigator wishes to determine the p(Blood|Positive Reaction). All crime scene investigators need to ensure that the conditional is not transposed when interpreting such results. Furthermore, this work provides a model for the assessment of a multiple test diagnostic system intended for investigators.