The application of TreadMatch scans to aid the process of footwear mark comparison

Abstract

Currently in the UK, if a person is arrested or charged with a recordable offence, they can have prints of their footwear taken whilst in custody. The tread pattern recorded in these prints can be searched for using the National Footwear Database to find out if the same footwear pattern has been recovered at previous crime scenes, generating forensic intelligence.TreadMatch is a digitised system for collecting footwear prints seized from detainees in custody for this purpose. Whilst its use for generating intelligence is accepted, validation experiments have not been conducted to understand its level of performance in assisting in forensic comparison purposes for identification, because in the absence of an incorporated scale, it is not known how well TreadMatch reproduces the pattern size of a tread, threatening the validity of the comparison. If it can be determined that the measurements of TreadMatch scans are consistent with the more commonly used aluminium powder test marks, this could save Police time and resources if the digital scans could be used for preliminary assessment prior to the footwear being physically submitted for evidential comparison. Therefore, this study set out to compare three different types of TreadMatch scans (‘dynamic’, ‘zoomed’ and ‘rolled’) for thirty different pieces of footwear, with test marks of the same footwear using the traditional method (fingerprint powder). Length and width measurements were obtained from each tread pattern using GNU Image Manipulation Program software. The resulting data were analysed to assess for agreement between TreadMatch scans and test marks using 95% Intraclass Correlation Coefficients (ICC) and 95% Bland-Altman plots of Limits of Agreement (LOA). Additionally, an intra-sample study using fifteen repeated measurements of the same piece of footwear for different TreadMatch scanning methods was carried out to support the larger validity study.95% ICC3,1 resulted in coefficients ranging from 0.99 to 1.00 across all measurements. 95% LOA displayed close agreement. There was less agreement and more variation displayed between the test marks and the TreadMatch rolled scans for both length and width measurements. This variation for hand-rolled prints must be taken into consideration and a standard approach developed. The study suggests TreadMatch can be used for preliminary assessments in assisting forensic comparisons, particularly for dynamic and zoomed prints.