Infrared Analysis of Urinary Stones: a Trial of Automated Identification

Abstract

A Search algorithm included in the Opus software of Bruker (Germany) was evaluated for analysis of urinary stones. Three reference libraries containing respectively 85 (single components), 1,059 (binary mixtures) and 4,565 (ternary mixture) digitized spectra were created and used to identify unknown spectra (n=320), applying the automatic procedure. Identification of the major component was correct in 83% of cases but the percentage of identification significantly decreased for the second and the third components. In cases of identification of the two first components, quantitative assessment was correct within tolerance limits ± 15%.

The computer results are judged unsatisfactory with regard to pathology because computer-aided identification is not sufficiently sensitive and specific to differentiate species with similar spectral pattern, even for the identification of main component, and also to detect minor components. It can be of assistance to guide spectral analysis, but it cannot replace human identification.