Clinical evaluation of the APAS® Independence: Automated imaging and interpretation of urine cultures using artificial intelligence with composite reference standard discrepant resolution

Abstract

Background: This study reports the outcome of the first evaluation of the APAS® Independence for automated reading and preliminary interpretation of urine cultures in the routine clinical microbiology laboratory. In a 2-stage evaluation involving 3000 urine samples, two objectives were assessed; 1) the sensitivity and specificity of the APAS® Independence compared to microbiologists using colony enumeration as the primary determinant, and 2) the variability between microbiologists in enumerating bacterial cultures using traditional culture reading techniques, performed independently to APAS® Independence interpretation.

Methods: Routine urine samples received into the laboratory were processed and culture plates were interpreted by standard methodology and with the APAS® Independence. Results were compared using typical discrepant result resolution and with a composite reference standard, which provided an alternative assessment of performance.

Results: The significant growth sensitivity of the APAS® Independence was determined to be 0.919 with a 95% confidence interval of (0.879, 0.948), and the growth specificity was 0.877 with a 95% confidence interval of (0.827, 0.916). Variability between microbiologists was demonstrated with microbiologist bi-plate enumerations in agreement with the consensus 88.6% of the time.

Conclusion: The APAS® Independence appears to offer microbiology laboratories a mechanism to standardise the processing and assessment of urine cultures whilst augmenting the skills of specialist microbiology staff.

Keywords: APAS; Artificial intelligence; Automation; Discrepant; Imaging.