Healthy scepticism surrounds the use of automated systems to reliably detect and count colonies on an agar plate. Compendial guidelines covering the assessment of alternative microbiology methods are well established (e.g. USP<1223>) which provide measurements of method suitability. The appetite for technology interrogation, however, is high and driven by a desire to understand any limitations as part of risk profiling for an application. One of the most common concerns for automated colony counting systems used in Environmental Monitoring is the ability to detect colonies on the perimeter and labels on the plate, for which there are no established methods or acceptance criteria. When performing primary validation for the APAS Independence, this particular challenge test was developed as part of robustness examinations. Bespoke AI application tools were used for microbiologists to annotate colonies on the APAS Independence images which were then objectively and computationally compared to the APAS Independence result, on a colony-by- colony basis and stratified by location (edge, label, non-edge/non-label). This is a unique approach to a unique challenge test developed for this type of colony counting technology. The high level of colony detections by the APAS Independence underpins performance across all validation tests in a statistically defensible manner.

Poster Presentation: Clever Culture Systems

Conference: PDA Pharmaceutical Microbiology Conference 2024

Date: October 2024

Authors: Giglio, S

Citation: Giglio, S. 2024. Environmental Monitoring: Developing tests to challenge artificial intelligence models for colony counting using the APAS® Independence. PDA Micro 2024. Washington.