Artificial intelligence (AI)-assisted antimicrobial susceptibility testing (AST): Automated imaging and the use of artificial intelligence for interpretation of disc-diffusion AST according to EUCAST, including rapid antimicrobial susceptibility testing (RAST)

Disc diffusion remains an essential part of AST and EUCAST has recently expanded its guidelines to include breakpoints for rapid disc diffusion reading at 4, 6 or 8 hours. Manual reading and interpretation of disc diffusion zone sizes is labour intensive, time consuming, and associated with other problems of manual methods such as dependence on the experience and skill of the performer. Automated reading and zone-size interpretation can eliminate these problems and greatly enhance the value of disc diffusion for routine AST.

A total of 80 organisms sourced from urine cultures which included E. coli, K. pneumoniae, P. aeruginosa, and S. aureus were processed and interpreted according to EUCAST methodology and breakpoints, including a rapid read at 6 hours, which generated over 740 unique antimicrobial reads. APAS-based AI-assisted reads and interpretations were compared with manual reads by two experienced microbiologists at 6 and 24 hours with regard to the categorical agreement (S/I/R, EUCAST) between the manual plate-in-hand (PIH) and APAS-AI-assisted (AAA) result.

The study demonstrated AAA yielded comparable results to PIH and provides a promising alternative to PIH for disc diffusion for standard and Rapid AST-EUCAST breakpoint reading. Differences in measurement between ruler, APAS-AMR, and on-screen measurement occurred. This demonstrated variability in zone size interpretation across methods. APAS-AMR has the opportunity to standardise disc diffusion testing with high throughput and accuracy.

Oral Presentation: by Labor Dr Wisplinghoff

Conference: ECCMID 2022, Lisbon

Date: July 2021

Authors: Giglio, S., Hill, R., Jazmati, N., Nowag, A., Radder, L., Berndsen, C. and Wisplinghoff, H.

Citation: Giglio, S., Hill, R., Jazmati, N., Nowag, A., Radder, L., Berndsen, C. and Wisplinghoff, H. 2022. Artificial intelligence (AI)-assisted antimicrobial susceptibility testing (AST): automated imaging and the use of artificial intelligence for interpretation of disc-diffusion AST according to EUCAST, including rapid antimicrobial susceptibility testing (RAST). ECCMID 2022. Lisbon.

Read Abstract