AssiST: convolutional neural network for analysis of antibiotic susceptibility testing

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Antibiotic susceptibility testing (AST) is crucial for guiding antimicrobial therapy.
• Current AST methods can be labor-intensive and time-consuming.
• There is a need for scalable and reproducible methods for drug susceptibility determination.

Purpose of the Study:

• To introduce AssiST, a convolutional neural network (CNN) pipeline for automated antibiotic susceptibility testing.
• To enable classification of microbial growth in 96-well broth microdilution plates from scanned images.
• To provide a flexible and scalable solution for inferring drug susceptibility.

Main Methods:

• Development of a CNN pipeline (AssiST) for image-based microbial growth classification.
• Utilizing scanned images of 96-well broth microdilution plates.
• Implementing user-configurable mapping for phenotype to susceptibility calls.

Main Results:

• AssiST accurately classifies microbial growth patterns indicative of drug susceptibility.
• The pipeline accommodates diverse microbial morphologies and experimental conditions.
• AssiST enables the conversion of scanner images into reproducible drug sensitivity readouts.

Conclusions:

• AssiST offers a scalable, automated, and reproducible method for antibiotic susceptibility testing.
• The pipeline supports flexible application across various microorganisms, media, and drugs.
• AssiST empowers laboratories to generate drug sensitivity data efficiently using standard hardware.