PalAST: A Cross-Platform Mobile Application for Automated Disk Diffusion Antimicrobial Susceptibility Testing

Abstract

Antibiotic resistance is the ability of bacteria to resist the effects of antibiotics, making infections more difficult to treat and increasing the risk of complications and death. One way to fight antibiotic resistance is by identifying the most effective antibiotics for treating bacterial infections. This can be done through a laboratory test called AST, which is used to determine the susceptibility of bacteria to antibiotics. However, manual AST has several limitations that include time delay, limited accuracy, limited testing capacity, and subjective interpretation of results. Therefore, there is an emergent need for a more reliable and efficient alternative to manual AST. Recently, few works have tried to automate disk diffusion AST through AI-based solutions and mobile applications. However, these works do not support advanced analysis and interpretation of results, do not present evaluation of detection performance, or are not publicly available to download and use. This work proposes PalAST, a cross-platform mobile application that supports automated disk diffusion AST. The application enables biologists to take AST photos and analyze them in real time with minimal human intervention. It uses image processing and a pre-trained machine learning model to detect antibiotic disks in the agar plate and predict bounding circles for inhibition zones. Then, it provides an interpretation of results including the diameters of the inhibition zones, the labels on the antibiotic disks, and the rating of the bacteria as susceptible, intermediate, or resistant to each antibiotic. PalAST also stores the results of tests, allowing users to access and review past test results. PalAST was tested using a number of real AST photos, and the detection performance was evaluated by using common metrics, i.e. precision, recall, and Intersection over Union. We also used expert evaluation through a questionnaire to assess the usability and ease of use of PalAST.