Advancements in AI-driven drug sensitivity testing research

Abstract

Antimicrobial resistance (AMR) constitutes a significant global public health challenge, posing a serious threat to human health. In clinical practice, physicians frequently resort to empirical antibiotic therapy without timely Antimicrobial Susceptibility Testing (AST) results. This practice, however, may induce resistance mutations in pathogens due to genetic pressure, thereby complicating infection control efforts. Consequently, the rapid and accurate acquisition of AST results has become crucial for precision treatment. In recent years, advancements in medical testing technology have led to continuous improvements in AST methodologies. Concurrently, emerging artificial intelligence (AI) technologies, particularly Machine Learning(ML) and Deep Learning(DL), have introduced novel auxiliary diagnostic tools for AST. These technologies can extract in-depth information from imaging and laboratory data, enabling the swift prediction of pathogen antibiotic resistance and providing reliable evidence for the judicious selection of antibiotics. This article provides a comprehensive overview of the advancements in research concerning pathogen AST and resistance detection methodologies, emphasizing the prospective application of artificial intelligence and machine learning in predicting drug sensitivity tests and pathogen resistance. Furthermore, we anticipate future directions in AST prediction aimed at reducing antibiotic misuse, enhancing treatment outcomes for infected patients, and contributing to the resolution of the global AMR crisis.

Keywords: antimicrobial resistance; antimicrobial susceptibility testing; artificial intelligence; machine learning; whole genome sequencing.

Figure 1

AI-assisted Antimicrobial Susceptibility Trials (AST) enhance clinical decision-making by integrating and analyzing multimodal data. After pretreating clinical samples, phenotypic and genotypic data, along with patient information, are collected. The AI model processes this data to predict drug resistance and suggest personalized treatment plans. If results are contradictory, the model updates using transfer learning. Traditional AST, however, takes 16–24 hours, relying on sample culture, drug sensitivity tests, and manual result interpretation.