Ai-Based Rapid Identification of Multidrug-Resistant Bacteria Using Imaging and Genomic Sequencing

Abstract

Aim: To evaluate the effectiveness of artificial intelligence (AI) for the rapid identification of multidrug-resistant (MDR) bacteria using imaging techniques and genomic sequencing, and to compare its diagnostic accuracy with conventional antibiotic susceptibility testing.

Materials and methods: This prospective laboratory-based study was carried out over 12 months to evaluate the utility of artificial intelligence (AI) for rapid detection of multidrug-resistant (MDR) bacteria using imaging and whole-genome sequencing (WGS). A total of 120 non-duplicate clinical isolates were collected from blood, urine, sputum, and wound swabs of patients admitted to a tertiary-care hospital. Standard culture and biochemical tests were used for initial bacterial identification before AI analysis.

Results: In our study, AI-based approaches showed high accuracy for rapid identification of multidrug-resistant (MDR) bacteria. The imaging AI achieved an accuracy of 86.2%, while genomic AI performed better at 93.5%; when combined, sensitivity, specificity, and overall accuracy increased to 97.1%, 95.8%, and 96.4%, respectively, closely matching conventional AST results. Genomic profiling further revealed predominant resistance genes, including blaCTX-M, tetA, and sul1 in E. coli; blaNDM, blaKPC, and oqxA in K. pneumoniae; mexA, blaVIM, and oprD in P. aeruginosa; and mecA, ermC, and tetK in S. aureus, with AI predictions showing over 92% concordance across all species.

Conclusion: AI-powered diagnostics represent a vital advancement in the fight against antimicrobial resistance, offering faster, more precise, and actionable insights into multidrug-resistant infections.