Leveraging Laser‐Induced Breakdown Spectroscopy and Machine Learning Methods for Rapid Detection of AMR Profiles in Pathogenic Bacterial Isolates

Rapid antimicrobial resistance (AMR) detection is critical for clinical treatment efficacy and infection control. We present a Laser-Induced Breakdown Spectroscopy (LIBS) method for antimicrobial susceptibility testing and detection of various resistance profiles in pathogenic bacteria. A controlled growth environment is utilized for regulated culture of different bacterial strains which induces strain-specific adaptation under controlled nutrient stress and also minimizes spectral variability due to differences in media composition and bacterial growth phases. By leveraging minimal yet robust spectral features (emission lines from Sodium, Potassium and Calcium), this method enables accurate AMR detection by employing machine learning algorithms. An accuracy of 94.7% and ROC-AUC > 0.99 was achieved using a Support Vector Machine model for classifying seven bacterial strains (3 susceptible and 4 resistant), without requiring outlier filtering or averaging of spectra. This tailored LIBS framework establishes a quick and cost-effective diagnostic tool for rapid AMR profiling compared to conventional culture-based susceptibility testing.