Dynamic Laser Speckle Imaging Meets Machine Learning to Enable Rapid Antibacterial Susceptibility Testing (DyRAST)

Area of Science:

• Microbiology
• Biophysics
• Machine Learning

Background:

• Rapid antibacterial susceptibility testing (RAST) is crucial for effective antibiotic treatment.
• Current RAST methods face limitations such as cost, bulky equipment, and potential interference from reagents.
• Existing methods may not detect gradual bacterial adaptation to antibiotics, leading to misdiagnosis.

Purpose of the Study:

• To develop a novel RAST approach using machine learning and dynamic laser speckle imaging (DLSI).
• To accurately and rapidly predict the minimum inhibitory concentration (MIC) of antibiotics against *Escherichia coli*.
• To overcome the limitations of current RAST techniques, particularly in resource-limited settings.

Main Methods:

• Utilized time-resolved dynamic laser speckle imaging (DLSI) to capture changes in bacterial motion.
• Applied machine learning algorithms to analyze DLSI data for antibiotic response prediction.
• Trained and validated the algorithm using results from a gold standard antibacterial susceptibility testing method.

Main Results:

• Accurately predicted the MIC of ampicillin and gentamicin for *E. coli* K-12 in 60 minutes.
• Achieved comparable accuracy to traditional overnight methods but with a significant reduction in time.
• Demonstrated effectiveness for various antibiotic classes, including β-lactams and aminoglycosides.

Conclusions:

• The developed DLSI-based RAST method offers a fast and accurate alternative for determining antibiotic susceptibility.
• This approach has the potential to improve clinical decision-making and combat antibiotic resistance.
• The technique is suitable for resource-limited environments due to its potential for reduced cost and equipment size.