Optical forward-scattering for identification of bacteria within microcolonies

Area of Science:

• Microbiology
• Biophotonics
• Machine Learning

Background:

• Rapid identification of pathogenic bacteria is crucial for clinical diagnosis and safety control.
• Current methods often require sample preparation or labeling, and can be destructive.
• Optical methods using light scattering offer a non-invasive, label-free alternative for bacterial identification.

Purpose of the Study:

• To evaluate the potential of light forward-scattering combined with machine learning algorithms for early-stage bacterial species and strain discrimination.
• To assess the method's efficacy on thin agar layers after short incubation periods.
• To determine the feasibility of using this technique in a clinical diagnostic setting.

Main Methods:

• Bacterial colonies were grown on thin agar layers (1 mm Tryptic Soy Agar) at 37°C for 6 hours.
• Light forward-scattering patterns (scatterograms) were acquired from bacterial colonies.
• Machine learning algorithms (Bayes Network, Continuous Naive Bayes, Sequential Minimal Optimisation) were employed for data analysis and classification.
• Discrimination was tested on gram-negative strains, Escherichia coli strains, and coagulase-negative Staphylococci species on commercial media.

Main Results:

• Nearly 80% recognition rate for 7 gram-negative strains after 6 hours of incubation.
• 82% recognition rate for discriminating four strains of Escherichia coli.
• Successful discrimination of Staphylococcus haemolyticus and Staphylococcus cohnii on a commercial diagnostic medium without opening the lid.

Conclusions:

• Light forward-scattering combined with machine learning enables rapid, non-invasive, and accurate identification of bacterial species and strains at an early growth stage.
• The method shows promise for automated microbiology laboratories, providing early clinical information and enhancing food/water safety.
• The non-invasive nature prevents cross-contamination and minimizes sample handling, making it suitable for clinical diagnostic workflows.