Rapid alignment-free bacteria identification via optical scattering with LEDs and YOLOv8

Area of Science:

• Microbiology
• Optical Physics
• Computer Science

Background:

• Rapid bacterial identification is crucial for effective infection treatment, especially in sepsis.
• Traditional and molecular methods have limitations in speed, cost, or complexity.
• Optical scattering offers label-free bacterial detection but often requires complex equipment.

Purpose of the Study:

• To develop an enhanced, affordable, and user-friendly bacterial detection system.
• To improve the speed and accuracy of bacterial identification using optical scattering.
• To differentiate bacterial strains, including challenging pathogens like MRSA.

Main Methods:

• Utilized RGB light-emitting diodes (LEDs) as the light source for optical scattering.
• Captured and combined diffraction images from multiple LED colors using image registration.
• Employed the YOLOv8 object detection model for high-accuracy bacterial strain analysis.

Main Results:

• Achieved an average accuracy of 97% (mAP50 of 0.97) in bacterial identification.
• Successfully differentiated closely related bacterial strains.
• Accurately identified the significant pathogen Staphylococcus aureus MRSA 1320.

Conclusions:

• The enhanced RGB LED-based optical scattering system provides a rapid, label-free, and accurate method for bacterial identification.
• This approach offers advantages in cost-effectiveness, usability, and workflow integration.
• Presents a viable alternative for timely bacterial identification in clinical settings.