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

Rapid and accurate bacterial identification is essential for timely treatment of infections like sepsis. While traditional methods are reliable, they lack speed, and advanced molecular techniques often suffer from cost and complexity. The bacterial detection technology based on optical scattering system offers a rapid, label-free alternative but traditionally relies on complex lasers and analysis. Our enhanced approach utilizes RGB light emitting diodes (LEDs) as the light source. Three diffraction images of bacterial colonies from different LED colors are separately captured by a USB camera and combined using an image registration algorithm to enhance image sharpness. Our approach utilizes an object detection model, i.e., YOLOv8, for analysis achieving high-accuracy differentiation between bacterial strains. We demonstrate the effectiveness of this approach, achieving an average accuracy of 97% (mAP50 of 0.97), including accurate discrimination of closely related strains and the significant pathogen Staphylococcus aureus MRSA 1320. Our enhancement offers advantages in affordability, usability, and seamless integration into existing workflows, providing an alternative for rapid bacterial identification.