Open-set deep learning-enabled single-cell Raman spectroscopy for rapid identification of airborne pathogens in...

Area of Science:

• Microbiology
• Spectroscopy
• Artificial Intelligence

Background:

• Pathogenic bioaerosols pose significant risks for airborne disease outbreaks.
• Accurate and rapid identification of airborne pathogens in complex environments is a major challenge.

Purpose of the Study:

• To develop an advanced method for identifying airborne pathogens in real-world air using open-set deep learning (OSDL) and single-cell Raman spectroscopy.
• To enhance pathogen identification accuracy and reduce false positives compared to conventional methods.

Main Methods:

• Constructed Raman datasets of aerosolized bacteria for testing and enhancement.
• Optimized OSDL algorithms and training strategies for pathogen identification.
• Utilized single-cell Raman spectroscopy for high-resolution analysis of airborne particles.

Main Results:

• Achieved 93% accuracy for five target airborne pathogens and 84% accuracy for untrained air bacteria.
• Reduced false positive rates by 36% compared to close-set algorithms.
• Demonstrated high detection sensitivity down to 1:1000 and simultaneous identification of multiple pathogens within an hour.

Conclusions:

• The developed Raman-OSDL method accurately identifies airborne pathogens and unknown bacteria in complex air environments.
• This single-cell tool significantly advances rapid surveillance of pathogens to prevent infection transmission.
• The method offers a sensitive and rapid solution for real-time monitoring of bioaerosols.