Deep Learning-Assisted Rapid Bacterial Classification Based on Raman Spectroscopy of Bacteria Lysed by Acoustically...

Area of Science:

• Analytical Chemistry
• Biotechnology
• Spectroscopy

Background:

• Accurate bacterial identification is crucial for clinical decisions and antibiotic resistance.
• Conventional methods using surface molecules have limitations in classification accuracy.
• Intracellular bacterial biomolecules offer richer spectral information but are masked by the cell envelope.

Purpose of the Study:

• To develop a novel method for enhanced bacterial classification.
• To improve Raman spectral information richness by exposing intracellular components.
• To achieve rapid and accurate pathogen identification for clinical applications.

Main Methods:

• Acoustofluidic lysis utilizing a vibrating fiber-tip to create a single-vortex within a capillary.
• Concentration of bacteria in high-shear regions for enhanced lysis efficiency.
• Raman spectroscopy combined with a deep learning residual neural network (ResNet) for automated classification.

Main Results:

• Effective exposure of intracellular components (nucleic acids, proteins, lipids) through acoustofluidic lysis.
• Significant enhancement in spectral resolution and information richness of bacterial Raman spectra.
• Achieved 98.9% classification accuracy across seven bacterial samples using the ResNet model, outperforming traditional classifiers.

Conclusions:

• The developed method significantly improves bacterial classification accuracy by analyzing intracellular biomolecules.
• This approach offers a promising advancement for rapid, label-free, on-site pathogen identification in clinical settings.
• The combination of acoustofluidic lysis, Raman spectroscopy, and deep learning facilitates faster and more effective clinical decision-making.