Rapid identification of pathogens by using surface-enhanced Raman spectroscopy and multi-scale convolutional neural...

Area of Science:

• Analytical Chemistry
• Microbiology
• Machine Learning

Background:

• Salmonella is a major global pathogen responsible for significant illness and death.
• Over 2600 Salmonella serovars exist, with Salmonella Enteritidis, Salmonella Typhimurium, and Salmonella Paratyphi being common foodborne pathogens.
• Accurate and efficient detection methods are crucial for public health.

Purpose of the Study:

• To develop an efficient analytical method for detecting and distinguishing common Salmonella serovars.
• To combine surface-enhanced Raman spectroscopy (SERS) with a multi-scale convolutional neural network (CNN) for Salmonella serotype identification.

Main Methods:

• Preparation of 34-nm gold nanoparticles (AuNPs) as label-free Raman substrates.
• Measurement of 1854 SERS spectra from Salmonella Enteritidis, Salmonella Typhimurium, and Salmonella Paratyphi.
• Development of a multi-scale CNN model for multi-dimensional SERS spectral feature extraction.

Main Results:

• The multi-scale CNN model achieved a recognition accuracy exceeding 97%.
• The study analyzed the impact of training iterations and sample size on recognition accuracy.
• Experimental data validated the model's high performance.

Conclusions:

• The combined SERS and multi-scale CNN approach is a feasible and effective method for Salmonella serotype identification.
• This technique shows potential for identifying other bacterial species and serovars.
• The developed method offers a significant advancement in rapid pathogen detection.