Hyperspectral upgrade solution for biomicroscope combined with Transformer network to classify infectious bacteria

Area of Science:

• Microbiology
• Spectroscopy
• Artificial Intelligence

Background:

• Bacterial infectious diseases present a major public health challenge, necessitating rapid and precise identification methods.
• Hyperspectral microscopic imaging (HMI) is a promising technique for microbial detection due to its nondestructive, rapid, and data-rich nature.

Purpose of the Study:

• To develop a cost-effective hyperspectral biomicroscope by augmenting a standard biomicroscope.
• To evaluate the efficacy of a Transformer-based classification network for identifying bacterial pathogens using HMI data.

Main Methods:

• A prism-grating-prism configuration was employed to upgrade a standard biomicroscope into a hyperspectral system.
• 600 hyperspectral data cubes were generated for four bacterial species: Listeria, Bacillus typhi, Bacillus pestis, and Bacillus anthracis.
• A Transformer-based neural network was designed and trained for pathogen classification.

Main Results:

• The developed hyperspectral biomicroscope successfully acquired spectral data from bacterial samples.
• The Transformer-based classification network achieved a high accuracy of 99.44% in distinguishing between the tested bacterial species.
• The proposed method demonstrated superior performance compared to traditional identification techniques.

Conclusions:

• The integration of hyperspectral microscopic imaging with an optimized Transformer network offers a powerful approach for the rapid and accurate detection of infectious disease pathogens.
• This cost-effective HMI system has significant potential for application in clinical diagnostics and public health surveillance.
• Further research can explore expanding the library of detectable pathogens and refining the classification algorithms.