Development of a cost-effective dual-mode hyperspectral imaging system with machine learning for enhanced bacterial...

Area of Science:

• Optics and Photonics
• Biotechnology
• Machine Learning Applications

Background:

• Bacterial identification is crucial in diagnostics and research.
• Existing methods can be time-consuming or lack specificity.
• Hyperspectral imaging (HSI) offers a non-invasive approach for analyzing sample spectral properties.

Purpose of the Study:

• To develop a cost-effective, dual-mode hyperspectral imaging (HSI) system.
• To integrate HSI with machine learning for enhanced bacterial detection and classification.
• To validate the system's performance for identifying specific bacterial species.

Main Methods:

• Constructed a dual-mode HSI system using COTS components and 3D printing.
• Implemented a compound prism-grating-prism spectrograph for simplified optics.
• Utilized reflectance and fluorescence modes with LED illumination (VIS-NIR and UV).
• Developed a machine learning framework for spectral feature fusion.

Main Results:

• Achieved spectral resolution of 1.55 nm and spatial resolutions of ~0.81 mm x 0.49 mm.
• Classified *Staphylococcus aureus* and *Pseudomonas aeruginosa* with high accuracy.
• Dual-mode classification accuracy reached 97.11%, surpassing single modes (92.55% reflectance, 93.48% fluorescence).

Conclusions:

• The developed dual-mode HSI system provides a scalable and economical solution for bacterial identification.
• Optical-computational approach enhances classification accuracy for high-throughput applications.
• Demonstrates the potential of integrated HSI and machine learning in microbiology.