Optimizing Excitation Light for Accurate Rapid Bacterial Species Identification with Autofluorescence

Area of Science:

• Analytical Chemistry
• Microbiology
• Biophysics

Background:

• Rapid bacterial identification in patient samples is vital for effective infectious disease treatment and healthcare economics.
• Autofluorescence spectroscopy offers a potential method for bacterial species identification, but optimal excitation wavelengths need to be determined.

Purpose of the Study:

• To investigate an algorithm for improving bacterial species identification accuracy using fluorescence spectroscopy.
• To identify optimal excitation wavelengths and their number for accurate bacterial diagnosis.

Main Methods:

• Developed and verified a machine learning classifier algorithm for bacterial species identification using autofluorescence.
• Evaluated diagnostic accuracy for ten bacterial species across various excitation wavelengths.
• Utilized Extra Tree (ET), Logistic Regression (LR), and Multilayer Perceptron (MLP) algorithms to determine key wavelengths.

Main Results:

• Identified key excitation wavelengths at 280 nm, 300 nm, 380 nm, and 480 nm, with 280 nm being most significant.
• Achieved diagnostic accuracy comparable to using 200 wavelengths with only two wavelengths for ET and LR, and three for MLP.
• Demonstrated that a limited number of wavelengths within an optimal range are sufficient for accurate bacterial identification.

Conclusions:

• An optimal range of excitation wavelengths exists for spectroscopic measurement of bacterial autofluorescence.
• Accurate bacterial species identification can be achieved using a minimal set of excitation wavelengths, enhancing diagnostic efficiency.