Improved Classification Performance of Bacteria in Interference Using Raman and Fourier-Transform Infrared...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Rapid and sensitive detection of bioaerosols is crucial for public health.
• Pollen interference can significantly impact the spectral identification of bacterial species.
• Raman and Fourier-Transform Infrared (FTIR) spectroscopy are key techniques for bioaerosol analysis.

Purpose of the Study:

• To investigate the impact of pollen on bacterial identification using Raman and FTIR spectra.
• To develop and evaluate machine learning algorithms for classifying spectral data and mitigating pollen interference.
• To assess the effectiveness of different classification models in identifying bacterial species.

Main Methods:

• Spectral data from fourteen bacterial classes were preprocessed and features extracted using machine learning.
• Classification models including Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machine (SVM), and Random Forest (RF) were employed.
• Raman and FTIR spectral data, as well as their fused data, were analyzed.

Main Results:

• PLS-DA achieved classification accuracies of 78.57% and 92.85%.
• SVM demonstrated 100% accuracy in classifying Raman spectral data.
• RF achieved 100% accuracy for both individual spectra and fused data, effectively eliminating pollen interference.

Conclusions:

• Machine learning algorithms, particularly SVM and RF, provide highly accurate methods for bacterial species identification from spectral data.
• The developed spectral processing algorithms efficiently eliminate pollen interference, enhancing bioaerosol detection reliability.
• This research offers a robust approach for public health protection through improved bioaerosol monitoring.