Fourier based partial least squares algorithm: new insight into influence of spectral shift in "frequency domain"

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Partial least squares (PLS) combined with spectroscopy is vital for chemical analysis.
• Spectral shift is a significant challenge, reducing accuracy and stability in practical spectroscopic applications.

Purpose of the Study:

• To develop a novel method for completely removing spectral shift in spectroscopic analysis.
• To enhance the robustness and practical applicability of PLS-spectroscopy methods.

Main Methods:

• A Fourier transform-based partial least squares (FT-PLS) method was proposed.
• Spectral shift was transformed into a phase term in the frequency domain using Fourier transform.
• The module of the Fourier transformed spectra was calculated to remove the phase term, thus eliminating spectral shift.

Main Results:

• The FT-PLS method effectively removed spectral shift, significantly improving model accuracy and stability.
• The developed model demonstrated insensitivity to spectral shift.
• The approach proved effective in practical applications, exemplified by blood species discrimination using Raman spectroscopy.

Conclusions:

• The FT-PLS method offers a new perspective for addressing spectral shift in spectroscopic analysis.
• This technique enhances the reliability of spectroscopic methods for real-world applications by reducing environmental measurement constraints.
• The study highlights the potential of FT-PLS for more robust and accurate chemical analysis.