Near-infrared calibration transfer based on spectral regression

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Near-infrared (NIR) spectroscopy is a powerful analytical technique.
• Transferring calibration models between different NIR instruments is challenging.
• Existing methods like piecewise direct standardization (PDS) have limitations.

Purpose of the Study:

• To propose and evaluate a novel calibration transfer method for NIR spectra.
• To compare the performance of spectral regression against PDS.
• To assess the impact of standardization subset size on method performance.

Main Methods:

• Development of a calibration transfer method based on spectral regression.
• Application of spectral regression to reveal low-dimensional manifold structures in high-dimensional spectroscopic data.
• Comparative analysis with piecewise direct standardization (PDS) on benchmark NIR datasets.

Main Results:

• The proposed spectral regression method demonstrated superior performance compared to PDS.
• Spectral regression showed competitive results even against PDS with background correction.
• Excellent performance was achieved when the standardization subset contained sufficient samples.

Conclusions:

• Spectral regression is a robust and effective method for NIR spectral calibration transfer.
• The method's ability to handle high-dimensional data makes it suitable for inter-instrument standardization.
• Sufficient data in the standardization subset is crucial for optimal performance of spectral regression.