Comparison of derivative preprocessing and automated polynomial baseline correction method for classification and...

Area of Science:

• Analytical Chemistry
• Spectroscopy

Background:

• Raman spectroscopy is crucial for analyzing diverse samples, including illicit narcotics.
• Baseline drift in Raman spectra complicates analysis and quantification.
• Effective baseline correction is essential for accurate spectral interpretation.

Purpose of the Study:

• To compare derivative preprocessing and the Lieber-Mahadevan-Jansen (LMJ) polynomial method for Raman spectra baseline correction.
• To evaluate the impact of these methods on factor analysis, analyte discrimination, and quantification.
• To assess the suitability of these methods for analyzing complex mixtures like diluted narcotics.

Main Methods:

• Application of derivative preprocessing and LMJ polynomial methods to Raman spectra of 85 solid illicit narcotic samples.
• Analysis of spectral data using principal component analysis (PCA).
• Quantification of analytes (cocaine, heroin, MDMA) using cross-validation error metrics.

Main Results:

• Both methods yielded comparable principal component analysis (PCA) models.
• Similar quantification errors were observed: ~8% for cocaine and heroin, ~3-4% for MDMA.
• The LMJ method retained original spectral peak shapes, simplifying interpretation.

Conclusions:

• Neither derivative preprocessing nor the LMJ method showed superior performance in terms of PCA models or quantification errors.
• The LMJ method is recommended for Raman spectroscopy research due to its ability to preserve spectral peak shapes.
• Baseline correction using the LMJ method facilitates more straightforward interpretation of preprocessed Raman spectra.