ATR-FTIR with chemometrics tools for classification and prediction of hydrogen peroxide in liquid milk
Hydrogen peroxide (H2O2) is a powerful oxidising agent that can serve as milk adulterant to extend its shelf life. Hydrogen peroxide (H2O2) is incorporated into milk in minimal quantities to suppress bacterial proliferation and enhance the milk's coloration. This study effectively employed Atte...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Academic Press
2025
|
| Subjects: | |
| Online Access: | https://eprints.ums.edu.my/id/eprint/44234/1/FULL%20TEXT.pdf https://eprints.ums.edu.my/id/eprint/44234/ https://doi.org/10.1016/j.jfca.2025.107223 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Hydrogen peroxide (H2O2) is a powerful oxidising agent that can serve as milk adulterant to extend its shelf life. Hydrogen peroxide (H2O2) is incorporated into milk in minimal quantities to suppress bacterial proliferation and enhance the milk's coloration. This study effectively employed Attenuated Total Reflectance–Fourier Transform Infrared (ATR-FTIR) spectroscopy alongside multivariate data analysis techniques, including Principal Component Analysis (PCA), Discriminant Analysis (DA), and Multiple Linear Regression (MLR), for the discrimination and prediction of H2O2 in milk. There were 3 different approaches of wavenumber used to see the suitable wavenumber window for identification of H2O2 in UHT milk which are full spectra (4000–500 cm−1), 3200–1020 cm−1, 900–600 cm−1. The PCA indicated a cumulative variability of 85.4 % with four groups including unadulterated and different level groups of adulterants. Consequently, the DA model exhibited an exceptional differentiation between unadulterated and adulterated UHT milk, achieving a classification accuracy exceeding 95.0 % during cross-validation and 82.0 % for testing dataset utilizing the full spectra. The MLR model yielded R² values of 0.957 and 0.923, with RMSE and MSE below 1 for calibration and validation respectively. Subsequently, MLR models were externally validated with a test dataset, revealing a result for t-test (p > 0.05) demonstrating MLR's capability to ascertain the percentage of H2O2 in UHT milk as low as 0.5 %v/v. The wavenumber 2826.06 cm−1 is an appropriate fingerprint region for identifying H2O2 in the UHT milk. |
|---|