How to Analyze Fourier Transform Infrared (FTIR) Spectra?

This article explores the application of Fourier Transform Infrared Spectroscopy (FTIR) in detecting the storage time of Agaricus blazei, and introduces two feature interval extraction algorithms—Backward Interval Partial Least Squares (BIPLS) and Synergy Interval Partial Least Squares (SiPLS)—to optimize the model. The study results indicate that, compared to traditional full-spectrum data, using feature interval extraction algorithms significantly enhances the predictive performance of the model. Among them, the BIPLS method demonstrates superior performance on both training and testing sets.

Agaricus blazei, also known as the Brazilian mushroom, is a nutrient-rich saprophytic fungus containing proteins, lipids, amino acids, and various polysaccharides. However, due to its high metabolic rate, the mushroom has a short shelf life. During storage, the mushroom undergoes a series of physicochemical and microbial activity changes. Therefore, developing fast and effective methods to detect its storage duration is crucial for ensuring food quality and safety.

Infrared spectroscopy is a non-destructive testing method that can quantitatively analyze samples by establishing models. However, spectral data from different feature intervals contain varying amounts of information. The BIPLS and SiPLS algorithms effectively improve model performance by calculating the contribution of intervals to the model and selecting and combining intervals accordingly.

In this study, the experimental Agaricus blazei samples were purchased from Kunming, Yunnan, and stored at (8±4)℃ for 0-6 days, measured every 2 days. The samples were dried at 55℃ to a constant weight and ground into a fine powder, which was mixed with potassium bromide at a ratio of approximately 16:1 and pressed into tablets for spectral testing. The instrument used was a Frontier FTIR spectrometer from Perkin Elmer. The spectral acquisition range was 4000–400 cm^-1, with a resolution of 4 cm^-1 and 16 scans. The potassium bromide background spectrum was subtracted, and a total of 90 samples were tested (0 days: 26, 2 days: 20, 4 days: 22, 6 days: 22). Figure 1 shows the original infrared spectra of Agaricus blazei at different storage days.

The PLS model established based on FTIR technology can effectively detect the storage time of Agaricus blazei. Combining the feature interval extraction algorithm significantly improves model performance. Especially the FTIR-PLS model improved using the BIPLS method showed the best results on both the training and testing sets. This work demonstrates that the PLS model based on FTIR technology is an efficient method for detecting the storage time of Agaricus blazei.

Figure 1 FTIR spectra of Agaricus blazei Murrillmushroom with different storage days

In the experiment, the raw spectra were preprocessed using second-order derivatives and normalization, and the Kennard-Stone algorithm was used to divide the samples into training and testing sets. Finally, the BIPLS and SiPLS algorithms were used to extract feature intervals from the infrared spectral data and establish corresponding PLS models. The performance of the Agaricus blazei storage time prediction models established using the extracted feature intervals (BIPLS, SiPLS) . The results show that the PLS model established based on FTIR technology can effectively determine the storage duration of Agaricus blazei. Compared with traditional full-spectrum data, the model performance is improved when combined with feature interval extraction algorithms, and the model improved using the BIPLS method showed the best results in both training and testing sets.

Figure 2 The results of feature interval extraction using BiPLS