Rapid beef quality detection using spectra pre‐processing methods in electrical impedance spectroscopy and machine learning

• Published in: International journal of food science & technology 2024-03, Vol.59 (3), p.1624-1634
• Main Authors: Qiu, Junhong, Lin, Yuduan, Wu, Jiaqing, Xiao, Yuhui, Cai, Honghao, Ni, Hui
• Format: Article
• Language: English
• Subjects: Accuracy; Beef; Beef quality inspection; Carrageenan; carrageenan adulteration; Data acquisition; Datasets; Electrical impedance; electrical impedance spectroscopy; Impedance; Impedance spectroscopy; Learning algorithms; Machine learning; Meat; Meat quality; Model accuracy; Reagents; repeated freeze–thaw; Sample preparation; Spectra; spectra pre‐processing; Spectroscopy; Spectrum analysis; Temperature requirements; water adulteration
• ISSN: 0950-5423; 1365-2621
• DOI: 10.1111/ijfs.16915

Summary The fraudulent practice of beef adulteration is a growing concern, as it violates consumer rights. Electrical impedance spectroscopy (EIS) combined with machine learning has emerged as a widely used approach to identify low‐quality meat. Unlike traditional biochemical methods that require expensive instruments, complex sample preparation, and chemical reagents, EIS is a cost‐effective alternative. However, EIS data are susceptible to temperature fluctuations, requiring a waiting period under consistent temperature conditions for data stabilisation before measurements. This process becomes impractical when dealing with a large number of samples. To overcome this limitation, standardisation, normalisation, and smoothing methods were introduced in the meat quality detection based on EIS data. A recognition model for detecting carrageenan adulteration in beef was established. Under an inconsistent temperature condition, by applying the spectra pre‐processing methods to the prediction dataset, the model accuracy reached 84%, whereas the accuracy of the unprocessed prediction dataset dropped to 54%. This study demonstrates that acquiring EIS data under consistent temperature conditions is unnecessary if proper spectra pre‐processing methods are applied. By eliminating the waiting time for data stabilisation, this practical approach enhances the efficiency and accuracy of meat quality detection. Electrical impedance spectroscopy (EIS) combined with machine learning is a popular method for meat quality inspection. However, EIS data is sensitive to temperature fluctuations, necessitating a waiting period under stable temperature conditions for data stabilisation. To address this, standardisation, normalisation, and smoothing techniques were introduced to EIS data. This study shows that consistent temperature conditions are not necessary when proper spectra pre‐processing methods are applied.