Quantitative determination of phosphorus in seafood using laser-induced breakdown spectroscopy combined with machine learning

Highlights

• Total phosphorus in seafood was determined by laser-induced breakdown spectroscopy.

• Matrix effect was assessed by using univariate and multivariate analysis methods.

• Machine learning helps to reduce the matrix effect for LIBS-based seafood analysis.

• Non-linear SVM model showed better analytical performance than linear PLS model.

Abstract

Quantitative determination of phosphates or total phosphorus in seafood is of great importance for the fraud detection as well as food security issues. In this work, laser-induced breakdown spectroscopy (LIBS) was applied as a rapid method for phosphorus determination in three types of seafood including codfish, scallop and shrimp. Both univariate and multivariate regression models were established with special attentions on the correction of matrix effect to improve the analytical performances of LIBS. The obtained results showed that compared with the traditional univariate model and the linear PLS model, the non-linear SVM model could provide the best figures-of-merit with R² of 0.9904, RMSEC, RMSEP and ARE of 1.68 g/kg, 1.42 g/kg and 3.70%, respectively. The average RSD of prediction of SVM is 5.18%, which is much lower than the value of PLS (9.40%) and is comparable to the value of univariate model (4.11%). This indicates that SVM may be more suitable to address the non-linear behaviors in LIBS spectra caused by the matrix effect, and therefore leads to a more robust calibration model. The present results demonstrated the capacity of LIBS combined with machine learning in phosphorus determination of seafood products, which could be potentially used for on-site phosphates detection within the food supply chains.

Introduction

Phosphates have been widely used as food additives in seafood with the legitimate functional aims including retention of natural moisture, inhibition of flavor and lipid oxidation, aiding emulsification and removal of shell fish shells and offering cryoprotection [1]. However, because phosphates can be used to retain moisture, there are concerns about the use of excess phosphates to retain “added water”, thus increasing the size and weight of seafood products. This can lead to unfair trade practices and result in economic fraud [2]. According to China's national standard GB2760-2014, phosphates can be added to frozen aquatic products and frozen surimi products to a maximum level of 5 g/kg (expressed as PO₄³⁻). The European Community Directive 95/2/EC and Regulation 1333/2008 state that a maximum level of 5 g/kg of phosphates can be added to frozen and deep-frozen mollusks and crustaceans (expressed as P₂O₅). As for the health impact, excess phosphorus intake may cause disturbance of the homeostasis and lead to the development of cardiovascular disease, renal impairment, and bone loss [3]. The Recommended Dietary Allowance (RDA) of phosphorus is 700 mg/day for adults. For these reasons, it is important to monitor the level of phosphorus in seafood which is relevant to the economic as well as food security issues.

A variety of analytical methods for phosphates or total phosphorus detection in seafood and meat products have been used, among which we can find the classical spectrophotometric method [4], ion-exchange chromatography (IEC) [5], capillary electrophoresis (CE) [6], or ³¹P nuclear magnetic resonance (NMR) [7]. Other established techniques have also been applied to phosphorus determination in food such as inductively coupled plasma optical emission spectrometry (ICP-OES) or mass spectrometry (ICP-MS) [8,9], or wavelength dispersive X-ray fluorescence (WD-XRF) [10]. Although these methods are effective, they in general need complex and time-consuming sample pretreatment, associated with expensive equipment and a certain amount of expertise that typically correspond to laboratory-based techniques. Heavy pretreatment procedures are needed especially for seafood or meat samples because of their complex organic matrix where substantial quantity of chemical reagents or solvents are usually required. There is therefore an obvious interest to develop an easier, quicker and environmentally friendly method, which becomes crucial for on-site food fraud detection and food screening within increasingly complex and globalized food supply chains [11].

Laser-induced breakdown spectroscopy (LIBS) is an optical emission spectroscopy technique that uses a laser pulse to vaporize, atomize, and excite a hot plasma as the spectroscopic emission source [12]. With the advantages of minimal sample preparation, rapid detection and multi-elemental analysis capabilities, LIBS has been proven to be an attractive technique for food analysis [13,14], and was applied in various food matrixes such as milk [15], bakery products [16], wines [17], rice [18], vegetable oils [19], flour [20], and tea samples [21]. For meat analysis, LIBS has been used in several applications including quantitative determination of calcium, magnesium, potassium and sodium in bovine and chicken meat [22], determination of chromium in pork [23], quantification of lithium in meatballs [24], quantification of copper in beef [25], and identification of meat species for adulteration detection [26,27]. The potential of LIBS for on-line monitoring of calcium in comminuted poultry meat was also reported for industrial application [28]. Although LIBS has demonstrated its versatility features in numerous fields, the quantitative analysis capability of LIBS is still considered as its Achilles' heel [29]. A major factor affecting the quantification performance is the matrix effect [30]. In particular, for LIBS analysis of seafood products, the matrix effect could be more pronounced due to the large variations in chemical composition and physical property between different types of seafood, such as fishes, mollusks (e.g., scallop), and crustaceans (e.g., shrimp). Spectral data treatment with univariate analysis often leads to poor analytical performances. Multivariate analysis based on chemometrics or machine learning method has been proven to be able to provide robust calibration models and can significantly reduce the matrix effect to improve the LIBS measurement accuracy [[31], [32], [33]].

In this work, we evaluated the potential of LIBS as a rapid method for determining phosphorus in seafood products for the first time. Special attentions were paid on the matrix effect in phosphorus quantification by preparing three series of seafood including codfish, scallop, and shrimp samples with different P concentrations. The reference P concentration in each sample was determined by the standard spectrophotometric method. Calibration models were built by univariate analysis method, as well as two multivariate analysis methods based on partial least square (PLS) and support vector machines (SVM). The analytical performances were compared between different calibration models.