Non-destructive assessment of instrumental and sensory tenderness of lamb meat using NIR hyperspectral imaging

Highlights

• We develop hyperspectral imaging for predicting instrumental & sensory tenderness.

• We use spectral features to predict these attributes in lamb meat.

• We build and optimise multivariate prediction models.

• We select tenderness related wavelengths using successive projections algorithm.

• We conduct analysis to correlate between textural features and tenderness values.

Abstract

The purpose of this study was to develop and test a hyperspectral imaging system (900–1700 nm) to predict instrumental and sensory tenderness of lamb meat. Warner–Bratzler shear force (WBSF) values and sensory scores by trained panellists were collected as the indicator of instrumental and sensory tenderness, respectively. Partial least squares regression models were developed for predicting instrumental and sensory tenderness with reasonable accuracy (R_cv = 0.84 for WBSF and 0.69 for sensory tenderness). Overall, the results confirmed that the spectral data could become an interesting screening tool to quickly categorise lamb steaks in good (i.e. tender) and bad (i.e. tough) based on WBSF values and sensory scores with overall accuracy of about 94.51% and 91%, respectively. Successive projections algorithm (SPA) was used to select the most important wavelengths for WBSF prediction. Additionally, textural features from Gray Level Co-occurrence Matrix (GLCM) were extracted to determine the correlation between textural features and WBSF values.

Introduction

Meat is a widely consumed product, it is also a perishable product. In order to enhance its quality, new techniques such as novel refrigeration processes (Sun and Eames, 1996, Sun et al., 1996, Sun, 1997, Sun, 1998, Sun, 1999, Desmond et al., 2000, McDonald and Sun, 2001, McDonald and Sun, 2001, Wang and Sun, 2002a, Wang and Sun, 2002a, Li and Sun, 2002, Sun and Hu, 2003, Sun and Li, 2003, Sun and Zheng, 2006) are continuously developed, at the same time, novel techniques are required for its efficient and effective quality measurement and control. Among many quality parameters, tenderness is regarded as one of the most important attributes that affects the eating quality of meat. It is positively correlated with juiciness and taste and it has a substantial influence on overall customer satisfaction (Naganathan et al., 2008a). Inconsistencies in meat tenderness have led to the decline in customer satisfaction and subsequently a reduction in market shares (Cluff et al., 2008, Naganathan et al., 2008b). On the other hand, guaranteeing tenderness would increase consumers’ confidence and significantly influences the decision to return (Shackelford, Wheeler, & Koohmaraie, 2005). Several studies have shown that consumers would be ready to pay more for a guaranteed tender steak (El Jabri et al., 2010, Sun et al., 2012). Tenderness is a property of a cooked meat and accurate measurement and prediction of this property from a fresh steak is a major concern for the meat industry. Despite its importance, it is one of the attributes that is most difficult to evaluate before purchase, because it is not visible and is highly variable. Ideally tenderness would be measured by a consumer panel as ultimately the customer must be satisfied. This is however an ambitious task, which is difficult to organise and is usually not feasible due to time and financial considerations (Jackman, Sun, & Allen, 2009). Alternatively, it can be measured by trained sensory panels or by instrumental methods. Although instrumental methods such as Warner–Bratzlar shear force (WBSF) measurement provide reliable information about meat tenderness (Wu, Peng, et al., 2012), they still have some disadvantages for industrial applications because these methods are slow and destructive. On the other hand, sensory analysis with trained panellists is still can be subjective, very expensive, time consuming, and also destructive, and hence, it is not possible to use as a routine analysis in the meat processing industry (Wang, Lonergan, & Yu, 2012). Although these methods for tenderness prediction are not suitable for a fast-paced production and processing environment, they are still widely used due to unavailability of smart alternatives to measure this palatability attribute. Therefore, it is desirable to develop a fast, non-destructive, accurate, and on-line technique to predict meat tenderness.

Conventional imaging techniques (Kumar and Mittal, 2009, Pallottino et al., 2010, Quevedo et al., 2010) have proved some success as an objective method in meat tenderness prediction especially when colour, marbling and texture features are used as the indicators in tenderness prediction (Jackman, Sun, & Allen, 2010). Also, as one of the major optical applications, spectroscopic techniques (Klaypradit et al., 2010, Liu et al., 2012, Liu et al., 2011, Quevedo and Aguilera, 2010, Shao et al., 2011) in both visible (Vis) and NIR regions have been widely used in predicting meat tenderness (Andrés et al., 2008, Barlocco et al., 2006). As a logic extension of imaging (Zheng et al., 2006, Sun and Brosnan, 2003, Du and Sun, 2005) and spectroscopy, hyperspectral imaging is an important technique that integrates both techniques in one system to attain both spatial and spectral information from an object. Hyperspectral imaging is a technique whereby hundreds of reflectance images are captured over a broad wavelength range at contiguous and narrow intervals, forming a three-dimensional structure of multivariate data (hypercube). When the hyperspectral data are appropriately processed, it is possible to automatically identify the location of features that display specific spectral signatures and to map the gradient and spatial distribution of specific attributes. In recent years, hyperspectral imaging techniques have received much attention for food quality and safety inspection (ElMasry et al., 2009, Mahesh et al., 2008, Nagata et al., 2006, Taghizadeh et al., 2011), particularly, considerable research endeavours have been conducted to evaluate the quality and safety traits of meat and meat products (ElMasry et al., 2011, ElMasry et al., 2011, Grau et al., 2011, Kamruzzaman et al., 2012, Kamruzzaman et al., 2011, Kamruzzaman et al., 2012a, Kamruzzaman et al., 2012b, Kim et al., 2004, Kim et al., 2006, Kobayashi et al., 2010, Nakariyakul and Casasent, 2008, Nakariyakul and Casasent, 2009, Qiao et al., 2007, Qiao et al., 2007, Qiao et al., 2007, Segtnan et al., 2009, Segtnan et al., 2009, Sivertsen et al., 2011, Sivertsen et al., 2011, Sone et al., 2012), including prediction of the tenderness of beef and pork (Cluff et al., 2008, ElMasry et al., 2012, Naganathan et al., 2008a, Naganathan et al., 2008b, Sugiyama, 1999, Tao et al., 2012, Wu et al., 2012). However, to the best of our knowledge, no research endeavours are yet executed for predicting tenderness of lamb meat using hyperspectral imaging. Therefore, it is of our aim to implement this emerging technology for predicting both instrumental and sensory tenderness in lamb meat.

The main aim of the current study was to examine the potential of using NIR hyperspectral imaging as a fast and non-destructive method to predict lamb meat tenderness assessed by instrumental measurement and sensory analysis. Specific objectives were to (1) establish a NIR hyperspectral imaging in the NIR range of 910–1700 nm, (2) build calibration models to quantitatively correlate spectral information and measured tenderness, (3) select the tenderness-related wavelengths by SPA for predicting instrumental tenderness, and (4) investigate the correlation between WBSF values and textural features extracted via GLCM.