Elsevier

Food Control

Volume 50, April 2015, Pages 630-636
Food Control

Classification of intact açaí (Euterpe oleracea Mart.) and juçara (Euterpe edulis Mart) fruits based on dry matter content by means of near infrared spectroscopy

https://doi.org/10.1016/j.foodcont.2014.09.046Get rights and content

Highlights

  • This study is the first attempt to evaluate the DM content of açaí and juçara intact fruit using NIR spectroscopy.

  • PCA-DA model has 72% accuracy in predicting DM content of new population.

  • Non-destructive classification of DM content can be used to optimize the volume of water added during processing.

Abstract

The processing of açaí (Euterpe oleracea Mart.) and juçara (Euterpe edulis Mart) fruit requires water addition for adequate pericarp extraction. Currently, the amount of added water is based on fruit moisture content as estimated using a convection oven method. In this study, diffuse reflectance FTNIR spectra (1000–2500 nm, 64 scans and spectral resolution of 8 cm−1) of intact açaí and juçara fruit were used to discriminate fruit batches based on the dry matter (DM) content using mature fruit collected over two years. Spectra were collected of ∼25 fruits per batch, placed on a 90 mm diameter glass dish in a single layer. The calibration set contained of 371 lots, while the prediction set consisted of 132 lots (of different locations, times). Spectra were subject to several pre-processing methods and models were developed using Partial Least Squares Regression (PLSR), Partial Least Squares-Discriminant Analysis (PLS-DA) and Principal Component Analysis Discriminant Analysis (PCA-DA). A PLSR model constructed using the wavelength range of 1382–1682 nm and full multiplicative scatter correction achieved a root mean square error for prediction on DM of 5.25% w/w with a ratio of the standard deviation of DM set to the bias corrected RMSEP of 1.5 on the test set. A PCA-DA model based on the same wavelength of region outperformed the PLS-DA method to segregate the test population into categories of high (>32 %DM) and low DM (<32% DM) with 74% accuracy achieved. The PCA-DA technique is recommended to the processing industry as a non-destructive and rapid method for optimisation of water added during processing using batch assess of fruit from incoming lots of fruits.

Introduction

Açaí (Euterpe oleracea Mart.) is endemic to Amazonian floodplain ecosystems and juçara (Euterpe edulis Mart.) is found in the Atlantic Forest Region (Pessoa and Teixeira, 2012, Inácio et al., 2013). Both fruits have been traditionally used as food in these regions. A typical fruit weights 1–3 g, of which 5–25% is the edible pulp (exocarp and mesocarp - pericarp) that surrounds a single seed (Borges et al., 2011, Pessoa and Teixeira, 2012, Schauss et al., 2006).

Both fruit have exceptionally high antioxidant activity (Aguiar et al., 2013, Inácio et al., 2013, Poulose et al., 2012), and were included in the top ten “super foods” in 2012 (Pessoa and Teixeira, 2012, Smith, 2013). These fruits have also been used in the nutraceutical and cosmetic industries due to the reports of efficacy of fruit extracts in, e.g., combating inflammatory and oxidative mediators involved in ageing (Poulose et al., 2012, Zhao et al., 2004).

Frozen and dried pulp are produced and exported, with exports to the United States beginning in 2000, and since expanding to include Australia, Europe and Japan. The value of the exported product reached US$ 2.1 million in 2002, and it increased to US$ 17 million in 2012, corresponding to about 6000 tons of pulp, mainly açaí (S.E.Agri, 2014).

In Brazil, the quality of fruit pulp is regulated by the Brazilian Ministry of Agriculture and Food Supply (MAPA), which established a specification based on total soluble solids content (TSS) of pulp. Three grades are recognized (>14%; 11–14%; and 8–11% TSS; Brazil, 2000).

The current production process involves reception, sieving, removal of fruit with visual defects (disease, bruises, insect damage), washing, sanitation, blanching, softening of acceptable fruit in water (35–40 °C, 20 min), processing to separate pulp from seed, homogenization, pasteurization, packaging, freezing and storage (Schwob, 2012). Currently sorting is based on human visual inspection, but the processing system would suit use of machine vision, as is done for cherry. Fruits are mechanically de-pulped in a 1:1 or 1:2 ratio of fruit:water (Rogez et al., 2012, Silva et al., 2013), followed by homogenization to the current standards. Product may also be dried by lyophilisation, atomization (spray-drying), vacuum or other methods (Schwob, 2012). The addition of too much water during processing will cause product to fail the specification, and thus involves the cost of later water removal. On the other hand, the addition of too little water results in poor pericarp (pulp) extraction from the fruit.

Thus the amount of water used in the softening process represents a significant management point. Commonly, fruit moisture content is measured at reception, prior to fruit sorting, using the convection oven method (Schwob, 2012). The oven method is time consuming (∼24 h), and suffers sampling issues, due to wide variation of dry matter (DM) among fruit in a given population. Near infrared (NIR) spectroscopy may be an alternative method, conditional on the ability to create a robust calibration for this indirect analysis technique. NIR spectroscopy is used to assess DM of other intact fruit, for example, Subedi and Walsh (2011) reported the use of interactance spectra (720–920 nm) of intact mango to predict DM content, with root mean square error for prediction (RMSEP) between 0.69 and 1.14 % w/w. Koizimi, Trevelin, Pessoa, Cunha Junior, and Teixeira (2013) reported use of reflectance spectra (1000–2500 nm; acquired using a Spectrum 100N FTNIR, PerkinElmer company) to assess the DM content of açaí reconstituted juice from commercial pulp extracts, reporting a root mean square error for cross validation (RMSECV) of 0.95% and ratio of standard deviation of DM of the calibration set and RMSECV (RPD) of 3.3.

The assessment of fruit DM content normally involves the use of an interactance optical geometry, with an effective ‘optically sampled volume’ to a depth of between 10 and 30 mm (Subedi & Walsh, 2011). In contrast, reflectance geometry is considered to result in optical sampling of intact fruit to a depth of <4 mm (Lammertyn, Peirs, Baerdemaeker, & Nicolaï, 2000). In the current application we desire to minimise information from the seed, and as the pericarp has a thickness between 3 and 8 mm (Pessoa & Teixeira, 2012), a reflectance geometry is appropriate.

NIR spectroscopy estimation of a continuous variable such as DM is typically undertaken using multivariate linear regression techniques (e.g. multiple linear regression; principal component regression; Partial Least Squares Regression), and can be undertaken using non-linear techniques (e.g. artificial neural networks; support vector machines; soft independent modelling of class analogies; Mukarev & Walsh, 2012). Supervised pattern recognition analysis can be used to classify samples based on predetermined categories, for example, linear discriminant analysis (LDA) and Partial Least Square-Discriminant Analysis (PLS-DA) (Naes, Isaksson, Fearn, & Davies, 2002). LDA was used by Santos, Nardini, Cunha Junior, Barbosa Junior, & Teixeira (2014) to distinguish juçara from açaí fruit based on rare earth element concentration, achieving an accuracy of external classification of 83.3%.

The evaluation of any proposed technique must include a validation step based on assessment of samples not included in the training set. For example, Brito et al., 2013, Salguero-Chaparro et al., 2013 and Sinelli, Cerretani, Egidio, Bendini, & Casiraghi (2010) developed discriminant analyses models based on only one population of fruit, randomly divided to calibration and validation sets. The reported accuracy of 80–95% of is thus optimistic, and not a true test of the ability of the model to predict or classify fruit outside the population used in calibration (Golic & Walsh, 2006). In this study, the use of reflectance FTNIR spectra of intact açaí and juçara fruit to assess fruit dry matter (DM) was investigated content.

Section snippets

Plant material

Açaí (E. oleracea Mart.) and juçara (E. edulis Mart.) fruit bunches were harvested at commercial maturity (deep purple fruit) several times during the cropping season at four localities (within São Paulo State, Brazil) and in two years (2012 and 2013). A total of fifty bunches were harvested, with each bunch coming from a separate tree. The açaí locations were: (i) from a privately owned garden in Américo Brasiliense (21°42′ S latitude, 48°01′ W longitude, and 646 m altitude), São Paulo State

Spectra

DM content of açaí and juçara fruit batches varied between 18.6 and 49.5 % (Table 1). On average, fruit with low DM showed higher apparent absorbance (log 1/R) values than fruit with high DM (also see difference spectra, Fig. 1A). This phenomenon is consistent with increased scattering of light by the lower DM fruit, inferring smaller cell size and more water–air interfaces. The absorbance spectra also displayed major features associated with the O–H first overtone region (1382–1682 nm) and O–H

Conclusions

Despite the thin and variable pericarp thickness, NIR reflectance spectroscopy was demonstrated to be a useful technique to sort intact açaí and juçara fruit based on dry matter content. PLSR models achieved a RMSEP of 5.03, while the supervised pattern recognition technique of PCA-DA achieved 72% accuracy in discriminating fruit into two categories (high, above 32%, and low, below 32% w/w DM content). While not allowing precise control of water addition, at-line NIR spectroscopy of incoming

Acknowledgements

The authors would like to thank the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) for the financial support of this research (Proc. 2008/51408-1, 2011/19669-2) and for providing the BEPE scholarship (Proc. 2013/0.6089-3).

References (25)

  • A.O.A.C

    Association of official analytical chemists: Official methods of analysis – Met. 925.23

    (1998)
  • D.S. Borges et al.

    Chemical characterization, bioactive compounds, and antioxidant capacity of jussara (Euterpe eduli) fruit from the Atlantic Forest in southern Brazil

    Food Research International

    (2011)
  • Cited by (27)

    • Association Between Diet, Health, and the Presence of Bioactive Compounds in Foods

      2019, Bioactive Compounds: Health Benefits and Potential Applications
    • Association Between Diet, Health, and the Presence of Bioactive Compounds in Foods

      2018, Bioactive Compounds: Health Benefits and Potential Applications
    • Cellular energy allocation analysis of multiple marine bivalves using near infrared spectroscopy

      2018, Ecological Indicators
      Citation Excerpt :

      Predicted versus measured glycogen content had 90.8% of samples within the RMSEP range and 91.6% of samples within a 95% confidence range (Fig. 2c). NIRS modelling of glycogen and sugar-related compounds has been shown to be effective in multiple food product studies (Cunha Junior et al., 2015; Li et al., 1996; Pan et al., 2015) and natural bivalve populations (Brown, 2011) with a range of modelling methods. As previously discussed, TCA hydrolysis of glycogen and colorimetric determination using phenol and hydrochloric acid has been shown to be an effective method for glycogen determination in bivalves (De Coen and Janssen, 1997; Dridi et al., 2007; DuBois et al., 1956).

    View all citing articles on Scopus
    View full text