Correlation of odor and color profiles of oysters (Crassostrea virginica) with electronic nose and color machine vision.

ABSTRACT Odor and color changes in oysters (Crassostrea virginica) stored at 1.8[degrees] and 7[degrees]C were measuredevery 3 days for up [o 13 days with an electronic nose, a computer vision system, and sensory panels. Electronic nose and odor sensory data was analyzed using Discriminant Function Analysis (DFA). Correct classification rates of 100% were obtained to group electronic nose data by days and sensory scores at each temperature. When all temperature and days data were pooled, DFA predicted sensory scores based on electronic nose readings with 94% accuracy. Oyster colors were analyzed using the 512 color block scheme. Eight colors were found to be significant. The correlations of these colors, as well as the average L*a*b* values with time and with color sensory data suggested prediction of color quality. This study suggests that electronic nose and color vision could be used to objectively evaluate the quality of oysters.

KEY WORDS: oyster, E-Nose, machine vision, odor and visual quality

INTRODUCTION

Oysters are a variable commodity worldwide. In 1998, United Slates' domestic landings of oysters by weight of meat was 15,213 metric tons, valued at $88 million (U.S. Department of Commerce, 1999). Imports increased from 4,506 metric tons in 1997 to 6,169 metric tons, valued at $25 million, in 1998.

Oysters have a distinctive odor and flavor. Fresh shellfish products indicates typical aquatic fresh odor, brilliant appearance and physical peculiarities of the species in good situation (Perkins et al. 1992). The odor of seafood has been widely used as one of the signs of quality for a long lime (Botta et al. 1995), the flesh color is a measure of the freshness of the seafood and research proves that consumers equate freshness with the vibrancy of the flesh color (Beaudoin, 1997).

Most studies regarding shellfish quality have focused on taste-acttive and flavor components (Sekiwa et al., 1997; Ok et al., 1996; Lee et al., 1995; Tanchoticul and Hsieh, 1991). Effects of chemical and other treatments on the taste and color of shellfish were also studied (Jawahar et al., 1994; Chellappan, 1991).

Fresh shellfish products should be marketed as fast as possible since their shelf life is very short. A critical commercial step after sorting is the inspection of the product. Current techniques concerning the quality assessments of oysters rely on sensory inspections that use the senses of vision, smell, etc. These are difficult to quantify to compare with objective standards and their results are not sufficiently repeatable. There is a necessity for objective, more reliable and quick technique to evaluate the quality of shucked oyster, to serve for these sector and international markets.

Developments in sensor technology and electronic noses (EN) have many potential applications in the food industry (Bartlett el al., 1997; Corcoran et al., 1993) including flavor and aroma profiles fresh squeezed orange juice, coffee powder, beer and cheese (Bazemore et al., 1996; Delaure et al., 1996; Pearce et al., 1993; Jou and Harper, 1998), quality control area of grapes, olive oil, ground beef, (Tokusoglu and Balaban, 2000; Aparicio et al., 2000; Winquist et al., 1993), aroma discrimination of black teas (Tokusoglu et al. 2002; Tokusoglu, 2001), separation of spice and grain mixture (Brezmes et al., 1997; Borjesson et al., 1996), among others. Some work was published on the correlation of EN sensor outputs with sensory data in seafood products. EN has been applied to odor evaluation of seafood (Dodd et al., 2000; Luzuriaga, 1999; Balaban et al., 1996; Balaban et al., 1994).

Computer vision is being used in some food and agricultural areas to automate the inspection of visual quality (Sarkar et al., 1991). Visual inspection of seafood products involves image processing software and video cameras to evaluate product color, appearance, size or shape. There is reported research in the application of the machine vision technology for quality evaluation of seafood (Balaban et al., 1994; Balaban et al., 1996; Luzuriaga et al., 1995; Luzuriaga et al., 1997; Newman et al., 1998), for grading and sorting (Par et al., 1995; Diehl et al., 1990) and for shape detection (So et al., 1996). No study could be found concerning quality determination of oysters with E-nose and computer vision for objective evaluation. The objectives of this study were 1) to develop predictive models of odor change in shucked oysters stored at 1.8, and 7[degrees]C using an electronic nose, 2) to correlate these readings with sensory evaluation data, 3) to develop models of color change in these samples using machine vision and correlate them with visual sensory evaluation data.

MATERIALS AND METHODS

Oyster Samples and Storage Conditions

Fresh oysters (Crassostrea virginica) front Apalachicola, FL in the Gulf of Mexico were obtained. The weight of each oyster with the shell was approximately 75 g; the meat was approximately 20% of the total weight, about 15 g. Dating from the first day, they were stored in 100 ml beakers, 2 oysters per covered beaker, in cold rooms at 1.8, and 7.0[degrees]C for up to 13 days. For the machine vision experiments, 2 oysters were placed in a small petri dish (60 nun diameter), the lids covered, and stored at the same temperatures. The samples were analyzed every three days. This study was repeated twice.

Electronic Nose Measurements

An Electronic Nose model 4000 (EEV Inc, UK) equipped with 12 conducting polymer sensors (types 458, 459, 460, 461, 462, 463, 464, 478, 483, 297, 298, 401) was used. Before the experiment started, E-nose was calibrated with propylene glycol (chromatographic grade. Fischer Scientific, No. P-355-20, Fair Lawn, N J; 75% v/v) as recommended by the manufacturer.

Every day before the experiments, compressed dry air (CGA Grade D, Strade Welding Supply Inc., Jacksonville, FL) was passed through the EN sensors for 20 min. A covered beaker containing oyster samples was taken out of the cold room one hour before analysis and equilibrated to room temperature (25[degrees]C). The beaker was then placed in the EN. Five consecutive readings were taken for each sample. Then, the replicate sample was taken out of the cold room and treated similarly. For each reading, the vessel and head were purged with dry air for 2 and 4 min, respectively. Then. sensor responses were read for 4 min. For each sample total analysis time was 10 min.

Computer Vision Analysis

The color machine vision system developed by Luzuriaga and Balaban (1997) was used to grab images. The light box had 3 sections consisting of a top lighting box, a bottom lighting box and a sample chamber. The illuminant was 15 watt Chroma 50 fluorescent light (F15T8 C50, General Electric, Cleveland, OH). The S-video output of a video camera (Sony SSC-S20, Sony Corp., Japan) was connected to a color frame grabber (Meteor, Matrox Electronic Systems, Ltd., Dorval, Quebec) mounted inside a Pentiara 133 MHz PC computer. In this experiment, the settings for the video camera were: hue = 0, saturation = 160, contrast = 160, and brightness = 148 (for a range from 0 to 255). During image acquisition a standard color the (beige, L = 63.59, a = 65.19, b = 33.39) was placed next to the oyster samples in the light box.

For each sampling day, the small petri dish was removed from the cold room, placed in the light box with front and back lighting, the image of oysters stored 1.8 and 7.0[degrees]C was captured and saved in a computer file. The petri dish was then placed back in the cold room. The same samples were used throughout the experiment. This study was repeated twice. Oysters remained at room temperature for no more than about 5 min daring image capture.

Images were calibrated by the color analysis software using the standard color the (Luzuriaga, 1999). The color of the calibrated images was reported in the RGB and L, a, b color systems. The 512-color blocks scheme was used for the RGB colors (Luzuriaga, 1999). The value for each color block was the percent of the view area of the oyster covered by that color. Colors that were more than 5% of the total area were considered.

Sensory Evaluation

The odor and color of oyster samples were analyzed by a 10-member trained sensory panel consisting of graduate students, professors, and personal, 24-50 y of age, from the Food Science and Human Nutrition Dept. at the University of Florida. Panelists were trained using oyster samples from the same supplier utilizing a descriptive sensory analysis method. They used the samples as "excellent" and "poor" as a guide in their decision to evaluate odor and color for the unknown samples, respectively. Samples stored at two different temperatures were evaluated using odor description lot the best sample: typical aquatic fresh odor, distinctive oyster aroma, and the worst sample; deteriorated odor, putrefactive aroma. For color description the best sample had brilliant color, good appearance, and the worst sample had deteriorated, dull, slimy appearance.

A 5-point scale was used in this study where 1 = poor, 2 fair, 3 = good, 4 = very good, 5 = excellent. Samples were evaluated for odor and color every three day for up to 13 days. This study was applied according to IFT Sensory Evaluation Division (IFT, 1994). Odor and color profile in descriptive sensory profiling was applied as a method (Piggott et al., 1998) and hedonic scale was carried out as an overall acceptability test in consumer and acceptance tests.

Data Analysis

Discriminant Function Analysis (DFA), a parametric procedure to identify relationships between qualitative variables such as sensory descriptors, odor, color classes, anti quantitative variables such as sensor response, was used. It tries to classify samples into known groups by constructing linear relationships for the sensor data or color data and the criterion set of variables. This method generates functions that are used to classify samples. Each discriminant function is calculated for which the F-ratio of the analysis of the variance is maximized (Gardner and Hines, 1997).

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Function 1 gives the most separation between groups, function 2 gives the next best separation, and so on. Generally the first two functions are sufficient to separate the groups. E-nose sensor readings, color vision data and sensory scores were analyzed in Statistica for Windows ('98 edition, Ver. 6.0, StatSoft Inc., Tulsa, OK) using discriminant function analysis (DFA) to develop the predictive functions for oysters stored at different temperatures and storage times, and to correlate these with sensory data.

In this study, the 12 sensor outputs in E-nose analysis and 8 color outputs obtained from color analysis were reduced to 2 discriminant functions. These functions were used to graph the data in two dimensional plots and observe separation of groups. For each function, coefficients and correct classification matrix were obtained.

RESULTS AND DISCUSSION

E-nose readings with odor sensory evaluation data for oyster stored at different temperatures for different times were correlated and also machine vision readings with color sensory evaluations for oyster stored at two different temperature for up to 13 days were correlated using Discriminant Function Analysis. Figure 1 indicates DFA correlations of EN readings with sensory criteria data at two storage temperature. For 1.8 and 7.0[degrees]C, all temperature and days pooled data, discriminant function analysis perfectly separated the E-nose data with correct classification rates of 100%. 100%, 94% respectively (Table 1).

[FIGURE 1 OMITTED]

There was no significant difference between the number of EN readings (n) used to obtain the discriminant functions (P < 0.01).

During sensory panel, panelists judged the differences in odors at two different temperature for up to 13 days (Table 2). Table 2 indicates odor sensory scores of oysters stored at 1.8 and 7.0[degrees]C by sensory panel. Oyster odor changed with storage time. The differences between the two replicate of sensory scores of both temperatures were not significant (P < 0.01).

At two temperatures, some panelists could evaluate udder differences with difficulty between successive days while EN detected these differences clearly. After Day 3, panelists could not discriminate odor of semideteriorated and deteriorated oyster samples and these were evaluated as putrefactive aroma. When all temperature and days data were pooled together, DFA predicted sensory scores based on electronic nose readings with 94% accuracy.

DFA was used to calculate two discriminant functions to correlate EN readings with storage time at both temperature. Table 3 shows coefficients for Discriminant Functions to estimate oyster odor based on storage times at 1.8[degrees]C, 7.0[degrees]C and both temperatures with EN.

DFA analysis of e-nose odor data correlated with storage time at 1.8, 7.0[degrees]C gave correct classification rates of 100% and 100%, respectively (Table 4). There was no differences between the number of Electronic Nose readings (n) used to obtain the discriminant functions (P < 0.01). Figure 2 shows a clear discrimination for up to 13 days at both temperature. As shown in these figures, there was no overlap in days, meaning that there were definite differences in the sensor readings for each day of storage.

[FIGURE 2 OMITTED]

Computer vision could clearly quantify non-uniform colors of oyster. Classification functions of the correlation between color data and storage time, as well as color sensory evaluation data, were obtained. When classified by day at each temperature, the classification rates for 1.8[degrees]C and 7.0 [degrees]C were 100% and 100%, respectively. Classification of color data by sensory readings gave classification rate of 100%, 100%, respectively.

Eight colors were found to be significant for oyster color quality with RGB color system. The correlations of these colors, as well as the average L*a*b* values with time and with color sensory data suggested prediction of color quality.

Figure 3 shows average L*a*b* values of oysters stored at 1.8[degrees]C. The lightness scale (L) differences of oyster samples for both replicates were not significant (p < 0.01). Redness or greenness (a) and yellowness or blueness (b) can reflect a single color function called color difference, and this difference is a measure of the distance in color space between two colors. According to L*a*b* results, the differences of oyster color stored for up to 13 days could not be clearly determined (Fig. 3).

[FIGURE 3 OMITTED]

The panelists judged differences in color quality in both temperatures during storage time and in the same samples evaluated for up to 13 days (Table 5). Table 5 indicates color sensory scores of oysters stored al 1.8 [degrees]C and 7[degrees]C by the sensory panel. Oyster color changed with storage time. The differences between the two replicates of sensory scores of both temperatures were not significant (p < 0.01).

At 1.8[degrees]C and 7.0[degrees]C. panelists could not evaluate color quality differences after day 3, whereas machine vision clearly determined these differences between semideteriorated and dull, slimy samples.

Each color block of oyster and RGB values of these colors were determined using the color machine vision system and the eight colors obtained are shown in Table 6.

Six of the 8 colors were identified as deteriorated color whereas the remaining two were represented by yellow and orange hues.

During storage time, good color, or Color 73 (RGB = 245,241,273), was decreased over time, and fresh oyster appearance was decreased as show in Fig. 4(a). The level of this color was decreased to around 17% for storage at 1.8[degrees]C and 8% for 7.0[degrees]C from around 29%. [Fig. 4(a)]. There were no significant differences between the two replicates of color analysis (p < 0.01). While computer vision was determined deterioration from after day 3, objectively; sensory criteria could not evaluate visual quality of oysters. Color (145 + 146) (RGB = 80,80,64) increased over time and level of this color was increased to around 27% for storage at 1.8[degrees]C and 36% for 7.0[degrees]C from around 16%, respectively [Fig. 4(b)]. These additional colors were responsible for the grayish-dark hue of deteriorated oyster.

[FIGURE 4 OMITTED]

E-nose and machine vision data can be used in pattern recognition for objective classification of oyster. With comprehensive study oyster samples by odor and color, the e-nose can be used reliably to classify and machine vision can objectively classify oyster samples by color.

DFA can be used satisfactorily as a pattern recognition analysis technique in food engineering area for the identification and determination of seafood odor and color.

TABLE 1.
Percent Correct Classification Rates for the DFA of EN Readings
Compared with Sensory Scores.

                1.8[degrees]C   7.0[degrees]C   Combined T

Sensory Score     (n = 10)        (n = 10)       (n = 20)

      1              100             100            97
      2              --              100           100
      3              100             100           100
      4              100             --             60
   Overall           100             100            94

n = number of EN readings used to obtain the discriminant functions.

TABLE 2.
Odor Sensory Scores Obtained by Sensory Panel. (a)

        Sensory scores   Sensory scores

ODOR    1.8[degrees]C    7.0[degrees]C

(DAY)      Average          Average

  0           5                5
  1           4                3
  3           3                2
  6           1                1
  9           1                1
 13           1                1

(a) (p < 0.01).

TABLE 3.
Coefficients for DFA of Oyster Odor Based on Storage Times
and EN Readings.

                  Coefficients for Discriminant Functions

              Storage at        Storage at        Both
            1.8[degrees]C     7.0[degrees]C     temperatures

 Sensor    Funct.   Funct.   Funct.   Funct.   Funct.   Funct.
  Type       1        2        1        2        1        2

Type 301    -1.90    10.70    -3.26     4.06    -5.06     4.71
Type 298   -14.77    37.13     3.49     8.69   -26.91    31.05
Type 297    47.72   -33.80    -5.76    -4.65    39.38   -38.07
Type 283     1.35   -14.18     3.54   -13.23    12.21    -5.75
Type 278   -12.55   -17.17     0.74    -6.48     0.69    -3.87
Type 264    -3.21    65.60    -7.01    17.08   -29.68    31.96
Type 263    -7.78   -26.87    22.52    -3.44    11.03    -2.82
Type 262    14.40    -4.97    -4.05     3.58     6.74   -10.92
Type 261    -3.78    -4.86   -13.25   -11.67    -8.20   -14.83
Type 260    18.71    18.76    -0.20    -0.25     9.76    13.22
Type 259    -9.43   -30.43     9.38     5.52     9.44    -8.57
Type 258   -29.02     0.19    -6.23     0.90   -19.00     3.84

TABLE 4.
Percent Correct Classification Rates for the DFA of
EN Readings Compared with Storage Time.

        1.8[degrees]C   7.0[degrees]   Combined T

(DAY)     (n = 10)        (n = 10)      (n = 20)

  1          100            100           100
  3          100            100           100
  6          100            100            95
  9          100            100           100
 13          100            100            95

n = number of EN readings used to obtain the discrimination
functions.

TABLE 5.
Color Sensory Scores Obtained by Sensory Panel. (a)

        Sensory scores   Sensory scores

COLOR   1.8[degrees]C    7.0[degrees]C

(DAY)      Average          Average

  0           5                5
  1           4                3
  3           3                2
  6           1                1
  9           1                1
 13           1                1

(a) (p < 0.01).

TABLE 6.
KGB Values of Each Color Block of Oyster.

COLOR NO   R value   G value   B value          COLOR

   72         48        48        16     Dark olive
   73        245       241       173     Pale orange yellow
  137         67       107        68     Dark yellowish green
  145         80        80        48     Greyish olive
  146         80        80        80     Dark grey
  209        112        80        48     Moderate brown
  210        112        80        80     Dark red
  218        112       112        80     Greyish olive

ACKNOWLEDGMENT

This research was supported in part by the Florida SeaGrant Project R/LR-Q-17 The part of this work was presented at the "2002 IFT Annual Meeting June 15-19 in Anaheim Convention Center, Anaheim, California, USA."

REFERENCES

Aparicio R., S. M. Rocha, I. Delgadillo & M. T. Morales. 2000. Detection of rancid defect in virgin olive oil by the electronic nose. J Agric. Food Chem. 48:853-860.

Balaban M. O., S. Yeralan & Y. Begmann. 1994. Determination of count and unite, fruity ratio of shrimp by machine vision. Journal of Aquatic Food Product Technolog. 3: 43-58.

Balaban M. O. & D. A. Luzuriaga. 1996. Automated quality evaluation of shrimp. Food Quality. 5: 56-66.

Bartlett P. N., J. M. Elliot & J. W. Gardner. 1997. Electronic noses and their application in the food industry. Food Technol. 51: 44-48.

Bazemore R.. R. Rouseff & C. Sims. 1996. Comparison of fresh squeezed orange juice analysis by electronic nose, sensory panel and GC FID. The 3rd International Symposium on Olfaction and Electronic Nose. Nov. 3-6, 1999. Miami. FL.

Beaudoin J. 1997. In search of the true hue. Seafood Business. Jan/Feb: 30-34.

Botta J. R. 1995. Evaluation of seafood freshness quality. New York, NY: VCH Publishers Inc.

Chellappan, J. 1991. Processing of oyster meat by freezing. Fishery Technology. 28: 122-124.

Borjesson, T., T. Eklov, H. Sundgren & J. Schnurer. 1996. Electronic nose for odor classification of grains. Cereal-Chemisty. 73: 457-461

Brezmes J., B. Ferreras, E. Liobet, X. Vilanova & X. Correig. 1997. Neural network based electronic nose for the classification of aromatic species. Analytica-Chimica-Acm. 348: 103, 503-509.

Corcoran, P. 1993. Electronic odor sensing systems. Electronic Communication Engineering J. 5: 303-308.

Delaure, J., C. Gretsch, R. Liardob & P. Visani. 1996. Detection of aroma above a coffee powder limits perspectives of electronic sensors. Proceedings of the 3rd International Symposium on Olfaction and Electronic Nose. Nov. 3-6, 1996. Miami, FL.

Diehl K. C., T. W. Awa, R. K. Byler, M. F. van Gelder, M. Koslav & C. R. Hackney. 1990. Geometric and physical properties of raw oyster meat as related to grading. Transactions of the ASAE. 33: 1270-1274.

Dodd, T. H., S. A. Hale & S. M. Blachard. 2000. An aroma based method for determining the storage time of fishery product. 2000. ASAE Annual International Meeting. July 9-12, 2000. Milwaukee, Wisconsin.

Gardner, J. W. & E. L. Hines. 1997. Pattern Analysis Technique. In: E. Kress-Rogers, Editor. Handbook of biosensors and electronic noses. Medicine, food and the environment. Boca Raton, FL: CRC Press: pp. 633-652.

IFT. 1994. Guidelines fur the preparation and review of papers reporting sensory evaluation data.

Jawahar, A. T., S. Balasundari & J. G. Indra. 1994. Influences of antioxidants on the sensory quality and oxidative randicity of frozen edible oyster. Journal of Food Science and Technology India 31: 168-170.

Jou, K. D., & W. J. Harper. 1998. Pattern recognition of Swiss cheese aroma compounds by SPME/GC and an electronic nose. Milchwissenschaft. 53: 5, 259-263.

Lee, J. S. 1995. Isolation and some properties of bitter taste compounds from cultured oyster, Crassastrea gigas. Journal of the Korean Fisheries Society. 28: 569-575.

Luzuriaga, D. A. 1995. Development and testing of an automated quality evaluation device for shrimp. Masters' Thesis. Gainesville, EL: University of Florida.

Luzuriaga, D. A., M. O. Balaban & S. Yeralan. 1997. Analysis of visual quality attributes of white shrimp by using machine vision. Journal of Food Science. 62:113-118, 130.

Luzuriaga, D. A. 1999. Application of computer vision and electronic nose technologies for quality assessment of color and odor of shrimp and salmon. Doctoral Thesis. Gainesville, FL: University of Florida.

Newman, D. 1998. Automation of quality evaluation of raw tuna. Masters' Thesis. Gainesville, FL: University of Florida.

Ok S. J., Il Seo Kwon, S. L. Young, H. L. Jong. D. C. Sang & H. S. Ki. 1996. Changes in contents of some taste compounds of dried mussel and baby clam during storage. Korean Journal of Food Science and Technology. 28; 882-887.

M. B. Par, R. K. Byler, K. C. Diehl & C. R. Hackney. 1995. Machine vision based oyster meat grading and sorting machine. Journal of Aquatic Food Product Technology. 3: 5-24.

Pearce, T. C., J. W. Gardner, S, Friel, P. N. Bartlett & N. Blair. 1993. Electronic nose for monitoring the flavour of beers. Analyst. 118: 371-377.

Perkins, C. 1992. The Advanced Seafood Handbook. Rockland, ME: Seafood Business Magazine.

Piggott, J. R., S. J. Simpson & A. R. Williams. 1998. Sensory analysis. International Food, Science and Technology. 33: 7-18.

Sarkar, N. 1991. Machine vision for quality control in the food industry. In: D. Y. Fung and R. E. Matthews., editors, instrumental methods for quality, assurance in Foods. New York, NY: Marcel Dekker Inc. pp.167-187.

Sekiwa, Y. K., K. Kubota & A. Kobayashi. 1997. Characteristic flavor components in the brew of cooked clam (Meretrix luxoria) and the effect of stage on flavor formation. Journal of Agricultural and Food Chemistry. 45: 826-830.

So, J. D. & E. W. Wheaten. 1996. Computer vision detection of oyster hinge lines. Transactions of the ASAE. 39: 1557-1566.

Tanchoticul, U. & T. C. Y. Hsieh. 1991. Analysis of volatile flavor components in steamed rangia clam by dynamic headspace sampling and simultaneous distillation and extraction. Journal of Food Science. 56: 327-331.

Tokusoglu, O., M. K. Unal & M. O. Balaban. 2002. Identification of aroma characteristics of various black tea infusions using electronic nose technology. IFT Annual Meeting Book of Abstracts

Tokusoglu, O. 2001. In the determination of the major phenolic compounds (flavanols, flavonols, Tannis) and aroma properties of black teas. Doctoral Thesis. Izmir, Turkey: University of Ege, Bornova.

Tukusoglu, O. & M. O. Balaban. 2000. Evaluation of odor and color changes of muscadine grapes stored at different temperatures by electronic nose and computer vision. IFT Annual Meeting Book of Abstracts. June 11-14, 2000. Dallas, Texas. pp. 61-62.

USDC. 1999. Fisheries of the United States. 1999. Current Fishery Statistics No. 9600. U.S.D.C. U.S. Government Printing Office, Washington, DC.

Winquist, F., E. G. Hornsten, H. Sundren & I. Lundstrom. 1993. Performance of an electronic nose for quality estimation of ground meat, Meas. Sci. Technol. 4: 1493-1500.

OZLEM TOKUSOGLU (1) * AND MURAT O. BALABAN (2)

(1) Ege University Engineering Faculty, Department of Food Engineering, 35100, Bornova, Izmir, Turkey; (2) University of Florida, Food Science and Human Nutrition Department, Gainesville, FL, USA 32611

* Corresponding author. E-mail: [e-mail]