Progress 10/01/06 to 09/30/11
Outputs
OUTPUTS: A sophisticated hyperspectral imaging laboratory consisting of CCD-based imaging system (400-1000 nm), InGaAs-based imaging system (900-1700 nm), various spectrometers, and illumination systems has been developed. 1. Hyperspectral imaging to predict beef tenderness: We have developed the portable hyperspectral imaging (HSI) system. The hyperspectral imaging system consisted of a mirror scanner, spectrograph imaging module, a CMOS camera, lens, dome lighting, computer, and an adjustable-height mobile HSI workstation. The spectral range of the system was 400 nm - 1000 nm. We travelled to a Tyson beef packing plant in Dakota City, Nebraska to collect samples for the project. After image acquisition of the ribeye on the hanging carcasses, strip loins were cut on the carcasses. All strip loins were aged for 14 d postmortem. Trays were wrapped with overwrap to prevent dehydration prior to cooking. Warner-Bratzler shear (WBS) force was then collected on cooked samples. The division between tender and tough was > 3.9 kg. Samples were sorted based on WBS values and every third sample was then allotted for validation, while the remaining samples were used for training, i.e., developing the model. 2. Imaging for detecting freeze injury in cucumbers: Hyperspectral imaging (two systems: 400-1000 nm and 900-1700 nm) and magnetic resonance imaging systems have been evaluated for detecting freeze injury in cucumbers. Pickling cucumber harvested from a private farm of Nebraska, USA, were transported to the laboratory within 24 h of harvesting. On day 1 after harvest, all the 176 cucumber were numbered and their physical properties viz. length, maximum diameter and weight were measured using a calliper gauge (least count: 0.1 mm) and analytical balance. Moisture content of the cucumber was determined using standard hot air oven method by keeping three samples (three replicates, each about 10 g) in oven at 102 C till a constant weight was achieved. Randomly selected forty four cucumber samples were bulk packed in a polyethylene bag (each bag contained around 10 fruits) and placed in a deep freezer maintained at -18 C. After this one time exposure to the freeze damage the samples were removed from the deep freezer after 150 min and transferred to a control atmosphere chamber maintained at 12 C and 85% RH. The freeze damaged samples and control samples were stored in this chamber for the rest of the experimental period. Texture profile analysis (TPA) of the cucumber samples was conducted using Texture Analyzer. MRI experiments were measured using a 9.4 T (400 MHz for protons), 89 mm vertical bore MR system (Varian, inc. Walnut Creek, CA) equipped with triple axis gradients (100 G/cm) and a 4 cm Millipede transmit/receive radiofrequency coil. T1 and T2 relaxation times of the samples were assessed using a multi-echo multi-slice (MEMS) sequence. Three cucumber samples each from freeze damaged and control lot were picked randomly for imaging. Same samples were imaged at 1, 2, 4 and 7 days after harvest, i.e. 0, 1, 3 and 6 days after inducing freeze injury, during the experiment. PARTICIPANTS: Shadi Othman, University of Nebraska; Nachiket Kotwaliwale, Central Institute of Agricultural Engineering, India; Chris Calkins, University of Nebraska; Ashok Samal, University of Nebraska, Renfu LU, USDA-ARS. TARGET AUDIENCES: Not relevant to this project. PROJECT MODIFICATIONS: Not relevant to this project.
Impacts
1. Hyperspectral imaging to predict beef tenderness: Tender certification accuracies (when the hyperspectral system certifies a sample as tender, what percentage of those samples is actually tender) are 71 and 72% for training and validation. Tough identification accuracies (what percentage of tough samples was correctly identified) are 71 and 65% for training and validation. The overall accuracies are 60 and 64% for training and validation. Validation results are slightly better than the training results, which indicates the model is not-overfitted and should generalize well for classifying new samples. We have demonstrated that the sophisticated hyperspectral imaging system can be made portable. The system was successfully implemented to collect the images on exposed ribeyes of the hanging carcasses and predict tenderness with high accuracy. Categorizing meat cuts by tenderness will help the U.S. beef industry expand its market and increase its share of the protein market. Labeling accurate quality factors on retail packaging will add value to the products. This system will enhance economic opportunities for cattle producers and processors by improving assessment of beef product quality to meet consumer expectations. Producers would enhance breeding efforts identified in relation to greater muscle tenderness and processors would be able to charge more for the most consistently tender product. 2. Imaging for detecting freeze injury in cucumbers: Images are currently being analyzed. Principal component analysis of hyperspectral images of cucumbers acquired on day 1 after induced freeze damage showed the area of damage, where mold growth was evident on day 8. Thus, hyperspectral imaging shows promise for early detection of freeze injury.
Publications
- Cluff, K., G. Konda Naganathan, J. Subbiah, C. Calkins and A. Samal. 2010. Pork tenderness evaluation using optical scattering with near-infrared (NIR) hyperspectral imaging. Annual International ASABE Meeting, Pittsburgh, PA. Paper No. 1009905.
- Konda Naganathan, G., K. Cluff, J. Subbiah, A. Samal, and C. Calkins. 2010. An on-line hyperspectral imaging system for tenderness-based grading of beef. Mid-central ASABE Meeting, St. Joseph, MO. Paper No. MC10-307.
- Konda Naganathan, G., D. Jonnalagada, R. Wehling and J. Subbiah. 2010. Determination of yolk contamination in liquid egg white using Raman spectroscopy. Annual International ASABE Meeting, Pittsburgh, PA. Paper No. 1009860.
|
Progress 10/01/08 to 09/30/09
Outputs
OUTPUTS: Hyperspectral Imaging: Development of reliable instruments to predict tenderness has long been a priority of the beef industry. In collaboration with Texas A&M University, University of Missouri, Oklahoma State University, and Colorado State University, three tenderness prediction instruments were validated under packing plant conditions. One of the instruments validated was the portable hyperspectral imaging system developed at the University of Nebraska-Lincoln. Cattle (n = 730) were selected from four processing plants in Western KS to vary in quality grade. Emphasis was placed on lower quality grades to develop a data set that included tougher carcasses. Cattle were scanned after grading and an approximate 7.62-cm section of the longissimus muscle was sampled from the striploin from each side of the carcass. Samples from the left side of each carcass were allotted to slice shear force (SSF) and the right side was allotted to Warner-Bratzler shear force (WBSF). All of the samples were aged for 14 d postmortem and shear force was conducted on the fresh, never frozen sample. The data set was then divided into a prediction equation refinement data set and a sequestered data set based on tenderness classification. Tenderness classification was based on SSF where tender < 24.9 kg and tough > 25.0 kg. Eggshell quality: Eggshell quality is an important factor in egg production. Different applications require different shell strength. Therefore, it is important to detect eggs with unfit shells and remove them from further incubation, and processing. The objective of this project was to develop ultrasound based instrumentation for investigating eggshell strength. Ultrasound signals from 157 egg samples were recorded and modeled with wavelet analysis to predict shell strength characteristics such as shell thickness, fracture force, and stiffness. In addition, physical parameters: egg weight, volume, and surface area, were used to develop regression equations to predict shell strength characteristics. Peanut moisture content: There are some commercial instruments available that use near-infrared (NIR) radiation measurements to determine the moisture content (MC) of a variety of grain products, such as wheat and corn, without the need of any sample grinding or preparation. However, to measure the MC of peanuts with these instruments, the peanut kernels have to be chopped into smaller pieces and filled into the measuring cell. This is cumbersome, time consuming, and destructive. An NIR reflectance method is presented here by which the average MC of about 100 g of whole kernels could be determined rapidly and nondestructively. The MC range of the peanut kernels tested was between 8% and 26%. Initially, NIR reflectance measurements were made at 1 nm intervals in the wavelength range of 1000 to 1800 nm, and the data were modeled using partial least squares regression (PLSR). The predicted values of the samples tested in the above range were compared with the values determined by the standard air-oven method. PARTICIPANTS: Chari Kandala, USDA-ARS, National Peanut Research Lab; C. L. Lorenzen, University of Missouri; R. K. Miller, Texas A&M University; J. B. Morgan, Oklahoma State University; K. E. Belk, Colorado State University; C. Calkins, University of Nebraska; A. Samal, University of Nebraska. TARGET AUDIENCES: Cattle producers, beef processors, peanut processors. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.
Impacts
Hyperspectral Imaging: Overall accuracy was determined as (the number predicted in to the correct tenderness classification/the total number predicted) x 100. The overall accuracies of all three instruments ranged from 79 to 81%. However, the hyperspectral imaging was superior in identifying tough carcasses, which was twice as accurate as the next best instrument in the study. Thus we have demonstrated that the developed portable hyperspectral imaging system can predict tenderness on the hanging carcasses in a non-destructive manner. This nondestructive system allows the packers to label the steaks as "guaranteed tender" and this adds value to the product. In additions, producers will be paid based on true value of the carcasses and consumer satisfaction for beef will be improved. Eggshell quality: The ultrasound wavelet model had correlation coefficient values of 0.66, 0.64 and 0.57 in predicting shell thickness, fracture force and stiffness, whereas the physical parameter model had 0.36, 0.39, and 0.41, respectively. The model developed with ultrasound wavelet parameters was more accurate than that of physical parameters. Hence, ultrasound measurements show potential in accurately determining shell strength characteristics. Peanut moisture content: The predicted values agreed well with the air-oven values, with an R2 value of 0.93 and a standard error of prediction (SEP) of 1.18. Using the PLSR beta coefficients, five key wavelengths were identified, and MC predictions were made using multiple linear regression (MLR). The R and SEP values of the MLR model were 0.91 and 1.09, respectively. Both methods performed satisfactorily and, being rapid, nondestructive, and noncontact, may be suitable for continuous monitoring of MC of grain and peanuts as they move on conveyor belts during their processing.
Publications
- Konda Naganathan, G., P. Chapain, I. Poudel, G.W. Froning, G.R. Bashford, G.E. Meyer, and J. Subbiah. 2009. Predicting Eggshell Strength Characteristics using Ultrasound. ASABE Annual International Meeting, Reno, NV. Paper No. 096753.
- Konda Naganathan, G., C. V. K. Kandala and J. Subbiah. 2008. NIR reflectance spectroscopy for nondestructive moisture content determination of peanut kernels. Transactions of the ASABE, accepted.
- Kumar, V., D. Jonnalagadda, J. Subbiah, A.P. Wee, H. Thippareddi, and S Birla. 2009. A 3-D heat transfer and fluid flow model for cooling of a single egg under forced convection. Accepted for publication in Transactions of the ASABE.
- Juneja, V.K., M.V. Melendres, L. Huang, J. Subbiah, and H. Thippareddi. 2009. Mathematical modeling of growth of Salmonella in raw ground beef under isothermal conditions from 10 to 45 degrees celcius. International Journal of Food Microbiology, 131(2-3): 106-111.
- Kandala, C. V. K., G. Konda Naganathan, and J. Subbiah. 2009. NIR Reflectance Spectroscopic Method for Nondestructive Moisture Content Determination of In-Shell Peanuts. Paper No. 095749, Annual ASABE International Meeting, Reno, NV.
- Kandala, C. V. K., J. Sundaram, G. Konda Naganathan, C.L. Butts, and J. Subbiah. 2009. Estimation of moisture and oil content of in-shell nuts with a capacitance sensor using discrete wavelet analysis. SPIE Optics and Photonics Paper No. 7294.
|
Progress 10/01/07 to 09/30/08
Outputs
OUTPUTS: A near-infrared (NIR) hyperspectral imaging was developed to forecast 14-day aged, cooked beef tenderness from the hyperspectral images of fresh ribeye steaks (n=319) acquired at 3 to 5 day post-mortem. A pushbroom hyperspectral imaging system (wavelength range: 900 - 1700 nm) with a diffuse-flood lighting system was developed. After imaging, steaks were vacuum-packaged and aged until 14 days postmortem. After aging, the samples were cooked and slice shear force (SSF) values were collected as a tenderness reference. After reflectance calibration, a Region-of-Interest (ROI) of 150 300 pixels at the center of longissimus muscle was selected. Partial least squares regression (PLSR) was carried out on each ROI image to reduce the dimension along the spectral axis. Gray-level textural co-occurrence matrix analysis with two quantization levels (64 and 256) was conducted on the PLSR bands to extract second-order statistical textural features. These features were then used in a canonical discriminant model to predict three beef tenderness categories, namely tender (SSF < 205.80 N), intermediate (205.80 N < SSF < 254.80 N), and tough (SSF > 254.80 N). PARTICIPANTS: Konda Naganathan, G., L. M. Grimes, C. R. Calkins, A. Samal, K. Cluff. TARGET AUDIENCES: Target audiences include beef packers and cattle producers. PROJECT MODIFICATIONS: Not relevant to this project.
Impacts
The hyperspectral imaging system correctly classified 242 out of 314 samples with an overall accuracy of 77.0%. We demonstrated that hyperspectral imaging can forecast tenderness of aged, cooked beef from the hyperspectral images of fresh beef in lab-settings. A patent has been filed. This nondestructive system allows the packers to label the steaks as "guaranteed tender" and this adds value to the product. In additions, producers will be paid based on true value of the carcasses and consumer satisfaction for beef will be improved.
Publications
- J. Subbiah,C. Calkins, A. Samal. 2008. System and method for analyzing material properties using hyperspectral imaging. United States Patent 20080199080. Cluff, K., G. Konda Naganathan, J. Subbiah, R. Lu, C. Calkins and A. Samal. 2008. Optical scattering in beef steak to predict tenderness using hyperspectral imaging in the VIS-NIR region. Sensing and Instrumentation for Food Quality and Safety 2(3): 189-196. Konda Naganathan, G., L. M. Grimes, J. Subbiah, C. R. Calkins, A. Samal and G. E. Meyer. 2008. Visible/near-infrared hyperspectral imaging for beef tenderness prediction. Computers and Electronics in Agriculture, 64 (2); 225-233. Konda Naganathan, G., L. M. Grimes, J. Subbiah, C. R. Calkins, A. Samal and G. E. Meyer. 2008. Partial least squares analysis of near-infrared hyperspectral images for beef tenderness prediction. Sensing and Instrumentation for Food Quality and Safety 2(3): 178-188. Grimes, L., G. Konda Naganathan, J. Subbiah and C. Calkins. 2008. Predicting aged beef tenderness with a hyperspectral imaging system. Nebraska Beef Cattle Report 137-139. Cluff, K., G. Konda Naganathan, J. Subbiah, C. Calkins and A. Samal. 2008. Beef tenderness evaluation using optical scattering with near-infrared (NIR) hyperspectral imaging. ASABE Annual International Meeting, Providence, RI Paper No. 085220. Konda Naganathan, G., L. Grimes, J. Subbiah, C. Calkins and A. Samal. 2008. Global and local principal component analysis of near-infrared hyperspectral images for beef tenderness evaluation. ASABE Annual International Meeting, Providence, RI Paper No. 085219.
|
Progress 10/01/06 to 09/30/07
Outputs
OUTPUTS: Modeling to Improve Safety of Shell Eggs: The overall objective of this study was to predict the growth of Salmonella Enteritidis (SE) in shell eggs under non-isothermal conditions such as cooling. To meet this overall objective, a dynamic microbial model to predict the growth of SE in egg yolk based on time-temperature profile and a heat transfer model to predict the center-point temperature of egg during chilling were developed. Growth data of SE in egg yolk was collected at constant temperatures of 10, 15, 20, 25, 30, 35, 37, 39, 41, and 43 degrees Celcius. A dynamic microbial model was developed by integrating the Baranyi model (primary model) and modified Ratkowsky equation (secondary model). The integrated dynamic model was then validated at several non-isothermal profiles such as exponential cooling, exponential heating, linear cooling, and sinusoidal. A two-dimensional axisymmetric transient heat transfer model was developed to study the temperature distribution
inside the shell egg during cooling. The model was solved using finite element numerical method in FEMLAB package with a time step of 30 s. The developed heat transfer model was validated for five ambient isothermal conditions and for three different exponential cooling scenarios. Growth of Enterobacter sakazakii in Infant Formula: Enterobacter sakazakii has been implicated in foodborne illnesses in infants and neonates with infant formula as the vehicle. The objective of this study is to develop predictive models for the growth of E. sakazakii in sterile milk and soy infant formula. Growth kinetics for a five strain cocktail of E. sakazakii was obtained at static temperatures of 8.5, 10, 15, 20, 25, 28, 32, 35, 37, 42, 45, and 47 degrees Celcius in reconstituted milk and soy infant formula. Primary and secondary models were developed. Hyperspectral Imaging to Predict Beef Tenderness: Beef tenderness is an important quality attribute for consumer satisfaction. The objective of this
study was to implement hyperspectral imaging to predict 14 day aged, cooked beef tenderness. A pushbroom hyperspectral imaging system (spectral wavelength : 400 - 1000 nm) with diffuse-flood lighting system was developed; spatial, spectral and reflectance calibrations were performed. Hyperspectral images of beef steaks (n=111) at 14 day postmortem were acquired. Slice shear force (SSF) values were used as a tenderness reference. Principal component analysis was carried out on the hyperspectral images to reduce dimension along the spectral axis. The first five principal components explained over 90% variance of all bands in the image. On the principal component images, co-occurrence matrix analysis was conducted to statistically extract textural features. A canonical discriminant model was developed to predict three beef tenderness categories namely tender (SSF=< 21 kg), intermediate (SSF=21.1 to 25.9 kg), and tough (SSF>=26 kg).
PARTICIPANTS: Gumudavelli, V - Graduate Student, University of Nebraska-Lincoln. Konda Naganathan, G - Graduate Student, University of Nebraska-Lincoln. Wesseling, A - Graduate Student, University of Nebraska-Lincoln. Velugoti, P.R - Visiting scientist, University of Nebraska-Lincoln. Thippareddi, H - Collaborator, University of Nebraska-Lincoln. Calkins, C.R - Collaborator, University of Nebraska-Lincoln. Samal, A - Collaborator, University of Nebraska-Lincoln.
TARGET AUDIENCES: Meat processors, egg processor, meat packing plants, cattle producers.
Impacts
Modeling to Improve Safety of Shell Eggs: Root mean squared error (RMSE) were less than 0.44 log10CFU/g and pseudo-R2 values were greater than 0.98 for the primary model fitting. For the secondary model, the root mean squared error and pseudo-R2 were 0.05 h-1 and 0.99, respectively. The dynamic model performed well with RMSE values less than 0.29 log10CFU/g for various temperature profiles, and can be used independently for prediction of SE in shell eggs either at a given static temperature or any other continuously varying temperature profile. The predictive model can be used to evaluate SE growth risk in case of temperature abuses during shell egg storage and distribution. For the heat transfer model, RMSE values between the predicted and observed temperatures for all the trials at constant ambient air velocities were found to be in the range of 0.42 to 0.91 degrees Celcius. The RMSE values for three different exponentially cooling ambient air temperature profiles
were in the range of 0.18 to 0.43 degrees Celcius. Predicted temperatures from the model agreed well with the experimental values collected at the center of the shell egg. The heat transfer models can be used for evaluating cooling rates of shell eggs stored at different ambient temperatures or under dynamic cooling rates. Integration of the heat transfer model with a dynamic microbial growth model can evaluate the risks of pathogen growth under time-varying temperature conditions such as cooling. Thus, this model can assist in establishing safe shell egg cooling regimes to minimize the risks of pathogen growth after the eggs are laid. Growth of Enterobacter sakazakii in Infant Formula: For primary model fitting of growth of E. sakazakii in milk formula, pseudo-R2 and RMSE ranged from 0.88 to 0.99 and 0.26 to 0.45 log CFU/mL, respectively. For soy formula, pseudo-R2 and RMSE ranged from 0.68 to 0.99 and 0.27 to 1.01 log CFU/mL, respectively. The model did not fit the growth data well
at 47 degrees Celcius in which E. sakazakii growth was erratic. For secondary model fitting of growth of E. sakazakii in milk formula, pseudo-R2 and RMSE were 0.99 and 0.07 h-1, respectively, and the corresponding values for the soy formula were 0.99 and 0.08 h-1, respectively. Primary and secondary models can be integrated and solved numerically to determine the growth of E. sakazakii at varying temperature profiles. The developed models can help in improving microbial risk assessment and developing appropriate risk management strategies. Hyperspectral Imaging to Predict Beef Tenderness: With a leave-one-out cross-validation procedure, the model predicted the three tenderness categories with an accuracy of 96.4% (Table 2). All five tough carcasses were correctly identified. Three out of 12 intermediate carcasses were misclassified as tender. Only one tender carcass was misclassified as intermediate. Hyperspectral imaging shows promise for predicting beef tenderness.
Publications
- Gumudavelli, V. 2006. An integrated model for heat transfer and dynamic growth of Salmonella Enteritidis in shell eggs. M.S. Thesis. Food Science & Technology, University of Nebraska-Lincoln.
- Gumudavelli, V., J. Subbiah, H. Thippareddi, and P.R. Velugoti. 2007. Dynamic Predictive Model for Growth of Salmonella Enteritidis in Egg Yolk. International Journal of Food Microbiology, 72(7): M254-M262
- Gumudavelli, V., J. Subbiah, H. Thippareddi, L. Wang, and C.Weller. 2007. Finite Element Modeling of Heat Transfer in Shell Eggs. Presented at the ASABE Annual Meeting, Minneapolis, MN.
- Gumudavelli, V., J. Subbiah, H. Thippareddi, and C.Weller. 2007. Development of an Integrated Model of Dynamic Growth of Salmonella Enteritidis and Heat Transfer in Shell Eggs. Presented at the ASABE Annual Meeting, Minneapolis, MN.
- Wesseling, A., P.R. Velugoti, J. Subbiah, H. Thippareddi. 2007. Development of a Predictive Model for the Growth of Enterobacter sakazakii in Reconstituted Milk and Soy Infant Formula. Presented at IFT Annual Meeting, Chicago, IL.
- Konda Naganathan, G., J. Subbiah, C.R. Calkins, and A. Samal. 2006. VNIR imaging for beef tenderness prediction. Paper No. 063036, Annual ASABE International Meeting, Portland, OR.
- Konda Naganathan, G. 2007. Prediction of beef tenderness using hyperspectral imaging. M.S. Thesis. Biological Systems Engineering, University of Nebraska-Lincoln.
|
|