NIR Prediction of Pork Tenderness
7/8/2013
Pork: NIR Prediction of Pork Tenderness NIR prediction of pork tenderness Steven Shackelford, Ph.D., Scientist, Meat Safety and Quality Research Unit, U.S. Meat Animal Research Center
Steven Shackelford, Andy King, and Tommy Wheeler USDA-ARS U.S. Meat Animal Research Center Clay Center, NE
1. Slice shear force
Background 1. Slice shear force 2. Fresh (never frozen) vs Frozen (and thawed) 3. Noninvasive tenderness prediction system for beef longissimus using visible and near-infrared (VISNIR) spectroscopy 4. VISNIR for pork 5. NPB tenderness prediction RFP/Project
1. Longissimus slice shear force Rep B slice shear force, kg
A. Beef – single measurement from 1 steak B. Pork – duplicate measurement using a single slice from each of 2 chops C. Lamb – single measurement (half slices from each of 2 chops combined)
Pork
40
Repeatability = 0.90 n = 744
35 30 25 20 15
Mean = 15.3 SD = 5.0
10 5 0 0
5
10
15
20
25
30
35
40
Rep A slice shear force, kg
1
7/8/2013
Correlation of WBSF and SSF with SPT 8
8
7
7
6
6
panel tenderness Sensory Tenderness
panel tenderness Sensory Tenderness
Correlation of pork longissimus SSF with SPT
5 4 3
r = -.77 n = 479
2 1
1
2
3
4
5
6
7
8
2
r = -.82 n = 479
Slice shear force, kg
Sensory panel tenderness
SSF values are much greater than WBS values
Sensory panel tenderness
Sensory panel tenderness
Sensory panel tenderness
3
0 5 10 15 20 25 30 35 40 45 50
Warner-Bratzler shear force, kg
2. Fresh (never frozen) vs Frozen (and thawed)
4
1 0
SSF values are much greater than WBS values
5
Effect of freezing and thawing on beef longissimus slice shear force
A. Most loin chops/loins are sold fresh and most of those loin chops/loins are likely cooked fresh. B. If freezing and thawing alters tenderness, then we are likely biasing effects with freezing and thawing. C. The bias in most cases would be an underestimate of treatment differences.
2
7/8/2013
Effect of freezing and thawing on pork longissimus slice shear force
Effect of freezing and thawing on beef longissimus slice shear force
Correlation of WBSF and SSF with SPT
2. Fresh (never frozen) vs Frozen (and thawed)
8
If SSF has the same ability 8to predict SPT 7 as WBS, why do we use SSF?
6 5
panel tenderness Sensory Tenderness
panel tenderness Sensory Tenderness
7
6 5
Throughput, which allows for large scale 4 (fresh) evaluations!
4 3
r = -.77 n = 479
2 1
3 2
r = -.82 n = 479
1 0
1
2
3
4
5
6
7
Warner-Bratzler shear force, kg
8
0 5 10 15 20 25 30 35 40 45 50 Slice shear force, kg
A. Most loin chops/loins are sold fresh and most of those loin chops/loins are likely cooked fresh. B. If freezing and thawing alters tenderness, then we are likely biasing effects with freezing and thawing. C. The bias in most cases would be an underestimate of treatment differences. D. Critical consideration when developing systems to control “tough” samples
3. Prediction of beef longissimus tenderness with VISNIR spectroscopy
4. VISNIR for pork A. Began work on VISNIR for pork in 2001 as a part of pork quality genetics evaluation. i.
We were unsuccessful, because there was very little “tough” pork!
3
7/8/2013
4. VISNIR for pork
4. VISNIR for pork
B. In 2007, we were approached by the industry to adapt VISNIR for pork.
B. In 2007, we were approached by the industry to adapt VISNIR for pork. i.
If there is very little “tough” pork, why should we do that?
Materials and Methods
4. VISNIR for pork B. In 2007, we were approached by the industry to adapt VISNIR for pork. i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that?
4. VISNIR for pork
4. VISNIR for pork
B. In 2007, we were approached by the industry to adapt VISNIR for pork.
B. In 2007, we were approached by the industry to adapt VISNIR for pork.
i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that? iii. Guess what happened ?
i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that? iii. Guess what happened ? Initially, we were unsuccessful, because there was very little “tough” pork!
4
7/8/2013
Materials and Methods
4. VISNIR for pork B. In 2007, we were approached by the industry to adapt VISNIR for pork. i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that? iii. Guess what happened ? Initially, we were unsuccessful, because there was very little “tough” pork! iv. Continued to build data set over the course of two years
Boneless pork loins (n = 901) were evaluated either on line on the loin boning and trimming line of large-scale commercial plants (n = 465) or at the U.S. Meat Animal Research Center abattoir (n = 436).
Materials and Methods Boneless loin sections were aged (2°C) until 14 days postmortem and two 2.54-cm thick chops were obtained from the 11th rib region. Fresh (never frozen) chops were cooked (71°C) and longissimus slice shear force (SSF) was measured on each of the two chops.
Those two values were averaged and that value was used for all analyses.
Materials and Methods Carcasses were blocked by • plant (n = 3), • production day (n = 24), • and observed SSF Mean = 13.9 kg; SD= 3.7; CV = 26.8%; Range 6.4 to 32.4 kg)
Materials and Methods
Calibration data set
65
n = 451 Comparison of means; SEM = 0.3 kg; P < 10-3 Comparison of % > 20 kg; P = 0.06
60 55 50
VISNIR predicted tender Mean SSF = 13.4 kg SD = 3.2 kg Range 8.1 - 28.2 kg 3.6% > 20 kg n = 279
45
Frequency, %
One-half of the carcasses were assigned to a calibration data set (CDS), which was used to develop regression equations, and one-half of the carcasses were assigned to a prediction data set (PDS), which was used to validate the regression equations.
Results - CDS
70
40 35 30
VISNIR not predicted tender Mean SSF = 14.5 kg SD = 4.1 kg Range 6.4 - 30.0 kg 8.1% > 20 kg n = 172
25 20 15 10
5 0 5 to 10
10 to 15
15 20 to to 20 25 Slice shear force at 14 d postmortem, kg
25 to 30
30 to 35
5
7/8/2013
Results - PDS
70
n = 450 Comparison of means; SEM = 0.3 kg; P < 10-7 Comparison of % > 20 kg; P < 10-5
60 55 50
VISNIR predicted tender Mean SSF = 13.3 kg SD = 2.8 kg Range 7.9 - 25.5 kg 1.8% > 20 kg n = 274
45 40 35 30
VISNIR not predicted tender Mean SSF = 15.2 kg SD = 4.9 kg Range 6.8 - 32.4 kg 13.6% > 20 kg n = 176
25 20 15 10
5
Observed longissimus intramuscular fat percentage
65
Frequency, %
Results – IMF prediction 8
Prediction data set
VISNIR conducted on ventral side of longissimus Calibration data set R^2 = 0.62 RSD = 0.62% n = 451
7 6
Prediction data set R^2 = 0.63 RSD = 0.63% n = 450
5
R² = 0.62 R² = 0.63
4 3 2 1
0 5 to 10
10 to 15
15 20 to to 20 25 Slice shear force at 14 d postmortem, kg
25 to 30
30 to 35
5. National Pork Board RFP/Project A. In response to the results of the NPB Consumer Preference Study, NPB issued a RFP for development of systems to predict pork tenderness in 2009 B. USMARC submitted a proposal to field test our VISNIR system C. TAMU submitted a proposal to test VISNIR and some less well-developed technologies. D. NPB staff encouraged us to combine efforts E. Collaborative effort was developed to maximize resources
5. National Pork Board RFP/Project
0 0
1
2
3
iv.
A total of 1,208 loins were sampled. a. b. c. d.
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009 Plant 3 -- 304 loins were sampled on Jan 12, 2010 Plant 4 -- 304 loins were sampled on Jan 14, 2010
5
6
7
8
5. National Pork Board RFP/Project F. USMARC established collaboration with four large-scale commercial packing plants i.
ii. iii.
To maximize the likelihood that we sampled ample variation in tenderness for testing the tenderness prediction systems, packing plants were selected to represent a diversity of processing systems and hog sources Two plants had conventional chilling systems with spray chill and two plants had blast-chilling systems All four plants had similar CO2 stunning systems
5. National Pork Board RFP/Project VISNIR predicted SSF class
F.
4
Predicted longissimus intramuscular fat percentage
Plant 1
Plant 2
Plant 3
Plant 4
1_< 13.4 kg 2_13.4 to 13.8 kg 3_13.8 to 14.2 kg 4_> 14.2 kg Grand Total
6
7/8/2013
5. National Pork Board RFP/Project
5. National Pork Board RFP/Project VISNIR predicted SSF class 1_< 13.4 kg
F. iv.
A total of 1,208 loins were sampled. a.
Plant 3
Plant 4
139
2_13.4 to 13.8 kg
80
3_13.8 to 14.2 kg
47
4_> 14.2 kg
34
Grand Total
300
5. National Pork Board RFP/Project VISNIR predicted SSF class
F.
Plant 1
Plant 2
139
26
2_13.4 to 13.8 kg
80
39
3_13.8 to 14.2 kg
47
62
4_> 14.2 kg
34
173
Grand Total
300
300
1_< 13.4 kg
A total of 1,208 loins were sampled. a. b.
Plant 2
Plant 1 -- 300 loins were sampled on Nov 17, 2009
5. National Pork Board RFP/Project
iv.
Plant 1
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009
5. National Pork Board RFP/Project
Plant 3
Plant 4
5. National Pork Board RFP/Project
G. Loin processing i.
ii. iii. iv. v.
vi.
Center-cut boneless loins were identified and USMARC conducted VISNIR on-line during or immediately following boneless loin production Loins were captured in combo bins Meat quality measurements and a second VISNIR measurement was obtained approximately 1 hour later to help facilitate development of a robust model Center-cut boneless loins were vacuum-packaged, boxed, and transported (-2.8ºC) to USMARC Loins arrived at USMARC within 12 hours of production and were immediately placed in a holding cooler, unboxed, sorted, inventoried, and placed on carts for aging (1.5ºC) The following day, vacuum-packaged loins were weighed for subsequent purge loss determination.
G. Loin processing vii. 14 days postmortem, loins were unpackaged viii. VISNIR spectra collected by USMARC
7
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. 14 days postmortem, loins were unpackaged viii. VISNIR spectra collected by USMARC ix. Impedance evaluated by TAMU
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss
8
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display
9
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem)
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b. c.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem)
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b. c. d.
Measurements made with two instruments
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem) 1 Chop for NIR, etc.
10
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b. c. d. e.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem) 1 Chop for NIR, etc. 1 Chop for Instrumental Color, pH, IMF (frozen) by TAMU
5. National Pork Board RFP/Project G. Loin processing
a. b. c. d. e. f.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut
1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem) 1 Chop for NIR, etc. 1 Chop for Instrumental Color, pH, IMF (frozen) by TAMU 1 Chop for CSU ??? and subsample for TMC ???
Plant differences in slice shear force NPB Project
70 60
Results
50 Frequency, %
vii. viii. ix. x. xi. xii.
Plant 1 Mean = 13.6 kg 1.3% > 25 kg n = 300
40
Plant 2 Mean = 18.8 kg 15.7% > 25 kg n = 300
30 20 10 0 < 10
10 to 15
15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 d postmortem, kg
> 45
11
7/8/2013
5. National Pork Board RFP/Project
5. National Pork Board RFP/Project VISNIR predicted SSF class
F. iv.
Plant 2
Plant 3
139
26
41
2_13.4 to 13.8 kg
80
39
45
3_13.8 to 14.2 kg
47
62
93
4_> 14.2 kg
34
173
125
Grand Total
300
300
304
1_< 13.4 kg
A total of 1,208 loins were sampled. a. b. c.
Plant 1
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009 Plant 3 -- 304 loins were sampled on Jan 12, 2010
5. National Pork Board RFP/Project
5. National Pork Board RFP/Project VISNIR predicted SSF class
F. iv.
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009 Plant 3 -- 304 loins were sampled on Jan 12, 2010 Plant 4 -- 304 loins were sampled on Jan 14, 2010
NPB Project
Frequency, %
Plant 1 Mean = 13.6 kg 1.3% > 25 kg n = 300
Plant 4 Mean = 14.3 kg 1.6% > 25 kg n = 304
Plant 2 Mean = 18.8 kg 15.7% > 25 kg n = 300
Plant 3 Mean = 20.7 kg 24.7% > 25 kg n = 304
40 30 20
Plant 3
Plant 4
139
26
41
60
2_13.4 to 13.8 kg
80
39
45
52
3_13.8 to 14.2 kg
47
62
93
80
4_> 14.2 kg
34
173
125
112
Grand Total
300
300
304
304
• Source of plant differences in tenderness will be discussed this afternoon during the reciprocation session entitled “The effect of chilling rate on tenderness” • But, implications for VISNIR will be discussed here
60 50
Plant 2
5. National Pork Board RFP/Project
Plant differences in slice shear force 70
Plant 1
1_< 13.4 kg
A total of 1,208 loins were sampled. a. b. c. d.
Plant 4
10 0 < 10
10 to 15
15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 d postmortem, kg
> 45
12
7/8/2013
Previous tenderness prediction model
Split-overall
Experiment 2 - predicted with model from Experinent 1 Comparison of means; SEM = 0.24 kg; P < 10-11 Comparison of % > 25 kg; P < 10-9
55 50
45 VISNIR predicted tender Mean SSF = 15.7 kg SD = 5.1 kg Range 7.9 to 47.1 kg 5.5% > 25 kg n = 604
40
Frequency, %
35
30
Plant
Predicted tender
1
247
Not predicted tender 53
2
94
206
3
121
183
4
142
162
25 VISNIR not predicted tender Mean SSF = 18.1 kg SD = 6.5 kg Range 7.7 to 40.6 kg 16.2% > 25 kg n = 604
20
15 10
This VISNIR model was not completely reflective of tenderness differences among plants
5 0 < 10
10 to 15 15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 days postmortem, kg
> 45kg
Average spectra for each plant
Correlation analysis
0.40 0.35 0.30
Reflectance
0.25 Plant 1
0.20
Plant 2
Plant 3 0.15
Plant 4
0.10 0.05
• To develop a robust model that would properly reflect tenderness variation among and within packing plants, correlation analysis was conducted for each plant to identify the wavelength range at which reflectance was most highly related to SSF. • For each plant, the strongest correlation was found at or near 822 nm.
0.00 400
450
500
550
600
650 700 750 Wavelength, nm
800
850
900
950
1000
Single variable model
Correlation analysis
55 50
45
VISNIR predicted tender Mean SSF = 15.1 kg SD = 4.5 kg Range 7.7 to 32.4 kg 3.6% > 25 kg n = 604
40
Frequency, %
• Also, variation in the plant means for reflectance at 822 nm accounted for virtually all of the variation in plant means for SSF (r = -0.99). • Thus, reflectance at 822 nm was indicative of variation in tenderness both among and within plants.
Experiment 2 - predicted with reflectance at 822 nm Comparison of means; SEM = 0.23 kg; P < 10-26 Comparison of % > 25 kg; P < 10-16
60
35 30 25
VISNIR not predicted tender Mean SSF = 18.7 kg SD = 6.6 kg Range 7.9 to 47.1 kg 18.0% > 25 kg n = 604
20 15 10
5 0 < 10
10 to 15 15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 days postmortem, kg
> 45kg
13
7/8/2013
Follow-up project single variable model Experiment 3 - predicted with reflectance at 822 nm Comparison of means; SEM = 0.28 kg; P < 10-9 Comparison of % > 25 kg; P < 10-5
50 45
• It works • So why hasn’t it been implemented
40 VISNIR predicted tender Mean SSF = 16.1 kg SD = 4.1 kg Range 8.4 to 37.8 kg 3.0% > 25 kg n = 300
Frequency, %
35 30
25 20
Conclusion
– Not for a lack of interest by the industry – Instrument supplier
VISNIR not predicted tender Mean SSF = 18.6 kg SD = 5.3 kg Range 8.5 to 37.9 kg 12.4% > 25 kg n = 299
15 10 5 0 < 10
10 to 15 15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 days postmortem, kg
Conclusion
> 45kg
Thank you
• It works • So why hasn’t it been implemented – Not for a lack of interest by the industry – Instrument supplier – We have sought other instruments capable of being used for beef and pork.
14
Pork: NIR Prediction of Pork Tenderness NIR prediction of pork tenderness Steven Shackelford, Ph.D., Scientist, Meat Safety and Quality Research Unit, U.S. Meat Animal Research Center
Steven Shackelford, Andy King, and Tommy Wheeler USDA-ARS U.S. Meat Animal Research Center Clay Center, NE
1. Slice shear force
Background 1. Slice shear force 2. Fresh (never frozen) vs Frozen (and thawed) 3. Noninvasive tenderness prediction system for beef longissimus using visible and near-infrared (VISNIR) spectroscopy 4. VISNIR for pork 5. NPB tenderness prediction RFP/Project
1. Longissimus slice shear force Rep B slice shear force, kg
A. Beef – single measurement from 1 steak B. Pork – duplicate measurement using a single slice from each of 2 chops C. Lamb – single measurement (half slices from each of 2 chops combined)
Pork
40
Repeatability = 0.90 n = 744
35 30 25 20 15
Mean = 15.3 SD = 5.0
10 5 0 0
5
10
15
20
25
30
35
40
Rep A slice shear force, kg
1
7/8/2013
Correlation of WBSF and SSF with SPT 8
8
7
7
6
6
panel tenderness Sensory Tenderness
panel tenderness Sensory Tenderness
Correlation of pork longissimus SSF with SPT
5 4 3
r = -.77 n = 479
2 1
1
2
3
4
5
6
7
8
2
r = -.82 n = 479
Slice shear force, kg
Sensory panel tenderness
SSF values are much greater than WBS values
Sensory panel tenderness
Sensory panel tenderness
Sensory panel tenderness
3
0 5 10 15 20 25 30 35 40 45 50
Warner-Bratzler shear force, kg
2. Fresh (never frozen) vs Frozen (and thawed)
4
1 0
SSF values are much greater than WBS values
5
Effect of freezing and thawing on beef longissimus slice shear force
A. Most loin chops/loins are sold fresh and most of those loin chops/loins are likely cooked fresh. B. If freezing and thawing alters tenderness, then we are likely biasing effects with freezing and thawing. C. The bias in most cases would be an underestimate of treatment differences.
2
7/8/2013
Effect of freezing and thawing on pork longissimus slice shear force
Effect of freezing and thawing on beef longissimus slice shear force
Correlation of WBSF and SSF with SPT
2. Fresh (never frozen) vs Frozen (and thawed)
8
If SSF has the same ability 8to predict SPT 7 as WBS, why do we use SSF?
6 5
panel tenderness Sensory Tenderness
panel tenderness Sensory Tenderness
7
6 5
Throughput, which allows for large scale 4 (fresh) evaluations!
4 3
r = -.77 n = 479
2 1
3 2
r = -.82 n = 479
1 0
1
2
3
4
5
6
7
Warner-Bratzler shear force, kg
8
0 5 10 15 20 25 30 35 40 45 50 Slice shear force, kg
A. Most loin chops/loins are sold fresh and most of those loin chops/loins are likely cooked fresh. B. If freezing and thawing alters tenderness, then we are likely biasing effects with freezing and thawing. C. The bias in most cases would be an underestimate of treatment differences. D. Critical consideration when developing systems to control “tough” samples
3. Prediction of beef longissimus tenderness with VISNIR spectroscopy
4. VISNIR for pork A. Began work on VISNIR for pork in 2001 as a part of pork quality genetics evaluation. i.
We were unsuccessful, because there was very little “tough” pork!
3
7/8/2013
4. VISNIR for pork
4. VISNIR for pork
B. In 2007, we were approached by the industry to adapt VISNIR for pork.
B. In 2007, we were approached by the industry to adapt VISNIR for pork. i.
If there is very little “tough” pork, why should we do that?
Materials and Methods
4. VISNIR for pork B. In 2007, we were approached by the industry to adapt VISNIR for pork. i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that?
4. VISNIR for pork
4. VISNIR for pork
B. In 2007, we were approached by the industry to adapt VISNIR for pork.
B. In 2007, we were approached by the industry to adapt VISNIR for pork.
i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that? iii. Guess what happened ?
i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that? iii. Guess what happened ? Initially, we were unsuccessful, because there was very little “tough” pork!
4
7/8/2013
Materials and Methods
4. VISNIR for pork B. In 2007, we were approached by the industry to adapt VISNIR for pork. i.
If there is very little “tough” pork, why should we do that? ii. Given that pork is not ribbed, how can we do that? iii. Guess what happened ? Initially, we were unsuccessful, because there was very little “tough” pork! iv. Continued to build data set over the course of two years
Boneless pork loins (n = 901) were evaluated either on line on the loin boning and trimming line of large-scale commercial plants (n = 465) or at the U.S. Meat Animal Research Center abattoir (n = 436).
Materials and Methods Boneless loin sections were aged (2°C) until 14 days postmortem and two 2.54-cm thick chops were obtained from the 11th rib region. Fresh (never frozen) chops were cooked (71°C) and longissimus slice shear force (SSF) was measured on each of the two chops.
Those two values were averaged and that value was used for all analyses.
Materials and Methods Carcasses were blocked by • plant (n = 3), • production day (n = 24), • and observed SSF Mean = 13.9 kg; SD= 3.7; CV = 26.8%; Range 6.4 to 32.4 kg)
Materials and Methods
Calibration data set
65
n = 451 Comparison of means; SEM = 0.3 kg; P < 10-3 Comparison of % > 20 kg; P = 0.06
60 55 50
VISNIR predicted tender Mean SSF = 13.4 kg SD = 3.2 kg Range 8.1 - 28.2 kg 3.6% > 20 kg n = 279
45
Frequency, %
One-half of the carcasses were assigned to a calibration data set (CDS), which was used to develop regression equations, and one-half of the carcasses were assigned to a prediction data set (PDS), which was used to validate the regression equations.
Results - CDS
70
40 35 30
VISNIR not predicted tender Mean SSF = 14.5 kg SD = 4.1 kg Range 6.4 - 30.0 kg 8.1% > 20 kg n = 172
25 20 15 10
5 0 5 to 10
10 to 15
15 20 to to 20 25 Slice shear force at 14 d postmortem, kg
25 to 30
30 to 35
5
7/8/2013
Results - PDS
70
n = 450 Comparison of means; SEM = 0.3 kg; P < 10-7 Comparison of % > 20 kg; P < 10-5
60 55 50
VISNIR predicted tender Mean SSF = 13.3 kg SD = 2.8 kg Range 7.9 - 25.5 kg 1.8% > 20 kg n = 274
45 40 35 30
VISNIR not predicted tender Mean SSF = 15.2 kg SD = 4.9 kg Range 6.8 - 32.4 kg 13.6% > 20 kg n = 176
25 20 15 10
5
Observed longissimus intramuscular fat percentage
65
Frequency, %
Results – IMF prediction 8
Prediction data set
VISNIR conducted on ventral side of longissimus Calibration data set R^2 = 0.62 RSD = 0.62% n = 451
7 6
Prediction data set R^2 = 0.63 RSD = 0.63% n = 450
5
R² = 0.62 R² = 0.63
4 3 2 1
0 5 to 10
10 to 15
15 20 to to 20 25 Slice shear force at 14 d postmortem, kg
25 to 30
30 to 35
5. National Pork Board RFP/Project A. In response to the results of the NPB Consumer Preference Study, NPB issued a RFP for development of systems to predict pork tenderness in 2009 B. USMARC submitted a proposal to field test our VISNIR system C. TAMU submitted a proposal to test VISNIR and some less well-developed technologies. D. NPB staff encouraged us to combine efforts E. Collaborative effort was developed to maximize resources
5. National Pork Board RFP/Project
0 0
1
2
3
iv.
A total of 1,208 loins were sampled. a. b. c. d.
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009 Plant 3 -- 304 loins were sampled on Jan 12, 2010 Plant 4 -- 304 loins were sampled on Jan 14, 2010
5
6
7
8
5. National Pork Board RFP/Project F. USMARC established collaboration with four large-scale commercial packing plants i.
ii. iii.
To maximize the likelihood that we sampled ample variation in tenderness for testing the tenderness prediction systems, packing plants were selected to represent a diversity of processing systems and hog sources Two plants had conventional chilling systems with spray chill and two plants had blast-chilling systems All four plants had similar CO2 stunning systems
5. National Pork Board RFP/Project VISNIR predicted SSF class
F.
4
Predicted longissimus intramuscular fat percentage
Plant 1
Plant 2
Plant 3
Plant 4
1_< 13.4 kg 2_13.4 to 13.8 kg 3_13.8 to 14.2 kg 4_> 14.2 kg Grand Total
6
7/8/2013
5. National Pork Board RFP/Project
5. National Pork Board RFP/Project VISNIR predicted SSF class 1_< 13.4 kg
F. iv.
A total of 1,208 loins were sampled. a.
Plant 3
Plant 4
139
2_13.4 to 13.8 kg
80
3_13.8 to 14.2 kg
47
4_> 14.2 kg
34
Grand Total
300
5. National Pork Board RFP/Project VISNIR predicted SSF class
F.
Plant 1
Plant 2
139
26
2_13.4 to 13.8 kg
80
39
3_13.8 to 14.2 kg
47
62
4_> 14.2 kg
34
173
Grand Total
300
300
1_< 13.4 kg
A total of 1,208 loins were sampled. a. b.
Plant 2
Plant 1 -- 300 loins were sampled on Nov 17, 2009
5. National Pork Board RFP/Project
iv.
Plant 1
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009
5. National Pork Board RFP/Project
Plant 3
Plant 4
5. National Pork Board RFP/Project
G. Loin processing i.
ii. iii. iv. v.
vi.
Center-cut boneless loins were identified and USMARC conducted VISNIR on-line during or immediately following boneless loin production Loins were captured in combo bins Meat quality measurements and a second VISNIR measurement was obtained approximately 1 hour later to help facilitate development of a robust model Center-cut boneless loins were vacuum-packaged, boxed, and transported (-2.8ºC) to USMARC Loins arrived at USMARC within 12 hours of production and were immediately placed in a holding cooler, unboxed, sorted, inventoried, and placed on carts for aging (1.5ºC) The following day, vacuum-packaged loins were weighed for subsequent purge loss determination.
G. Loin processing vii. 14 days postmortem, loins were unpackaged viii. VISNIR spectra collected by USMARC
7
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. 14 days postmortem, loins were unpackaged viii. VISNIR spectra collected by USMARC ix. Impedance evaluated by TAMU
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss
8
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display
9
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem)
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b. c.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem)
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b. c. d.
Measurements made with two instruments
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem) 1 Chop for NIR, etc.
10
7/8/2013
5. National Pork Board RFP/Project G. Loin processing vii. viii. ix. x. xi. xii. a. b. c. d. e.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut 1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem) 1 Chop for NIR, etc. 1 Chop for Instrumental Color, pH, IMF (frozen) by TAMU
5. National Pork Board RFP/Project G. Loin processing
a. b. c. d. e. f.
14 days postmortem, loins were unpackaged VISNIR spectra collected by USMARC Impedance evaluated by TAMU VISNIR spectra collected by TAMU Loins weighed for purge loss Chops cut
1 Chop overwrapped and placed in simulated retail display 2 Chops for SSF (next day … 15 d postmortem) 2 Chops for WBSF (frozen … 14 d postmortem) 1 Chop for NIR, etc. 1 Chop for Instrumental Color, pH, IMF (frozen) by TAMU 1 Chop for CSU ??? and subsample for TMC ???
Plant differences in slice shear force NPB Project
70 60
Results
50 Frequency, %
vii. viii. ix. x. xi. xii.
Plant 1 Mean = 13.6 kg 1.3% > 25 kg n = 300
40
Plant 2 Mean = 18.8 kg 15.7% > 25 kg n = 300
30 20 10 0 < 10
10 to 15
15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 d postmortem, kg
> 45
11
7/8/2013
5. National Pork Board RFP/Project
5. National Pork Board RFP/Project VISNIR predicted SSF class
F. iv.
Plant 2
Plant 3
139
26
41
2_13.4 to 13.8 kg
80
39
45
3_13.8 to 14.2 kg
47
62
93
4_> 14.2 kg
34
173
125
Grand Total
300
300
304
1_< 13.4 kg
A total of 1,208 loins were sampled. a. b. c.
Plant 1
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009 Plant 3 -- 304 loins were sampled on Jan 12, 2010
5. National Pork Board RFP/Project
5. National Pork Board RFP/Project VISNIR predicted SSF class
F. iv.
Plant 1 -- 300 loins were sampled on Nov 17, 2009 Plant 2 -- 300 loins were sampled on Nov 19, 2009 Plant 3 -- 304 loins were sampled on Jan 12, 2010 Plant 4 -- 304 loins were sampled on Jan 14, 2010
NPB Project
Frequency, %
Plant 1 Mean = 13.6 kg 1.3% > 25 kg n = 300
Plant 4 Mean = 14.3 kg 1.6% > 25 kg n = 304
Plant 2 Mean = 18.8 kg 15.7% > 25 kg n = 300
Plant 3 Mean = 20.7 kg 24.7% > 25 kg n = 304
40 30 20
Plant 3
Plant 4
139
26
41
60
2_13.4 to 13.8 kg
80
39
45
52
3_13.8 to 14.2 kg
47
62
93
80
4_> 14.2 kg
34
173
125
112
Grand Total
300
300
304
304
• Source of plant differences in tenderness will be discussed this afternoon during the reciprocation session entitled “The effect of chilling rate on tenderness” • But, implications for VISNIR will be discussed here
60 50
Plant 2
5. National Pork Board RFP/Project
Plant differences in slice shear force 70
Plant 1
1_< 13.4 kg
A total of 1,208 loins were sampled. a. b. c. d.
Plant 4
10 0 < 10
10 to 15
15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 d postmortem, kg
> 45
12
7/8/2013
Previous tenderness prediction model
Split-overall
Experiment 2 - predicted with model from Experinent 1 Comparison of means; SEM = 0.24 kg; P < 10-11 Comparison of % > 25 kg; P < 10-9
55 50
45 VISNIR predicted tender Mean SSF = 15.7 kg SD = 5.1 kg Range 7.9 to 47.1 kg 5.5% > 25 kg n = 604
40
Frequency, %
35
30
Plant
Predicted tender
1
247
Not predicted tender 53
2
94
206
3
121
183
4
142
162
25 VISNIR not predicted tender Mean SSF = 18.1 kg SD = 6.5 kg Range 7.7 to 40.6 kg 16.2% > 25 kg n = 604
20
15 10
This VISNIR model was not completely reflective of tenderness differences among plants
5 0 < 10
10 to 15 15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 days postmortem, kg
> 45kg
Average spectra for each plant
Correlation analysis
0.40 0.35 0.30
Reflectance
0.25 Plant 1
0.20
Plant 2
Plant 3 0.15
Plant 4
0.10 0.05
• To develop a robust model that would properly reflect tenderness variation among and within packing plants, correlation analysis was conducted for each plant to identify the wavelength range at which reflectance was most highly related to SSF. • For each plant, the strongest correlation was found at or near 822 nm.
0.00 400
450
500
550
600
650 700 750 Wavelength, nm
800
850
900
950
1000
Single variable model
Correlation analysis
55 50
45
VISNIR predicted tender Mean SSF = 15.1 kg SD = 4.5 kg Range 7.7 to 32.4 kg 3.6% > 25 kg n = 604
40
Frequency, %
• Also, variation in the plant means for reflectance at 822 nm accounted for virtually all of the variation in plant means for SSF (r = -0.99). • Thus, reflectance at 822 nm was indicative of variation in tenderness both among and within plants.
Experiment 2 - predicted with reflectance at 822 nm Comparison of means; SEM = 0.23 kg; P < 10-26 Comparison of % > 25 kg; P < 10-16
60
35 30 25
VISNIR not predicted tender Mean SSF = 18.7 kg SD = 6.6 kg Range 7.9 to 47.1 kg 18.0% > 25 kg n = 604
20 15 10
5 0 < 10
10 to 15 15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 days postmortem, kg
> 45kg
13
7/8/2013
Follow-up project single variable model Experiment 3 - predicted with reflectance at 822 nm Comparison of means; SEM = 0.28 kg; P < 10-9 Comparison of % > 25 kg; P < 10-5
50 45
• It works • So why hasn’t it been implemented
40 VISNIR predicted tender Mean SSF = 16.1 kg SD = 4.1 kg Range 8.4 to 37.8 kg 3.0% > 25 kg n = 300
Frequency, %
35 30
25 20
Conclusion
– Not for a lack of interest by the industry – Instrument supplier
VISNIR not predicted tender Mean SSF = 18.6 kg SD = 5.3 kg Range 8.5 to 37.9 kg 12.4% > 25 kg n = 299
15 10 5 0 < 10
10 to 15 15 to 20 20 to 25 25 to 30 30 to 35 35 to 40 40 to 45 Pork longissimus slice shear force at 15 days postmortem, kg
Conclusion
> 45kg
Thank you
• It works • So why hasn’t it been implemented – Not for a lack of interest by the industry – Instrument supplier – We have sought other instruments capable of being used for beef and pork.
14
