The relationship between video image analysis (VIA), visual classification, and saleable meat yield of sirloin and fillet cuts of beef carcasses differing in breed and gender

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

Author's personal copy

The relationship between video image analysis (VIA), visualclassification, and saleable meat yield of sirloin and fillet cutsof beef carcasses differing in breed and gender

C.R. Craigie a,b,n, D.W. Ross c, C.A. Maltin d, R.W. Purchas e, L. Bünger a,R. Roehe a, S.T. Morris b

a Animal and Veterinary Sciences Group, Scotland's Rural College, West Mains Road, Edinburgh EH9 3JG, Scotland, UKb Institute of Veterinary Animal and Biomedical Sciences, Massey University, Private Bag 11-222, Palmerston North, New Zealandc Future Farming Systems Group, Scotland's Rural College, West Mains Road, Edinburgh EH9 3JG, Scotland, UKd Quality Meat Scotland, The Rural Centre, Ingliston, Edinburgh EH28 8NZ, Scotland, UKe Institute of Food, Nutrition and Human Health, Massey University, Private Bag 11-222, Palmerston North 4442, New Zealand

a r t i c l e i n f o

Article history:Received 9 July 2013Received in revised form19 September 2013Accepted 20 September 2013

Keywords:BeefBreed differencesCarcass classificationGradingVIASaleable meat yield

a b s t r a c t

Carcass quality of 72 steers, 48 heifers and 21 bulls from continental and dairy genotypeswere compared on the basis of conformation, fatness and saleable meat yield (SMY%) ofthe fillet and trimmed boneless sirloin cuts. Comparisons between genotype and gendergroups showed that steers from beef breeds had higher EUROP conformation scores thanthose from dairy breeds which corresponded to a higher SMY% of sirloin and fillet. Resultssuggested that the EUROP grid may underestimate the sirloin SMY% of Charolais heifersdue to their higher muscle-to-bone ratio. Furthermore, the 141 carcasses were used toassess the accuracy with which video image analysis (VIA) and visual classification in acommercial abattoir predicted the weight and SMY% of the sirloin and fillet. Both VIA andthe visual carcass classification systems resulted in similar accuracies for prediction ofsirloin SMY% (R2¼58%, RSD¼0.35% for VIA and R2¼57%, RSD¼0.35% for visual classifica-tion) but fillet yield was poorly predicted by both VIA and visual classification systems.Including the weight of excess fat removed during sirloin fabrication as an additionalcovariate for sirloin SMY% prediction did not offer any substantial improvement inpredictive ability. Inclusion of bone weight as an additional covariate did show somepromise for improving the prediction accuracy of fillet SMY%. The fact that no statisticallysignificant correlations between fillet yield and EUROP carcass classification categories(assigned visually or by the VIA) could be identified after adjustment for genotype furtherjustifies the need to consider alternative modes of carcass evaluation to better reflect thedistribution of meat throughout the carcass.

Crown Copyright & 2013 Published by Elsevier B.V. All rights reserved.

1. Introduction

Within the European Union, adult bovine carcasses areclassified according to the EUROP carcass classification

scheme under the European Community regulations1208/81 and 2930/81 (European Community, 1981a,1981b). The EUROP classification scheme includes carcassconformation scores on a 15-point scale with 5 mainclasses, E (excellent conformation), U, R, O and P (poorconformation), and 10 sub-classes, and five main fatnessscores (1 (low fatness), 2, 3, 4 and 5 (excessive fatness))also with 10 sub-classes (Fisher, 2007). During develop-ment of the EUROP classification scheme in the early 1980sno attempt was made to relate it to the amount of lean

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/livsci

Livestock Science

1871-1413/$ - see front matter Crown Copyright & 2013 Published by Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.livsci.2013.09.014

n Correspondence to: Quality Meat Scotland, The Rural Centre,Ingliston, Edinburgh EH28 8NZ, UK. Tel.: þ44 131 472 4040;fax: þ44 131 472 4038.

E-mail addresses: [email protected],[email protected] (C.R. Craigie).

Livestock Science 158 (2013) 169–178

Author's personal copy

meat as a percentage of the dressed carcass, known assaleable meat yield (SMY%), which is lean meat sold with acertain amount of fat still attached (Kempster et al., 1980).The reason for the lack of alignment between EUROPclassification and yield traits was because there was nostandardised definition of SMY% (Allen, 2003). It has beenargued that a carcass classification system based on SMY%would give clear production objectives to producersthrough a value-based marketing system (Cross andWhittaker, 1992).

As a result, recent research has focused on relatingcarcass classification to SMY% (Conroy et al., 2009, 2010a,2010b). The EUROP scheme for carcass classification hasbeen widely criticised on account of its subjective (visual)mode, even when assessed by a trained classifier usingphotographic references. Although there is little evidence tosupport this criticism, it is impossible to demonstrate theobjectivity of the system as long as human classifiers areinvolved (Allen, 2003). To address this weakness, objectivecarcass classification methods based on video image analy-sis (VIA) were developed which allowed a large number ofvariables (lengths, widths, areas, volumes etc.) to be mea-sured on a carcass in a matter of seconds. VIA systems thatassess whole sides of beef can be installed on-line inabattoirs to operate autonomously at line speeds up to800 beef carcasses per hour (Ross et al., 2011) and up to 800lamb carcasses per hour (Rius-Vilarrasa et al., 2009).

Three different brands of commercially availablewhole-side VIA systems are used in Europe to classify beefcarcasses according to the EUROP conformation and fatclass grid, including the BCC-2 (Borggaard et al., 1996), VBS2000 (Augustini et al., 1997) and the Normaclass Machineà classer (MAC) (Allen, 2007). In an experiment under-taken by Teagasc in the Republic of Ireland, the VIAscan,VBS 2000 and BCC-2 systems were able to predict SMY%from the same carcasses with similar accuracies (residualstandard deviation (RSD) values between 1.1% and 1.2%(Allen and Finnerty, 2001). VIA has since been successfullyapplied to carcass classification in the Republic of Irelandon an industrial scale since 2004 using VBS 2000 machines(Allen, 2007; Pabiou et al., 2011). Many studies haveinvestigated the ability of the VBS 2000 to assess carcassesfrom a variety of cattle populations in Germany (Augustiniet al., 1997; Branscheid et al., 1998; Brinkmann, 2007;Sonnichsen et al., 2006), Norway (Jørgenvåg et al., 2009)and the Republic of Ireland (Allen, 2005; Allen andFinnerty, 2000, 2001; Pabiou et al., 2009, 2011). Most ofthese experiments have assessed the ability of VIA toclassify carcasses according to the EUROP grid, yet allwhole-side VIA systems are able to directly predict carcassSMY%. Even though SMY% is an important value compo-nent of the carcass, few studies have investigated thedirect prediction of SMY% using VIA under commercialabattoir conditions.

The objectives of the present study were as follows:

(i) Compare visual and VIA carcass classification using thetrimmed, boneless yield of sirloin and fillet cuts usingcarcasses from cattle differing in gender and genotype.

(ii) To assess the prediction of sirloin and fillet cut saleablemeat yield by visual and VIA classification.

2. Material and methods

2.1. Animals

Between March and May 2009, 141 cattle below 30months of age were selected for inclusion in the experi-ment at the point of inspection and classification in acommercial abattoir located in Scotland. Each week for 6weeks, 4 steers and 4 heifers were selected from Charolaisand Limousin breeds and 4 bulls and 4 steers wereselected from dairy breeds based on passport breed codes,age, and genders typically slaughtered by the processor.From the start of kill, carcasses matching the criteria wereselected for inclusion until four were reached in eachbreed and gender category giving a maximum of 24carcasses per week. After six weeks the data set comprised24 Charolais heifers (CH), 23 Charolais steers (CS), 25Limousin heifers (LH), 24 Limousin steers (LS), 24 dairysteers (DS) and 21 dairy bulls (DB). Breed codes areentered on the passports by the producer and are derivedfrom the sire breed (Todd et al., 2011). All Charolais andLimousin carcasses were crossbreds, whereas of the 45dairy animals, 34 were Holstein-Friesians, 4 were HolsteinFriesian crosses, four were Holsteins and three wereBritish Friesians according to the breed code descriptions(Todd et al., 2011). Ethical approval was not required forthis experiment as no measurements were recorded onlive animals.

2.2. Abattoir protocol

Cattle were stunned using a captive bolt, exsanguinated,and subjected to electrical stimulation (90 V for 30 s at 10 minpost-mortem) during hide removal. Carcasses were split inhalf down the midline and dressed to UK specification asdescribed in the Meat and Livestock Commercial ServicesLimited beef authentication manual (www.mlcsl.co.uk).

Visual carcass classification for conformation and fat-ness was performed on each carcass by a trained Meat andLivestock Commercial (MLC) human assessor using themore-restricted version of EUROP classification scale usedin the UK (MLCuk) that uses eight of the 15 possiblecategories for conformation (MLCCuk) and seven out of apossible15 categories for fatness (MLCFuk) (EuropeanCommunity, 1981a, 1981b). Both MLCCuk and MLCFukvalues were expressed on the full 15 point scale equiva-lents for analysis (the equivalent values from UK scale tothe 15 point scale are provided in Table 1).

A VIA system (VBS 2000, EþV GmbH, Germany) alsoestimated EUROP conformation and fatness on the UKscale (eight categories for conformation and seven forfatness) (referred to as VIACuk and VIAFuk, respectively,or VIAuk collectively). In addition, the VIA also predictedconformation and fatness on the full 15 point scales(VIAC15 and VIAF15, respectively or VIA15 collectively). Adirect prediction of whole-carcass SMY% was also obtainedfrom the VIA system (VIA-SMY%). VIA data was availableon 137 out of 141 carcasses as four carcasses werepresented to the VIA incorrectly. The VIA system wasoperating on-line and independently from the humanclassifier. Hot carcass weight (HCW) was recorded at the

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178170

Author's personal copy

same point as carcass classification and was used as oneinput parameter to the VIA system.

2.3. Carcass cutting protocol

All 141 carcasses were quartered between the 10th and11th ribs at 48 h post-mortem into hind (pistola) andforequarters. Only the left half of the carcass was usedfor the experiment, a schematic of the cutting protocol isshown in Fig. 1. The complete sirloin, comprising the 11thrib to the last lumbar vertebrae (inclusive of the fillet) wasremoved from the hind quarter by cutting along the cranialedge of the ilium and extending a cut through the flankapproximately 30 mm beyond the ventral part of theM. longissimus lumborum in a cranial direction to thequartering point (Kempster et al., 1980). The untrimmedfillet (FIL) (containing M. Psoas major and M. Psoas minor)was removed from the complete sirloin and weighed, thebone (BON) and excess fat (XSF) (fat trimmed to a max-imum depth of 9 mm at the 3/4 point across the long-issimus muscle at the 10th rib) of the sirloin were alsoweighed. The resulting trimmed, boneless saleable meat ofthe sirloin (SS), which included parts of M. longissimuslumborum, M. multifidus dorsi and M. longissimus costarum,was weighed, and the yield of saleable sirloin [SS/HSW]and fillet [FIL/HSW] were calculated as a percentage of thehot side weight (0.5�HCW), while sirloin muscle-to-boneratio (M:B) was calculated as SS divided by BON.

2.4. Statistical analysis

Three parts of statistical analysis were undertaken todetermine differences between genotypes, to compare VIA

and visual classification, and to establish whether additionalcovariates could account for more variation in the yieldprediction of sirloin and fillet cuts. All statistical analysiswas undertaken using SAS (SAS Inst. Inc., Cary, NC). For partone, traits analysed included hot carcass weight, age atslaughter, carcass traits assessed by visual and VIA carcassclassification, weights of the sirloin and its dissected compo-nents, and yields of the latter. A general linear model (PROCGLM) was used to estimate least-squares means for thegenotype–gender effects (CH, CS, LH, LS, DS and DB) afteradjusting for batch (determined by slaughter day as a secondfixed effect, n¼6). For each model, HCW was fitted as acovariate for all traits except for animal age. Comparison ofleast-squares means among genotype–gender effects wereperformed using t-tests (Bonferroni adjusted to account formultiple comparisons). Moreover, three non-orthogonal con-trasts were employed to make comparisons between geno-types and genders using “estimate” statements in SAS. Forpart two pair-wise residual correlations among traits wereestimated using the MANOVA option in PROC GLM, in a firstanalysis adjusting for batch effects only, and in a secondanalysis by additionally adjusting for genotype–gendereffects. Residual correlations measure the correlationsbetween pairs of variables after removing the slaughter dayand genotype–gender effects.

In the third part of the analysis, eight general linearmodels (all adjusted for slaughter batch effect) are com-pared to evaluate how genotype–gender, visual classifica-tion or VIA information relate to the sirloin and fillet cutweights, saleable meat yield and sirloin M:B ratios usingthe coefficient of determination (R2, higher indicates agreater percentage of total variation is explained by themodel) and residual standard deviation (RSD, lower isbetter). Model one included HCW as the sole explanatoryvariable, model 2: HCWþgenotype–gender, model 3:HCWþMLCCukþMLCFuk, model 4: HCWþVIACukþVIAFuk,model 5: HCWþgenotype–genderþMLCCukþMLCFuk,model 6: HCWþgenotype–genderþVIAC15þVIAF15, model7: HCWþgenotype–genderþVIACukþVIAFuk and model

Table 1The distribution (percentage) of carcasses (n¼141) used in the currentexperiment based on visually-assigned EUROP conformation (MLCCuk)and fat (MLCFuk) classes. The distribution (%) of all prime beef animalsslaughtered in Great Britain in 2009 is included for comparison.

Fatness (MLCFuk)ab GB 2009(%)c

2(5)

3(8)

4L(10)

4H(12)

5L(13)

5H(15)

Total

Conformation (MLCCuk)ab

Uþ (12) 0 0 0.7 1.4 0.4 0 2.5 2.3�U (10) 0 1.4 6.4 5.0 0 0 12.8 13.0R (8) 0.7 4.3 20.6 19.2 0.7 0 45.5 44.2Oþ (6) 0 4.3 8.5 1.4 0.7 0 14.9 26.6�O (4) 1.4 11.4 7.8 3.6 0 0 22.8 11.5Pþ (3) 0.4 0.9 0 0 0 0 1.3 1.8�P (1) 0.7 0 0 0 0 0 0.7 0.3Total 1.1 21.3 44.0 30.5 1.42 0 100GB 2009(%)c

10.8 30.8 44.0 12.1 0.9 0.1

a Conformation and fatness classification scores determined by atrained assessor on the UK scale, note there were no carcasses that had anE on the conformation or a 1 on the fatness scale, so these categories havebeen omitted.

b Numbers in brackets are the corresponding categories on the15-point EUROP scale (Fisher, 2007).

c GB 2009 refers to all prime cattle (all finished steers, young bullsand heifers that have not been bred) slaughtered in England, Scotlandand Wales between 1st January and 31st December 2009 (courtesy ofKim Matthews, English Beef and Lamb Executive).

Left side of carcass

Fillet (FIL)

Bone (BON)

Fat (XSF)

CompleteSirloin

SaleableSirloin (SS)trimmed to9 mm fat

Removedcomponents

Fig. 1. Schematic of carcass cutting procedure used in the experiment.

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178 171

Author's personal copy

8: HCWþgenotype–genderþVIA saleable meat yieldprediction.

3. Results and discussion

The distribution of carcasses for EUROP conformationand fat class (UK scale) in the present analysis wascompared to that of the distribution of prime cattle (allfinished steers, young bulls and heifers that have not beenbred) slaughtered in Great Britain (GB) in 2009 (Table 1).The majority of the carcasses were classified as R (45.5%),�O (22.8%) and Oþ (14.9%) for conformation, and 4L(44.0%), 4H (30.5%) and 3 (21.3%) for fatness. The sampledistribution was similar to that in the GB prime cattlepopulation slaughtered in 2009, but there was a higherproportion of 4H carcasses (30.5% vs. 12.1% in the GBpopulation), fewer in the Oþ and more in the �Oconformation classes in the sample set.

3.1. Genotype–gender effects

Least-squares means of genotype–gender effects forcarcass classification (all on a scale of 1–15) are shown inTable 2. Significant batch effects were present for VIAFuk,VIAF15 and VIA-SMY% (data not shown).

There was no significant difference between steers forfatness, but DS were significantly fatter than DB accordingto both visual and VIA classifications (Po0.001) and thedifference was more pronounced for the VIAF15 system(3.2970.44, Po0.001).

In contrast, there was no significant difference in con-formation class between DS and DB according to MLCCuk,whereas VIACuk and VIAC15 determined that DB had higherconformation scores than DS with differences of 0.9270.39(P¼0.02) and 0.9770.36 (Po0.008) respectively. As mightbe expected, beef steers (LS and CS) had significantly higherconformation class scores than dairy steers (DS), with thisbeing more pronounced with visual classification, thanVIACuk, or VIAC15 (Po0.001 in all cases).

Overall, heifers were significantly fatter than steers (3rdcontrast in Table 2) with the largest difference being foundwhen assessed by MLCFuk (1.7570.36, Po0.001). Thisdifference was considerably less when assessed by VIAFuk(1.5970.40, Pr0.001) and VIAF15 (1.4970.39, Po0.001).Heifers were found to have higher conformation scoresthan steers, but only when assessed by VIACuk (0.8870.35,P¼0.01), this difference was no longer statistically signifi-cant based on VIAC15 (0.5670.32, P¼0.08).

Carcass conformation and fatness as determinedvisually and by VIA were considerably different. MLCFukscores were consistently lower than VIAF15 and VIAFuk inboth beef genotypes, except for fat class where VIAFukpredicted the fat class of DB to be approximately a sub-class (1/15) lower than values determined by MLCFuk. Inthe comparison between DS and DB, the VIAF15 estimatedfatness difference between the genders were higher in DSby around 3 sub-classes whereas MLCFuk estimated thedifference to be around 1.5 sub-classes. In a previousreport where dairy bulls and steers were assessed by VIAin the Republic of Ireland on the 15-point EUROP scale,the difference in mean fat-class between genders was 0.7

sub-classes in favour of the steers (Conroy et al., 2010b).The small sample size in the present experiment, maypartly explain the greater difference in VIA-predicted fatclass observed between DS and DB, but of the prime cattleslaughtered in the UK in 2009 (1.95 million head), the fatclass of young bulls was 2–3 sub-classes (visually assessedon the UK scale but converted to the 15 point scale) lowerthan steers and heifers (personal communication, KimMatthews, English Beef and Lamb Executive, EBLEX).

The difference between genotype–gender effects isimportant (particularly in terms of SMY%), but the abilityof visual and VIA carcass evaluation systems to detectthese differences is a more appropriate basis of compar-ison. Further investigation into the significance of thedifferences between VIA and visual classification need toinclude carcass classification as a whole, i.e. fat class andconformation class since both are used together to describethe merits of a carcass.

3.2. Carcass classification

Further enquiry into the ability of the EUROP carcassclassification to determine SMY% is only possible wheneither fat class or conformation class (but not both) aresignificantly different in a given comparison at a commonlevel of fat trimming (Kempster, 1986). The present resultssatisfy this condition in the first contrast in Table 2: “Beefsteers vs. Dairy steers” and between genders, within theCharolais genotype only (data not shown). In the firstcontrast, the beef steers had significantly higher confor-mation scores than the dairy steers according to MLCCuk,VIACuk and VIAC15, but there were no significant differ-ences in fat class between the genotypes. In Charolais, CHhad higher fat class scores than CS according to MLCFuk,VIAFuk and VIAF15 but there was no significant differencein carcass conformation. SMY% traits SS/HSW and FIL/HSWwere both greater in the beef steers than DS (P¼0.08 andP¼0.001 respectively) as was the conformation score(Po0.001). Within the Charolais genotypes, CH had sig-nificantly higher fatness scores than steers (CS) andshowed significantly greater yields of SS/HSW (P¼0.005).No significant differences in FIL/HSW (P¼0.95) weredetected between CH and CS.

The inconsistent relationship between EUROP classifi-cation and SMY% in the current results may have severalcauses; including over fat carcasses (Purchas and Wilkin,1995), or may be due to an inconsistent relationshipbetween muscularity and M:B, particularly in the case ofheifers, where the M:B may be higher than that of bullsand steers (Purchas et al., 2002). The inconsistenciesbetween genders within genotypes are not surprising,given that a meta-analysis on 903 carcasses in 11 experi-ments failed to find a significant relationship betweencarcass conformation class and muscle proportion (Keaneet al., 2000). On the other hand, the same authors reportedfat class was negatively associated with both muscle andbone proportions in seven of the 11 experiments (Keaneet al., 2000). The inconsistency between carcass shape andSMY% has also been found in lambs where prediction oflean meat yield based on carcass shape, will be under-estimated in females and overestimated in males (Johnson

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178172

Author's personal copy

et al., 2005). Upon further investigation in the currentanalysis, CH did have a higher M:B ratio than CS but thedifference was not statistically significant (P¼0.24). Moreimportantly, the differences in M:B may be biased ifcarcass shape is used as a predictor of M:B (Purchaset al., 2002) and ultimately SMY%. Further investigationinto the relationship between carcass conformation andfatness and SMY% is possible by comparing the correla-tions between the traits. Table 3 shows residual correla-tions (adjusted for batch only, as well as adjusted for batch

and genotype–gender effects) between HCW, carcass clas-sification and SS/HSW, FIL/HSW and M:B.

After adjusting for batch effects, MLCFuk was positivelycorrelated to HCW (r¼0.2), and carcass conformation waspositively correlated to HCW in all cases (r¼0.47–0.55)(Table 3). Both fatness and conformation were moderatelypositively correlated to SS/HSW. Conformation classes hadthe closest positive correlations to SS/HSW in the firstinstance (r¼0.49–0.52) with little difference between VIA(UK and 15-pt scale) and MLCCuk, but after adjusting for

Table 2Least-squares means for genotype-gender groups adjusted for hot carcass weight and batch effects as well as contrasts between genotype–gender groups.

Trait(abbreviation)

Least-squares means of genotype–gender groups11 Contrast Effects12 R2 [%](RSD)13

CH CS LH LS DS DB Beef steersvs. dairysteers

Dairysteers vs.dairy bulls

Beef steersvs. beefheifers

HCW Group

n n¼24 n¼23 n¼25 n¼24 n¼24 n¼21Hot carcassweight(HCW) (kg)

318.8a 400.6b 301.2a 351.1b 322.5a 303.7a o0.001 0.06 o0.001 – o0.001 52.2 (34.55)

Age atslaughter(Age) (d)

753.8a 704.0a 741.2a 709.6a 756.3a 477.5b 0.03 o0.001 0.03 – o0.001 56.3 (88.70)

MLCFuk1 11.50a 9.80bcd 11.19ab 9.38cd 9.96c 8.43d 0.35 o0.001 o0.001 0.004 o0.001 40.8 (1.37)MLCCuk

2 8.13b 8.29b 8.07b 8.15b 4.73a 4.47a o0.001 0.41 0.66 o0.001 o0.001 80.1 (1.01)

Video image analysisVIAFuk3 11.94a 10.54ab 11.63a 9.85b 9.84b 7.06c 0.42 o0.001 o0.001 0.24 o0.001 59.6 (1.51)VIAF154 11.70a 10.54ab 12.01a 10.00b 10.44b 7.62c 0.64 o0.001 o0.001 0.23 o0.001 62.5 (1.46)VIACuk5 7.83b 7.14b 7.83 b 6.76 b 4.35a 5.26a o0.001 0.02 0.01 o0.001 o0.001 63.8 (1.30)VIAC156 7.65b 7.26b 7.64b 6.92b 4.65a 5.62a o0.001 0.008 0.08 o0.001 o0.001 65.8 (1.18)VIA-SMY (%)7 77.39c 76.78ac 77.26c 77.09c 75.23b 75.84ab o0.001 0.03 0.13 o0.001 o0.001 63.1 (0.94)

WeightsSaleablesirloin (SS)(kg)

7.73b 7.21ab 7.65b 7.26ab 6.79a 6.72a 0.01 0.69 0.005 o0.001 o0.001 81.8 (0.61)

Fillet (FIL)(kg)

3.56c 3.56bc 3.58c 3.51bc 3.23ab 3.15a 0.001 0.12 0.70 o0.001 o0.001 79.3 (0.27)

Excess fat(XSF) (kg)

1.21b 1.10ab 1.21b 1.11b 1.18b 0.77a 0.38 o0.001 0.19 o0.001 o0.001 47.6 (0.30)

Bone (BON)(kg)

3.53bc 3.56bc 3.41bc 3.47bc 3.61bc 3.85a 0.38 0.05 0.70 o0.001 0.008 53.7 (0.40)

YieldsSS/HSW (%)8 4.67b 4.32ab 4.62b 4.36ab 4.06a 4.02a 0.08 0.67 0.002 0.23 o0.001 53.7 (0.36)FIL/HSW (%)9 2.14d 2.14cd 2.16bd 2.11cd 1.97ac 1.88ab 0.001 0.07 0.51 0.09 o0.001 32.4 (0.16)M:B10 2.22b 2.04ab 2.26b 2.10ab 1.98ab 1.73a 0.43 0.05 0.14 0.02 o0.001 23.4 (0.42)

Note: For VIA traits the following numbers apply: CS (n¼22), LH (n¼24) and LS (n¼22). Least-squares means within a row sharing a common superscriptletter are not significantly different (PZ0.05).

1 MLCFuk¼meat and livestock commission fatness (operating on the UK scale, but expressed on 15-pt scale).2 MLCCuk¼meat and livestock commission conformation (operating on the UK scale, but expressed on 15-pt scale).3 VIAFuk¼video image analysis fatness (operating on the UK scale, but expressed on 15-pt scale).4 VIAF15¼video image analysis fatness operating on the 15-pt scale.5 VIACuk¼video image analysis conformation (operating on the UK scale, but expressed on 15-pt scale).6 VIAC15¼video image analysis conformation operating on the15-pt scale.7 VIA-SMY(%)¼video image analysis prediction of total carcass saleable meat yield.8 SS/HSW(%)¼yield of saleable sirloin meat (boneless with fat trimmed to 9 mm).9 FIL/HSW(%)¼yield of fillet.10 M:B¼muscle-to-bone ratio of the sirloin (excluding the fillet).11 CH¼Charolais heifers, CS¼Charolais steers, DS¼dairy steers, DB¼dairy bulls, LH¼Limousin heifers and LS¼Limousin steers. Pairwise comparisons

between the means are provided for information purposes, contrasts are used to draw specific comparisons between the genotype–gender groups.12 Effects (P values) HCW¼hot carcass weight (included as a covariate), group¼genotype–sex group (included as a fixed effect, n¼6), batch was

included as a fixed effect (n¼6) for all traits (data not shown).13 R2¼coefficient of determination (RSD¼residual standard deviation).

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178 173

Author's personal copy

the genotype–gender effects, the correlation was greatlyreduced (r¼0.28–0.31). This is especially evident in theFIL/HSW where conformation and fatness were not corre-lated to FIL/HSW after the genotype–gender effects wereremoved. M:B was positively correlated to HCW, but asexpected M:B and fatness were not correlated after adjust-ment for genotype–gender effects which is probably dueto the lower M:B in the DB genotype–gender group(Table 2). Conformation correlated to M:B but largeimprovements were seen when no adjustments weremade for genotype–gender effects (Table 3). This is mostlikely because M:B varies by gender and genotype.

In a previous experiment (Kempster and Harrington,1980), the correlation between carcass conformation andthe percentage of high-priced cuts (as a percentage of totalsaleable meat weight rather than total side weight) wassimilarly low, although only six classification categorieswere used. The highest correlations between sirloin SMY%and composition in the current results were obtainedbetween VIA operating on the 15 point scales, whichsuggests the scale is appropriate for the UK prime cattlepopulation. The correlations between yield and classifica-tion should be interpreted with caution because a positiverelationship between carcass fatness and SS/HSW wasobserved in CH, and MLCFuk was positively correlated toSS/HSW across all genotype–gender groups. Furthermore,it is also possible that the total carcass SMY% may not havethe same relationship to conformation as was presentedfor the sirloin region in the current results.

After adjusting for genotype–gender effects and batch,in the current analysis, there were no significant correla-tions between conformation and fat class, which corrobo-rates previous findings (Conroy et al., 2009, 2010a, 2010b;Drennan et al., 2007, 2008) who found fatness and con-formation were not significantly correlated in the absence

of genotype and gender effects. It has been previouslysuggested that breed generally provides a more preciseprediction of carcass composition than carcass conforma-tion in steers (Kempster and Harrington, 1980) and thecurrent results support this hypothesis.

Carcass classification needs to be an accurate predictorof SMY% if it is to be an effective mode of carcassevaluation since saleable meat is recognised as the majorvalue component of a carcass. Maximisation of SMY% isperhaps more of a target for meat processors than produ-cers at the present time because higher SMY% equates toless waste and greater plant efficiency. SMY% of a carcasswill increase when M:B increases at any given percentageof fat (FAT%) or if FAT% decreases at a constant M:B(Purchas, 2003). In order to maximise efficiency, theproducer must be able to finish cattle to the target weight,conformation and fat class for the lowest possible cost.Assuming conformation class (carcass shape) relates to M:B and fat class is a proxy estimation of total carcass FAT%,carcass classification should be indicative of SMY%, but thecorrelation between the two end points is variable andoften not statistically significant (Keane et al., 2000). Thepoor association between EUROP carcass classification andSMY% becomes problematic when classification is used asa basis of carcass evaluation and payment. When pricesignals are based on carcass classification and weightrather than SMY%, price incentives to producers may notallow the value chain to optimise production and improvethe efficiency of the wider industry.

Considering the relationship between the full 15-pointscale and the restricted UK scale; the un-adjusted (full)correlations between VIACuk and VIAC15 was r¼0.87(Po0.001) and VIAFuk and VIAF15 was r¼0.92 (Po0.001)(results not shown in tables). A full correlation less thanr¼1.00 between the VIA classification parameters is probably

Table 3Residual correlation and P-values (where significant) between carcass classification and hot carcass weight (HCW), yield traits (SS/HSW, FIL/HSW), and loinregion muscle to bone ratio (M:B) for 137 carcasses where VIA were present. Each analysis was successively adjusted for batch (B) and both batch andgenotype–gender (BþG).

Trait HCWa SS/HSW (%)h FIL/HSW (%)i M:Bj

Effects B BþG B BþG B BþG B BþG

HCWa – – 0.09 0.11 �0.03 �0.16 0.28 (0.001) 0.23 (0.01)

MLCFukb 0.20 (0.02) 0.28 (0.002) 0.41 (o0.001) 0.18 (0.05) 0.16 �0.16 0.28 (0.001) 0.10VIAFukc 0.16 0.10 0.38 (o0.001) 0.10 0.27 (0.002) �0.16 0.29 (o0.001) 0.01VIAF15d 0.16 0.11 0.42 (o0.001) 0.02 0.27 (0.002) �0.17 0.30 (o0.001) 0.02

MLCCuke 0.54 (o0.001) 0.47 (o0.001) 0.52 (o0.001) 0.30 (o0.001) 0.44 (o0.001) 0.12 0.44 (o0.001) 0.28 (0.002)VIACukf 0.47 (o0.001) 0.42 (o0.001) 0.51 (o0.001) 0.31 (o0.001) 0.32 (o0.001) 0.00 0.40 (o0.001) 0.25 (0.005)VIAC15g 0.55 (o0.001) 0.49 (o0.001) 0.49 (o0.001) 0.28 (0.001) 0.32 (o0.001) 0.04 0.45 (o0.001) 0.34 (o0.001)

a HCW¼hot carcass weight.b MLCFuk¼meat and livestock commission fatness (operating on the UK scale, but expressed on 15-pt scale).c VIAFuk¼video image analysis fatness (operating on the UK scale, but expressed on 15-pt scale).d VIAF15¼video image analysis fatness operating on the 15-pt scale.e MLCCuk¼meat and livestock commission conformation (operating on the UK scale, but expressed on 15-pt scale).f VIACuk¼video image analysis conformation (operating on the UK scale, but expressed on 15-pt scale).g VIAC15¼video image analysis conformation (15-pt scale).h SS/HSW (%)¼yield of saleable sirloin meat (boneless with fat trimmed to 9 mm).i FIL/HSW (%)¼yield of fillet.j M:B¼muscle-to-bone ratio of the sirloin (excluding the fillet).

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178174

Author's personal copy

a result of converting the full 15-point scale into thecategories employed in the UK.

The residual correlation coefficients (r) between theVIA classification and the visual classification are firstlyadjusted for batch only and secondly for batch, genotype–gender effects and HCW (Table 4). Results indicate that thesame characteristic (fatness or conformation) assessed bythe two systems were highly correlated.

Fatness and conformation scores were positively corre-lated with each other because carcasses with higher con-formation tend to be fatter. This is mostly due to gender,genotype and weight effects. After adjusting for genotype–gender, batch and HCW effects, fatness was not correlatedwith conformation using either the visual or VIA system.The residual correlation between VIAFuk and MLCFuk(r¼0.69, Po0.001) was stronger than the residual correla-tion between VIACuk and MLCCuk (r¼0.49, Po0.001). Thissuggests that the classifier and VIA are assessing differentaspects of carcass shape.

3.3. Prediction of sirloin and fillet weights, yields, and sirloinM:B

The accuracies (R2 and RSD) and significance of termsused in a range of models for predicting sirloin componentweights and SMY% of beef carcasses are shown in Table 5.As expected, HCW explained the majority of the variationin weight traits (SS and FIL). Batch effects were significantfor SS, XSF, BON, SS/HSW and FIL/HSW as date of slaughterhad a large influence on the level of fat trimming (data notshown).

Addition of conformation and fatness predictors offeredsmall improvements in prediction accuracy of SS/HSW andFIL/HSW SMY% traits (models 3 and 4 vs. model 2 inTable 5). VIA15 (model 6) was able to predict SS/HSW withthe highest accuracy overall (R2¼58.5%, RSD¼0.35). Thedifferences between VIA and visual carcass classificationwere small in terms of accuracies (o2%). After accounting

for genotype–gender, batch and HCW effects, carcassclassification offered little in terms of additional accuracy(r5%) for the prediction of FIL/HSW. Similarly, the directprediction of SMY% from VIA (VIA-SMY%) was equally poorat predicting FIL/HSW.

Compared to visual classification, VIA is better detect-ing differences in carcass regions that are difficult tomeasure (e.g., the loin). These results are in agreementwith R2¼57% for VIA and R2¼51% for visual classificationfor predicting yield of cube roll, strip-loin and fillet (veryhigh value cuts) reported by Drennan et al. (2007). Thepresent results are also in agreement with the R2¼56%(RSD¼0.30%) for sirloin reported in a recent trial under-taken by EBLEX (personal communication Kim Matthews).Prediction accuracies for M:B obtained from HCW andMLCuk completely explained the genotype–gender effectson M:B (Table 5). The same predictors were also the mostaccurate at determining the joint composition (percen-tages of saleable meat, excess fat and bone) of the sirloinregion (data not shown).

Considering the high value of the sirloin joint relativeto the rest of the carcass, prediction of loin SMY% with ahigher degree of accuracy is desirable for the meatindustry. Carcass classification encompasses the wholecarcass and is not solely focused on the sirloin region.Several previous experiments that have sought to predictboth the whole side and the very-high value cuts havefound that classification was considerably more accurate atpredicting the half carcass SMY% than very high value cutyields respectively: R2¼68% vs. 51% for the visual and 75%vs. 57% for VIA classification (Drennan et al., 2007),R2¼70% vs. 29% for visual classification (Drennan et al.,2008). Furthermore, the median overall accuracy (coeffi-cient of determination) of the VBS 2000 across a numberof published experiments was 76% (Craigie et al., 2012).In the current experiment, carcass classification was notable to measure the SMY% of the sirloin region as accu-rately as the SMY% of the whole or half carcass reported in

Table 4Residual correlations and P values (where significant) adjusted for batch [B] and residual correlations after adjusting for both B and group [G] between VIA-predicted and visually assessed carcass classification methods on 137 carcasses where VIA records were present.

Trait MLCFuka VIAFukb VIAF15c MLCCukd VIACuke

Effects B BþG B BþG B BþG B BþG B BþG

VIAFukb 0.78 (o0.001) 0.65(o0.001)

1 1 – – – – – –

VIAF15c 0.79 (o0.001) 0.69(o0.001)

0.96(o0.001)

0.92(o0.001)

1 1 – – – –

MLCCukd 0.39 (o0.001) 0.16 0.47

(o0.001)0.01 0.49

(o0.001)0.07 1 – – –

VIACuke

0.37 (o0.001) 0.15 0.38(o0.001)

�0.01 0.38(o0.001)

0.00 0.79(o0.001)

0.49(o0.001)

1 –

VIAC15f 0.34 (o0.001) 0.15 0.34(o0.001)

�0.02 0.33(o0.001)

�0.02 0.83(o0.001)

0.61(o0.001)

0.95(o0.001)

0.90(o0.001)

a MLCFuk¼meat and livestock commission fatness (operating on the UK scale, but expressed on 15-pt scale).b VIAFuk¼video image analysis fatness (operating on the UK scale, but expressed on 15-pt scale).c VIAF15¼video image analysis fatness operating on the 15-pt scale.d MLCCuk¼meat and livestock commission conformation (operating on the UK scale, but expressed on 15-pt scale).e IACuk¼video image analysis conformation (operating on the UK scale, but expressed on 15-pt scale).f VIAC15¼video image analysis conformation (15-pt scale).

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178 175

Author's personal copy

other experiments with R2 values ranging between 60%and 74% (Conroy et al., 2009, 2010a, 2010b; Drennan et al.,2007, 2008).

Total carcass value and the price consumers pay for apackage of meat are highly weight-dependent, so there isa case for predicting cut weight. Variation in cut weights islargely explained by HCW, but evaluation based on HCWalone would not account for variations in cattle types

(where large differences exist in the proportions of meat,fat and bone). After adjustment for batch effects in thecurrent results, HCW explained 74%, 48% and 33% of thevariation in the weights of SS, BON, and XSF in thecomplete sirloin (CSL), respectively (Table 5). It wasrecently reported that cold carcass weight could explain74% (RSD¼2.28 kg) of the variation in very high value cut(rib-roast, strip-loin and fillet) weights (Pabiou et al., 2011)

Table 5Models used to compare manual classification and VIA parameters for predicting loin weight, SMY% and muscle-to-bone ratio on 137 carcasses afteradjusting for batch effects. The significance of various covariate effects and the main genotype–gender effects are reported.

Model/trait Weight Yield M:Bo

SS (kg)j BON (kg)k XSF (kg)l SS/HSW (%)m FIL/HSW (%)n

Standard deviation 1.87 0.56 0.4 0.51 0.19 0.46

1. R2 (%) (RSD) 74.1 (0.71) 47.9 (0.41) 33.0 (0.33) 30.0 (0.43) 5.1 (0.19) 9.6 (0.45)HCWa o0.001 o0.001 o0.001 0.3 0.6 0.002

2. R2 (%) (RSD) 81.8 (0.61) 53.7 (0.40) 46.8 (0.30) 53.4 (0.36) 32.4 (0.16) 23.4 (0.42)HCWa o0.001 o0.001 o0.001 0.23 0.09 0.02Genotype–genderb o0.001 0.008 o0.001 o0.001 o0.001 o0.001

3. R2 (%) (RSD) 82.5 (0.59) 55.7 (0.38) 51.6 (0.28) 55.0 (0.35) 30.1 (0.16) 24.9 (0.41)HCWa o0.001 o0.001 0.001 0.004 o0.001 0.66MLCCukc o0.001 o0.001 0.41 o0.001 o0.001 o0.001MLCFukd 0.02 0.17 o0.001 0.005 0.6 0.14

4. R2 (%) (RSD) 81.1 (0.61) 51.9 (0.38) 55.4 (0.27) 53.8 (0.36) 22.0 (0.17) 21.7 (0.41)HCWa o0.001 o0.001 o0.001 0.02 0.01 0.18VIACuke o0.001 0.009 0.8 o0.001 o0.001 0.003VIAFukf 0.02 0.09 o0.001 0.008 0.06 0.06

5. R2 (%) (RSD) 83.4 (0.58) 57.8 (0.38) 55.6 (0.28) 57.7 (0.35) 36.4 (0.16) 27.7 (0.41)HCWa o0.001 o0.001 0.04 0.68 0.05 0.3Genotype–genderb 0.28 0.28 0.05 0.15 0.03 0.43MLCCukc 0.002 0.002 0.69 0.002 0.03 0.009MLCFukd 0.26 0.27 o0.001 0.16 0.05 0.7

6. R2 (%) (RSD) 83.1 (0.60) 55.8 (0.38) 58.4 (0.27) 58.5 (0.35) 36.5 (0.16) 28.4 (0.41)HCWa o0.001 o0.001 0.01 0.45 0.04 0.43Genotype–genderb 0.33 0.3 0.5 0.23 o0.001 0.33VIAC15g o0.001 o0.001 0.37 o0.001 0.17 0.002VIAF15h 0.32 0.52 o0.001 0.18 0.1 0.88

7. R2 (%) (RSD) 82.3 (0.60) 53.7 (0.39) 57.1 (0.27) 57.1 (0.35) 35.5 (0.16) 25.1 (0.42)HCWa o0.001 o0.001 0.01 0.85 0.08 0.13Genotype–genderb 0.11 0.47 0.3 0.09 o0.001 0.35VIACuke 0.01 0.02 0.33 0.003 0.5 0.05VIAFukf 0.93 0.48 o0.001 0.81 0.12 0.98

8. R2 (%) (RSD) 82.7 (0.59) 56.5 (0.37) 48.1 (0.30) 57.7 (0.35) 35.5 (0.16) 28.4 (0.41)HCWa o0.001 o0.001 0.02 0.6 0.02 0.38Genotype–genderb 0.004 0.17 o0.001 o0.001 o0.001 0.09VIA-SMY%i 0.004 o0.001 0.13 o0.001 0.08 0.002

Note: All models were corrected for batch (fixed effect, n¼6, data not shown).lXSF¼weight of excess fat trimmed from the saleable sirloin.

a HCW¼hot carcass weight (covariate).b Genotype–gender effect (fixed, n¼6).c MLCCuk¼meat and livestock commission conformation (operating on the UK scale, but expressed on 15-pt scale) (covariate effect).d MLCFuk¼meat and livestock commission fatness (operating on the UK scale, but expressed on 15-pt scale) (covariate effect).e VIACuk¼video image analysis conformation (operating on the UK scale, but expressed on 15-pt scale) (covariate effect).f VIAFuk¼video image analysis fatness (operating on the UK scale, but expressed on 15-pt scale) (covariate effect).g VIAC15¼video image analysis conformation (15-pt scale) (covariate effect).h VIAF15¼video image analysis fatness operating on the 15-pt scale (covariate effect).i VIA-SMY%¼video image analysis prediction of total carcass saleable meat yield (covariate effect).j SS¼weight of saleable sirloin.k BON¼weight of bone removed from the loin region.m SS/HSW(%)¼yield of saleable sirloin meat (boneless with fat trimmed to 9 mm).n FIL/HSW(%)¼yield of fillet.o M:B¼muscle-to-bone ratio of the sirloin (excluding the fillet).

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178176

Author's personal copy

and the current results corroborate this. But, HCW doesnot account for all or much of the variation in sirloin XSFor BON weights. Addition of genotype–gender effects orcarcass classification information offered improvementson weight prediction accuracy (82% and 83% respectively).Conformation was not a significant predictor for any of theXSF weight models, but EUROP fatness score did accountfor additional variation over and above genotype–gendereffects and HCW. Carcass fatness classification did accountfor an additional 10% of the variation in XSF and BON withVIA15 being slightly (2–3%) more accurate than the visualclassification (models 6 and 7 vs. model 5). VIA-SMY (%)showed similar accuracies for SS R2¼83% (RSD¼0.59 kg)and XSF, but accuracies for BON were lower than withcarcass classification values.

Carcass SMY% is affected by variations in carcass fatpercentage at a constant M:B ratio and variations in M:Bratio at a constant fat percentage, therefore it washypothesised that weights of excess fat and bone removedduring sirloin fabrication may account for additionalvariation in the SS/HSW and FIL/HSW. The weights of boneand excess fat removed during the sirloin fabrication weretested as additional covariates along with MLCuk, VIAuk

and VIA15 conformation and fat classes. It was found thatinclusion of XSF and BON weight as covariates in thecurrent analysis offered very small improvements in pre-diction accuracy for SS/HSW and FIL/HSW (R2 increase of4.5–6.8% for BON ando2.5% for XSF) (data not shown).

An early model of whole-side VIA, the BCC-1 (Sørensenet al., 1988) had a probe for determining fat depth, butnone of the current whole-side VIA-systems have a backfat depth probe, presumably because a measure of fatdepth did not increase prediction accuracy. Furthermore,the distribution of fat throughout the carcass varies bygenotype, gender, maturity and diet (McPhee et al., 2009)making the use of a back fat probe futile unless all (or atleast some) of the other factors are taken into account.

3.4. Future challenges for beef VIA

A high prediction accuracy for SMY% at a constant M:Bratio is dependent on accurate measures of carcass fatpercentage (Purchas et al., 2002). Because the VBS 2000VIA system captures images from the exterior of the halfcarcass, only subcutaneous fat is visible. Therefore, allpredictions of carcass composition are based on theassumption that fat distribution throughout the carcass isconsistent, but fat distribution has been shown to varybetween beef and dairy genotypes (Fisher and Bayntun,1984; McPhee et al., 2009). Further research is needed toaddress the differences in fat distribution and how thisaffects the accuracy with which VIA can predict SMY%.

Most analyses using VIA information from the VBS2000 system have used predicted EUROP classificationvariables. Further research is needed to establish whetherusing raw VIA data (such as primal yield predictions andcarcass dimensions) directly, rather than VIA-predicted theEUROP offer improvements in accuracy. Two recentreports have investigated the relationships between var-ious carcass dimensions and yield traits and have shownthat direct VIA outputs can predict SMY% with a higher

degree of accuracy than EUROP classification scores (Oliveret al., 2010; Pabiou et al., 2011), but information in thisarea is still lacking.

4. Conclusions

Although the present sample set was small, both VIAand visual classification systems predicted sirloin regionweights, yields and M:B with similar accuracies for beefcarcasses of different genotype and gender, but on balance,the VIA operating on the 15-point scale had slightly higheraccuracies. According to previous findings, it is likely thatthe accuracies obtained in the current experiment wouldhave been higher for both visual and VIA classificationsystems if the SMY% of the whole or half carcass wasavailable rather than just the sirloin region. Irrespective ofthis, both VIA and visual systems were relatively poor atpredicting the yield of fillet as there was no statisticallysignificant correlation between fillet yield and carcassclassification categories. No substantial improvements inprediction accuracy were gained by including the weightof excess fat trim from the sirloin as a covariate in thecurrent analysis; yet the use of sirloin bone weight as anadditional covariate did offer promising improvements inprediction of fillet SMY%. These findings should be vali-dated on a larger dataset with a wider range of SMY%,conformation and fatness scores. Accurate prediction ofthe carcass sirloin region SMY% is hugely important if thebeef industry wishes to adopt a value-based marketingapproach to carcass evaluation – especially if meat eatingquality parameters are to be included. Further refinementto current carcass evaluation systems is required toaddress this point.

Conflict of interest statement

None.

Acknowledgements

This research was funded by Quality Meat Scotland andthe C. Alma Baker Trust as part of C.R. Craigie's Ph.D.research. The support of EþV GmbH and abattoir staff isgreatly appreciated. The technical assistance of LesleyDeans (SRUC) is also greatly appreciated.

References

Allen, P., 2003. Beef carcass grading in Europe and USA – the prospects forusing VIA systems. In: Proceedings of 49th International Congress ofMeat Science and Technology. Campinas, Brazil, pp. 96–101.

Allen, P., 2005. Evaluating video image analysis (VIA) systems for beefcarcass classification. In: Proceedings of the 8th annual Langford FoodIndustry Conference, vol. 8, pp. 9–11.

Allen, P., 2007. New methods for grading beef and sheep carcasses. In:Lazzaroni, C., Gigli, S., Gabiña, D. (Eds.), Evaluation of Carcass andMeat Quality in Cattle and Sheep, Wageningen Academic Publishers,Wageningen, pp. 39–47. (EAAP Publication No. 123).

Allen, P., Finnerty, N., 2000. Objective Beef Carcass Classification — AReport of a Trial of Three VIA (Video Image Analysis) ClassificationSystems. The National Food Centre, Teagasc and Department ofAgriculture, Food and Rural Development, Dublin, pp. 1–34.

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178 177

Author's personal copy

Allen, P., Finnerty, N., 2001. Mechanical grading of beef carcasses. Teagasc,19 Sandymount Avenue Ballsbridge Dublin 4, 1–28.

Augustini, C., Dobrowolski, A., Spindler, M., 1997. Zur Videobildauswer-tung an Schlachtkörpern von Jungbullen. Mitteilungsblatt BAFF, derBundesanstalt für Fleischforschung 36, 117–122.

Borggaard, C., Madsen, N.T., Thodberg, H.H., 1996. In-line image analysisin the slaughter industry, illustrated by beef carcass classification.Meat Science 43, S151–S163.

Branscheid, W., Dobrowolski, A., Spindler, M., Augustini, C., 1998. Instru-mental classification and determination of the slaughter value forcattle with help of the video image analysis. Fleischwirtschaft 78,1046.

Brinkmann, D., 2007. Untersuchungen über die Eignung der Video-Image-Analyse (VIA) des VBS 2000 zur Beurteilung der Schlachtkör-perqualität von Kälbern und Jungbullen. University of Bonn1–150(unpublished Ph.D. thesis).

Conroy, S.B., Drennan, M.J., Kenny, D.A., McGee, M., 2009. The relation-ship of live animal muscular and skeletal scores, ultrasound measure-ments and carcass classification scores with carcass composition andvalue in steers. Animal 3, 1613–1624.

Conroy, S.B., Drennan, M.J., Kenny, D.A., McGee, M., 2010a. The relation-ship of various muscular and skeletal scores and ultrasound mea-surements in the live animal, and carcass classification scores withcarcass composition and value of bulls. Livestock Science 127, 11–21.

Conroy, S.B., Drennan, M.J., McGee, M., Keane, M.G., Kenny, D.A., Berry, D.P., 2010b. Predicting beef carcass meat, fat and bone proportions fromcarcass conformation and fat scores or hindquarter dissection. Animal4, 234–241.

Craigie, C.R., Navajas, E.A., Purchas, R.W., Maltin, C.A., Bünger, L., Hoskin,S.O., Ross, D.W., Morris, S.T., Roehe, R., 2012. A review of thedevelopment and use of video image analysis (VIA) for beef carcassevaluation as an alternative to the current EUROP system and othersubjective systems. Meat Science 92, 307–318.

Cross, H.R., Whittaker, A.D., 1992. The role of instrument grading in a beefvalue-based marketing system. Journal of Animal Science 70,984–989.

Drennan, M.J., Keane, M.G., McGee, M., 2007. Relationship of liveanimal scores/measurements and carcass grades with carcasscomposition and carcass value of steers. In: Lazzaroni, C., Gigli, S.,Gabiña, D. (Eds.), Evaluation of Carcass and Meat Quality in Cattleand Sheep, Wageningen Academic Publishers, Wageningen,pp. 159–170. (EAAP Publication No. 123).

Drennan, M.J., McGee, M., Keane, M.G., 2008. The value of muscular andskeletal scores in the live animal and carcass classification scores asindicators of carcass composition in cattle. Animal 2, 752–760.

European Community, 1981. Commission Regulation (EEC) No 2930/81 of12 October 1981 adopting additional provisions for the application ofthe community scale for the classification of carcases of adult bovineanimals. Official Journal of the European Communities 293/L,143–144.

European Community, 1981. Council Regulation (EEC) No 1208/81 of 28April 1981 determining the community scale for the classification ofcarcases of adult bovine animals. Official Journal of the EuropeanCommunities 123/L, 191–193.

Fisher, A.L., 2007. Beef carcass classification in the EU: an historicalperspective. In: Lazzaroni, C., Gigli, S., Gabiña, D. (Eds.), Evaluation ofCarcass and Meat Quality in Beef and Sheep, Wageningen AcademicPublishers, Wageningen, pp. 19–30. (EAAP publication No. 123).

Fisher, A.V., Bayntun, J.A., 1984. The effect of breed type on the relativevariability of different fat depots in cattle. Animal Production 38, 543.

Johnson, P.L., Purchas, R.W., McEwan, J., Blair, H.T., 2005. Carcass compo-sition and meat quality differences between pasture-reared ewe andram lambs. Meat Science 71, 383–391.

Jørgenvåg, I.A., Røe, M., Malmo, T., Horsen, P.M., Alvseike, O., 2009.Prediction of yield from carcass grading by EUROP and video imageanalysis. In: Proceedings of the 55th International Congress of MeatScience and Technology. Copenhagen, Denmark, pp. 18–19.

Keane, M.G., Connolly, J., Muldowney, D., 2000. Relationships betweencarcass grade and composition in beef cattle. In: Proceedings of theAgricultural Research Forum, pp. 105–106.

Kempster, A.J., 1986. Estimation of the carcass composition of differentcattle breeds and crosses from conformation assessments adjustedfor fatness. Journal of Agricultural Science 106, 239–254.

Kempster, A.J., Cook, G.L., Smith, R.J., 1980. The evaluation of a standar-dized commercial cutting technique for determining breed differ-ences in carcass composition. Journal of Agricultural Science 95,431–440.

Kempster, A.J., Harrington, G., 1980. The value of ‘fat-corrected’ confor-mation as an indicator of beef carcass composition within andbetween breeds. Livestock Production Science 7, 361–372.

McPhee, M., Oltjen, J., Fadel, J., Mayer, D., Sainz, R., 2009. Parameterestimation and sensitivity analysis of fat deposition models in beefsteers using acslXtreme. Mathematics and Computers in Simulation79, 2701–2712.

Oliver, A., Mendizabal, J.A., Ripoll, G., Albertí, P., Purroy, A., 2010.Predicting meat yields and commercial meat cuts from carcasses ofyoung bulls of Spanish breeds by the SEUROP method and an imageanalysis system. Meat Science 84, 628–633.

Pabiou, T., Fikse, W.F., Cromie, A.R., Keane, M.G., Nasholm, A., Berry, D.P.,2011. Use of digital images to predict carcass cut yields in cattle.Livestock Science 137, 130–140.

Pabiou, T., Kreuchwig, L., Cromie, A.R., Drennan, M.J., Keane, M.G., Fikse,W.F., Naesholm, A., Berry, D.P., 2009. Use of Digital Images to PredictCarcass Cut Yields In cattle. Interbull Meeting 2009. Barcelona, Spain.

Purchas, R.W., 2003. Factors affecting carcass composition and beefquality. In: Smeaton, D.C. (Ed.), Profitable Beef Production — A Guideto Beef Production in New Zealand, New Zealand Beef Council,Auckland, pp. 124–152.

Purchas, R.W., Fisher, A.V., Price, M.A., Berg, R.T., 2002. Relationshipsbetween beef carcass shape and muscle to bone ratio. Meat Science61, 329–337.

Purchas, R.W., Wilkin, G.H., 1995. Characteristics of lamb carcasses ofcontrasting subjective muscularity. Meat Science 41, 357–368.

Rius-Vilarrasa, E., Bunger, L., Maltin, C., Matthews, K.R., Roehe, R., 2009.Evaluation of video image analysis (VIA) technology to predict meatyield of sheep carcasses on-line under UK abattoir conditions. MeatScience 82, 94–100.

Ross, D.W., Craigie, C., Maltin, C., Lambe, N.R., Hyslop, J.J., Navajas, E.A.,Roehe, R., 2011. Assessment of animal, carcass and meat quality usingimage based techniques. In: Proceedings of the 1st InternationalSymposium on Animal Evaluation and Meat Quality: Technologies forCarcass and Meat Quality assessment. University of Sao Paulo,Pirassununga, pp. 37–42.

Sonnichsen, M., Dolbrowolski, A., Spindler, M., Branscheid, W., Brink-mann, D., 2006. Video image analysis of calf carcases. Fleisch-wirtschaft 86, 107–110.

Sørensen, S.E., Klastrup, S., Petersen, F., 1988. Classification of bovinecarcasses by means of video image analysis and reflectance probemeasurements. In: Proceedings of the 34th International Congress ofMeat Science and Technology, pp. 635–638.

Todd, D.L., Wolliams, J.A., Roughsedge, T., 2011. Gene flow in a nationalcross-breeding beef population. Animal 5, 1874–1886.

C.R. Craigie et al. / Livestock Science 158 (2013) 169–178178