See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/51781537

Prediction of water-holding capacity and

composition of porcine meat by comparative
spectroscopy

Article in Meat Science June 2000

DOI: 10.1016/S0309-1740(99)00141-2 Source: PubMed

CITATIONS READS

145 640

6 authors, including:

L. Munck Poul Henckel

University of Copenhagen Aarhus University
102 PUBLICATIONS 2,360 CITATIONS 67 PUBLICATIONS 2,361 CITATIONS

SEE PROFILE SEE PROFILE

Anders H Karlsson Sren Balling Engelsen

Swedish University of Agricultural Sciences University of Copenhagen
96 PUBLICATIONS 3,482 CITATIONS 288 PUBLICATIONS 7,034 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

MACSYS - Development of an objective method to perform quality classification of comminuted

poultry meat. View project

All content following this page was uploaded by Sren Balling Engelsen on 04 March 2017.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 Meat Science 55 (2000) 177185
www.elsevier.com/locate/meatsci

Prediction of water-holding capacity and composition of porcine

meat by comparative spectroscopy
Jesper Brnduma,b, Lars Muncka, Poul Henckelc, Anders Karlssonc,
Eva Tornbergd, Sren B. Engelsena,*
a
Food Technology, Royal Veterinary and Agricultural University, Rolighedsvej 30, 1958 Frederiksberg, Denmark
b
SFK Technology, Transformervej 9, 2730 Herlev, Denmark
c
Department of Animal Product Quality, Danish Institute of Agricultural Sciences, Research Centre Foulum, PO Box 50, 8830 Tjele, Denmark
d
Department of Food Engineering, Lund University, PO Box 124, SE-22100 Lund, Sweden

Received 10 July 1999; accepted 7 September 1999

Abstract
Four spectroscopic instruments, a bre optical probe (FOP), a visual (VIS) and near infrared (NIR) reectance spectro-
photometer, a reectance spectrouorometer and a low-eld 1H nuclear magnetic resonance (LF-NMR) instrument were used to
perform measurements on two muscles (longissimus dorsi and semitendinosous) from 39 pigs, 18 of which were carriers of the
Halothane gene. Water-holding capacity (drip loss and lter paper wetness) and chemical composition (intramuscular fat and
water) of the muscle samples were determined for spectroscopic calibration. Prediction models were established by partial least
squares regression to evaluate the potential of using the spectroscopic techniques in an on-line slaughterhouse system. VIS data
gave good prediction models, indicating that current industrial colour systems can be advanced into more specic meat evaluation
systems by including the entire visible spectral range. The FOP and uorescence measurements were less successful, and suered
from sampling problems since they measure only a small area. The best regression models were obtained from LF-NMR data for all
reference quality measures and yielded a correlation coecient of 0.75 with drip loss. LF-NMR proved able to distinguish between
the two muscles and the results for their longitudinal relaxation times, T21, were proportional to their average myobrillar cross-
sectional areas reported in the literature. # 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Chemometrics; Fluorescence; VIS; NIR; NMR; Porcine meat; WHC

1. Introduction have proposed the use of ultraviolet uorescence for

Instrumental techniques for rapid screening of meat
properties (e.g. water-holding capacity and chemical
composition) which can be used to improve control and
classication of the product in mechanised processes are
of great interest for both the industry and the consumers.
The use of rapid spectroscopic techniques for (on-line)
quality inspection of meat has been successfully demon-
strated in a few applications. Jensen, Reenberg and
Munck (1989) used ultraviolet uorescence for detection
of the tissue components of meat. Swatland (1995a),
Swatland, Brooks and Miller (1998), Swatland and Find-
lay (1997), Swatland, Gullett, Hore and Buttenham
(1995b) and Swatland, Nielsen and Andersen (1995c)
* Corresponding author. Tel.: +45-35-28-3205; fax.: +45-3528-3245.
E-mail address: se@kvl.dk (S.B. Engelsen).
0309-1740/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0309-1740(99)00141-2
measurement of connective tissue in relation to tender-
ness in beef. The technique was validated by measure-
ments on muscle of sensory properties, dynamic
electromechanical toughness, and resistance to penetra-
tion of a needle. Swatland (1995a) used visual reec-
tance to determine water-holding capacity, pH and light
reectance of porcine meat. Borggaard, Andersen and
Barton-Gade (1989) and Andersen, Borggard and Niel-
sen (1995) studied the relationships between a single
near infrared wavelength (950 nm) and the water-hold-
ing capacity (WHC) and intramuscular fat (IMF) in
porcine meat. The NIR study on WHC by Borggaard et
al. has recently been extended by taking measurements
with bre optic probes at many NIR wavelengths by
Forrest, Sheiss, Morgan and Gerrard (1997).
There are many reports of the use of low-eld nuclear
magnetic resonance (LF-NMR) to measure water and
178 J. Brndum et al. / Meat Science 55 (2000) 177185
its distribution in meat, especially pork muscle. NMR
imaging is widely used for diagnostic purpose in medical
applications. From measurements on the muscles it can
distinguish between pigs carrying the Halothane-posi-
tive gene from normal animals (Decanniere, Van Hecke,
Vanstapel, Ville & Geers, 1993). Low-eld NMR has
been applied to measure and predict meat quality prop-
erties in several studies (Borisova & Oreshkin, 1992;
Fjelkner-Modig & Tornberg, 1986; Renou, Monin &
Sellier, 1985; Renou, Kopp, Gatellier, Monin & Kozak-
Reiss, 1989; Tornberg, Andersson, Goransson & von
Seth, 1993). It is a technique that can realistically be
adapted to measurement on-line or at-line and Trout
(1988) recommends NMR as a means of measuring of
water-binding in muscle.
Garrido, Padauye, Banon, Lopez and Laencina,
(1995) discussed the use of electrical conductivity, pH
and light scattering for on-line classication of extremes
of porcine meat properties. They considered that pH
proved to be the best discriminator at 45 min post
mortem, but none of these techniques were satisfactory
alone. They found, however, that combining pH and
light scattering resulted in reasonable classication of the
``quality'' groups. Despite intensive research eorts, no
instrumental method has yet been developed that can be
used online in industry to measure or predict the func-
tional properties of pork meat. The closest approaches to
the use of spectroscopic instruments online for carcass
grading measurements are the Colormeter Colour vision
(BCC-2, SFK Technology, Herlev, Denmark), the use of
the reectance value from the Fat-O-Meater (FOM;
SFK Technology) and the Hennessy probe (Hennessy
Europe, EH Rijewijk, NL) from seem to be the closest
step to on/at-line use of spectroscopy so far. Few com-
parisons of spectroscopic techniques in relation to meat
properties and meat composition have been reported.
The aim of this study was therefore to compare the
potential of dierent spectroscopic techniques for on-
line measurements to predict the WHC parameters and
chemical composition (water and fat contents) of por-
cine meat. A wide range of spectroscopic techniques
were selected for this study, an insertion bre optical
probe (FOP), visible (VIS) and near infrared (NIR)
reectance spectroscopy, uorescence spectroscopy and
low-eld 1H nuclear magnetic resonance (LF-NMR).
To create a large variation in the muscle samples, two
conditions were used. First, stress sensitive and non-
stress sensitive animals were used to provide meat with
dierent WHC. Secondly, samples were taken from
both the loin muscle (M. longissimus dorsi, LD) and the
ham muscle (M. semitendinosus, ST) in order to have
muscles of dierent composition. The techniques of
exploratory multivariate analysis (Martens & Ns, 1993)
were used as essential tools to investigate the highly
covariate data structure derived from the spectroscopic
analyses.
2. Material and methods
2.1. Experimental design and meat sampling
A total of 39 DDLY (Danish Duroc as terminal sire and
Danish LandraceLarge White dams) pigs were slaugh-
tered (live weight 105 kg) at the Research Centre Foulum
during three dierent weeks. Within litter group and
slaughter week, 18 of the 39 pigs were free of the Halo-
thane gene (HAL-NN) and 21 of the 39 pigs were hetero-
zygote (carriers) for that gene (HAL-Nn). At 24 h post
mortem FOP measurements were collected from fresh
meat samples of approximately 300 grams which were
dissected from the loin muscle (LD) at the 10th rib and
from one ham muscle (ST). A subsample of approxi-
mately 50 g from each muscle type was frozen and stored
at 20 C until measurements of water and IMF contents
were performed. Approximately 100 g from each sample
was used for the water-holding capacity determination. At
24 h post mortem, the rest of the fresh meat samples were
kept at 8 C and transported (4 h) to the spectroscopy
laboratory. NIR, VIS, uorescence, and LF-NMR mea-
surements were collected at 4854 h post mortem. LF-
NMR was only measured in the two rst weeks on sam-
ples from 26 pigs, and uorescence was only measured in
the two last weeks on samples from 26 other pigs.
2.2. Water-holding capacity and chemical analyses
Water-holding capacity was measured as (1) drip loss,
whereby the loss of water from a 2.5 cm thick slice of
muscle taken 24 h post mortem hanging for 2 days in
double plastic bags at +4 C was registered (Honikel,
1987), and as (2) by lter paper wetness expressed by
percentage weight gain (Kaufmann, Eikelenboom, van
der Waal, Engel & Zaar, 1986). Fat was determined as
Stoldt-Fat (Stoldt, 1952), and water was measured by
weighing the sample before and after freeze-drying fol-
lowed by 4 h of heating at 100 C.
2.3. Fibre optic insertion probe (FOP)
Internal reectance was measured with a custom
designed FOP (Ocean Optics Europe, Top Sensor Sys-
tems, Eerbek, NL) optimised for meat use with a sharp
tip, thus enabling insertion into the meat through the
skin. This approach is a potential on-line sampling sys-
tem, while the sampling with the other spectroscopic
systems used in this study were typical for laboratory
equipment. The FOP system is constructed with a dou-
ble channel optical cable, with one channel optimised
for the ultraviolet to visual (UVVIS) range (280730
nm) and one channel optimised for the visual to near
infrared (VISNIR) range (500980 nm). Each channel
consists of seven bres, one for illumination of the meat
and six for collection of the surface reectance. Thus, a
 J. Brndum et al. / Meat Science 55 (2000) 177185
total of 14 bres exist in the optic cable. Each channel is
connected to an illumination source and a spectro-
photometer. A total area with a diameter of approximately
5 mm is measured with the probe. The UV-illumination
source is a combined deuteriumhalogen lamp (DH-2000)
and for the VIS a halogentungsten lamp (HG-1000) is
employed. The spectrophotometers for the two channels
are combined in a dual line spectrophotometer (PSD-1000)
connected externally to a portable PC via a PCMCIA pro-
tocol. The spectra are measured at a sampling frequency of
33 kHz by averaging four samples. Using the reference
measurement of a BaSO4 solution the raw reectance
measurement is transformed into absorbance units.
2.4. Near infrared and visual reectance
Near infrared and visual reectance (NIR and VIS)
were recorded with a NIR Systems 6500 (Silver Spring,
MD, USA). The incident light was illuminated on the
sample from 180 and reectance was measured at a 45
angle. The samples were positioned in a sample cup
equipped with a quartz window with a diameter of 25
mm. A compressible paper-disk was used to force the
slice of meat against the quartz window. The measured
spectra were separated into the VIS range from 400 to
800 nm and the NIR range 802 to 2500 nm.
2.5. Fluorescence
Fluorescence data were collected on a PE-LS-50B (Per-
kinElmer Ltd., Buckinghamshire, UK) grating spectro-
photometer on a round fragment of the meat positioned in
a solid sample holder with a uorescence-free quartz win-
dow. A pulsed xenon lamp excites the sample, and the
uorescence signal is registered with a photomultiplier
detector. A scan speed of 8 nm/min and a slit width of 8
nm were used on both excitation and emission. Complete
excitation-emission uorescence landscapes were col-
lected from two meat samples that covered the main var-
iation in the data material. Four excitation wavelengths:
lex1=280 nm, lex2=320 nm, lex3=365 nm, and lex4=395
nm were selected from these uorescence landscapes for
subsequent measurements. Emission spectra were recor-
ded from the excitation wavelength plus 20 nm, to avoid
Rayleigh scattering, and up to 600 nm.
2.6. Low-eld 1H nuclear magnetic resonance
LF-NMR measurements were performed on a Maran
Benchtop Pulsed NMR analyser (Resonance Instru-
ments, Witney, UK), operating at a frequency of 23.2
MHz and equipped with an 18 mm variable temperature
probe head. Transverse relaxation, T2, was measured
using the CarrPurcellMeiboomGill sequence
(CPMG; Carr & Purcell, 1954; Meiboom & Gill, 1958),
and longitudinal relaxation (T1) was measured using the
179
inversion recovery sequence (INVREC; Vold, Waugh,
Klein & Phelps, 1968). LF-NMR relaxation data were
acquired as 16 scan repetitions for the CPMG pulse
sequences and four scan repetitions for the INVREC
pulse sequences using a two-second relaxation delay
between successive pulse scans. The samples were equili-
brated for 20 min at the desired measurement temperature
(26 C) and the sample probe temperature was kept con-
stant at 26 C by a continuous airow. The meat samples
were introduced into the NMR probe by placing cylind-
rical samples (14 mm dia.) stamped out of the centre of the
meat samples into sealed glass tubes, which matched the
inner diameter of the 18 mm NMR sample tubes.
2.7. Data analysis
The numerous spectral data points obtained from the
spectrophotometers were processed using multivariate
data analysis with the purpose of developing calibration
models for predicting the reference information from
the water-holding capacity traits and from the chemical
composition of the meat. Predictions based on spectral
information were performed with partial least squares
regression (PLSR) which (after centering of the data)
projects the spectral data onto common orthogonal
structures, called latent variables, by describing the
maximum covariance between the spectral information
and the references. In spectroscopy the numerous data
points are highly co-linear and can thus be reduced to
only a few factors called scores. These scores give a more
robust and concentrated representation of the samples,
discarding only a minimum of non-systematic informa-
tion (Martens & Ns, 1993). The latent variables and the
scores combined form the principal components (PC).
In this study predictions are validated with full cross-
validation (leave one out), which is an accepted method
for estimating the number of PC's and the standard error
of prediction (SEP) (Martens & Dardenne, 1998; Martens
and Ns, 1993). The multivariate data analysis was
performed with the chemometric program Unscrambler
7.1 (CAMO, Trondheim, Norway).
Bi-exponential tting analysis of longitudinal relaxa-
tion data was carried out using an in-house program
(Bechmann, Pedersen, Nrgaard & Engelsen, 1999)
written in Matlab (The Mathworks Inc., Natick, MA,
USA). By this procedure characteristic relaxation time
constants, T21 and T22, and their corresponding ampli-
tude parameters, M21 and M22, were extracted for each
longitudinal (CPMG) LF-NMR measurement.
3. Results and discussion
Statistics of the reference parameters for all the sam-
ples combined and separated into four groups according
to HAL-genotype and the muscle group are shown in
180 J. Brndum et al. / Meat Science 55 (2000) 177185

Table 1. The variation in drip loss of the included meat

tends to be above the average level for porcine meat in
Denmark (Karlsson, Henckel, Pedersen & Andersson,
1997). This is a known eect when pigs carrying the
halothane positive gene are included in an experimental
group. The correlation between the two WHC methods
was 0.61. This relatively low correlation indicated that
the two methods represent dierent information on the
WHC in meat, as previously discussed by Trout (1988).
While the drip loss reveals the amount of free water that
exudes under the force of gravity from the muscle bres,
the lter paper method measures the water that is
extracted by applying a capillary force to the meat. The
force applied in the lter paper method is, however, too
small to cause any damage to the meat structure and is
therefore not likely to be responsible for the dierences
observed between the two methods. The lter paper
Fig. 1. Spectra from two pork meat samples collected with the FOP
method suers from a large dependence on the rough-
system. The UVVIS channel to the left and the VISNIR channel to
ness and composition of the meat surface where the l- the right: LD muscle ( ); ST muscle ( ).
ter paper is applied. This is also reected in a very large
standard deviation (Table 1). The drip loss and lter
paper wetness were signicantly (P<0.001) higher for Soret band) and 555 nm for deoxymyoglobin. It appears
the LD muscle than for the ST muscle for both geno- from the VIS part of the FOP UVVIS and from the
types. With respect to the mean within each muscle FOP VISNIR data that the absorbance values above
type, the only signicant (P<0.05) dierence between 450 nm are higher for the LD samples. This dierence
the genotypes was the higher drip loss for HAL-Nn between the two muscles is generally observed for all
compared to HAL-NN. Also, the levels of intramus- measurements (data not shown) and reects scattering
cular fat (IMF) appear to be higher than normal varia- properties of the muscles. Despite the fact that three
tion, especially for the ST muscle. However, low insertions were made with the FOP, it is possible that the
variation in the water content of the meat is observed small area measured was not suciently representative.
for both muscles and both genotypes. The NIR Systems VIS spectra for the LD and the ST
The absorbance FOP spectra for the LD and ST muscles of two normal samples are shown in Fig. 2.
muscles for two normal samples are shown in Fig. 1 for Peaks in the log(1/R) data were observed at 430 and 560
both spectroscopic channels. The FOP UVVIS spectra nm, very close to the absorbance peaks observed for the
show two absorbance peaks at 425 and 550 nm, whereas FOP data (Fig. 1). Consistent with the FOP measure-
the FOP VISNIR spectra display one absorbance peak ments, the correlation between the absorbance peaks
at 545 nm. The correlation between these two absor- was high (r=0.96), and the peaks are thus likely to be
bance peaks for the FOP UVVIS spectra is 0.96. This caused by the existence of myoglobin. The NIR spectra
high correlation indicates that the two peaks are caused of the LD and ST muscles from two standard samples
by the same phenomenon, namely myoglobin. Swatland are shown in Fig. 3. Local maxima in the NIR data are
(1995a) reported absorption maxima at 434 nm (for the found at 980 nm (OH stretch, second overtone), 1450

Table 1
Measurements of drip loss under gravity, moisture loss by the lter paper method, intramuscular fat (IMF) and total moisture content, expressed as
overall means and standard deviations, in the l. dorsi (LD) and semitendinosus (ST) muscles from DDLY pigs either carriers (HAL-Nn) or not
(HAL-NN) of the halothane sensitive gene

All samples LD ST

Hal-Nn Hal-NN Hal-Nn Hal-NN

(N=78) (N=18) (N=21) (N=18) (N=21)
Mean SD Mean SD Mean SD Mean SD Mean SD

Drip loss (%) 6.1 3.2 9.4 1.9 7.1 2.5 4.2 2.5 3.5 2.0
Filter paper (%) 38.0 20.4 46.0 23.0 36.7 15.7 36.6 19.9 31.2 21.0
IMF (%) 3.6 1.9 2.7 1.3 2.7 1.5 4.7 2.4 4.5 1.6
Water (%) 72.9 1.4 72.7 1.0 72.7 1.2 73.3 1.4 75.5 2.3
 J. Brndum et al. / Meat Science 55 (2000) 177185
Fig. 2. Typical VIS spectra from two pork meat samples obtained on
the LD muscles ( ) and the ST muscle ( ).
Fig. 3. Typical NIR spectra from two pork meat samples: LD muscles
( ), ST muscle ( ).
(OH stretch, rst overtone), 1920 nm (OH stretch+O
H deformation) (all due to water) and 1200 nm (CH
stretch, second overtone) and 1800 nm (CH stretch, rst
overtone or combination band). Spectral information due
to proteins is usually found around 1500 nm (NH stretch,
rst overtone) and above approx. 2000 nm (combination
bands), but the protein information in the spectra of Fig. 3
is hidden by the large signal from water.
Fluorescence emission peaks for both muscles from
the same pig were observed at 455 nm for lex=320 nm,
at 460 nm for lex=365 nm (with a signicant shoulder
at 485 nm) and at 465 nm for lex=395 nm (Fig. 4).
When using the excitation wavelength, lex, of 280 nm
only a weak shoulder emission was observed at about
455 nm. In the sample displayed in Fig. 4 it is remark-
able that both the scatter and the emission of the ST
181
Fig. 4. Four uorescence emmision spectra obtained from one sample
of LD muscle ( ) and one sample of ST muscle ( ).
muscle is lower at all excitations, however this trend
was not general. The most signicant uorophore is
observed around 460 nm for all excitations. Two factors
are believed to contribute to the uorescence. The rst is
the co-enzyme, NADH, which is involved in muscle
energy metabolism. An emission peak at 460 nm for
excitation 366 nm was explained as being due to NADH
by Jensen et al. (1989). Secondly, Swatland (1995a)
explained uorescence emission peaks at 440 nm (exci-
tation at 375 nm) in beef by the presence of connective
tissue (collagen type I and III). In our experiment the
Rayleigh peaks for especially lex=280 nm and lex=320
nm were very wide, and for the ST sample, the uores-
cence signal was almost submerged into the Rayleigh
signal. This is likely to be due to the scatter eect from
the UV reection of the fat, a conclusion supported by
the higher eect observed for the ST sample (see Table 1).
A higher level of connective tissue in ST than in LD is also
most likely a main factor for the separation of the two
muscles by the uorescence signal. We found no separa-
tion between the meats from HAL-NN and HAL-Nn pigs
from scrutiny of the NADH uorescence parameter (ex.
395 nm, em. 460 nm).
LF-NMR on samples with high water content, such
as meat, generally provides information on the state of
the water or, more precisely, on diusive domains in the
samples. The LF-NMR relaxation curves for LD and
ST from two pigs are shown in Fig. 5. From both, the
INVREC (T1) relaxation and the CPMG relaxation
curves, a dierence between the LD and the ST muscles
is observed (see also Table 2).
Bi-exponential analysis (two-component model) of the
CPMG relaxation curves was judged to be appropriate
to describe the muscle samples. The CPMG measure-
ments were initially performed using ve dierent values
of , which is half of the 180 interpulse spacing. Fig. 6
182 J. Brndum et al. / Meat Science 55 (2000) 177185

LD compared to ST (see Table 1). However, the dier-

Fig. 5. LF-NMR relaxation curves (INVREC and CPMG) obtained
from two samples of LD muscle ( ) and two samples of ST muscle ( ).
displays the  dependence of the four t parameters:
T21, T22, M21 and M22 on one selected sample. There is
a linear response regime between 100 and 1000 ms and
we have accordingly selected the CPMG relaxation at
t equal 500 ms for subsequent analysis. The statistics
for the amplitude (M21 and M22) and the time variables
(T21 and T22) are shown in Table 2. A signicant
(P<0.01) dierence between the two muscles is seen for
M22, T21 and T22, but most prominently for T21. Fig. 7
shows T21 for all 52 samples, displaying a clear separa-
tion between LD and ST (mean of 38 and 43 ms,
respectively) muscles. The parameter T21 has previously
has been related to sensory and rheological attributes of
meat and the signal has been considered as being due to
intracellular water (Tornberg et al., 1993).
The type of bres in the muscle has often been related
to meat quality parameters (Swatland, 1995a). Petersen,
Henckel, Oksbjerg and Srensen, (1996) reports 81 and
74% Type IIB muscle bres for LD and ST, respec-
tively. This bre type exhibits fast glycolysis post mor-
tem with the consequent decrease in pH resulting in
lower WHC a fact consistent with the lower WHC for
ence in proportions of the bre type does not account
for the signicant dierence in T21 between the two
muscles. Petersen et al. (1996) also reported a sig-
nicantly higher mean cross-sectional area for the mus-
cle bre types in ST (5354 mm2) than in LD (4763 mm2).
While T21 provides information regarding the diusive
restrictions within the muscles, it is also highly likely
that T21 is directly related to the geometric restrictions
within the myobrils of the muscles. We therefore pro-
pose that it is this characteristic dierence in the dia-
meter between the two muscles which is measured by
the dierence in the T21. There is a close proportional
relationship between the cross-sectional bre area of
two muscles and the T21 result: bre area/T21 being
4763/38=125.3 for LD and 5354/43=124.5 for ST.
This preliminary conclusion is supported by a recent
LF-NMR study on beef, in which post-mortem swelling
in the myobrillar system, causing the mean cross-sec-
tional area to increase, was followed by an increase in
T21 (Tornberg, Wahlgren, Brndum & Engelsen, 1999).
Table 3 presents the prediction results for the spec-
troscopic techniques. For the WHC parameters, LF-
NMR was superior among the spectroscopic techniques
with correlation values of 0.75 to drip loss and 0.53 to
lter paper wetness. The relation between the predicted
and the measured drip loss data from the CPMG mea-
surements is displayed in Fig. 8. Compared to a repeti-
tion coecient of r=0.89 for the drip loss method and
0.88 for the lter paper method (Karlsson, unpublished
results), the prediction results are encouraging. The
CPMG measurements resulted in a slightly better per-
formance (SEP) than the T1 measurements (except for
the IMF predictions). The CPMG data is acquired in a
few seconds, whereas the acquisition time for the T1
data exceeded 5 min. Thus, the potential for using the
CPMG data in an online measuring system in industry
is more realistic. The usual approach for analysing LF-
NMR relaxation curves is to use exponential data t-
ting. However, for predictive purposes, there tends to be
an improvement by including the entire information of
the relaxation curves in a multivariate calibration. As

Table 2
Overall means and standard deviations for the LF-NMR t parameters obtained from measurements on the l. dorsi (LD) and semitendinosus (ST)
muscles from DDLY pigs, either carriers (HAL-Nn), or not (HAL-NN), of the halothane sensitive gene

All samples LD ST

Hal-Nn Hal-NN Hal-Nn Hal-NN

(N=52) (N=13) (N=13) (N=13) (N=13)
Mean SD Mean SD Mean SD Mean SD Mean SD

M21 4935 197 4989 254 4926 170 4901 89 4917 232
M22 1033 59 948 154 963 84 1117 180 1118 124
T21(ms) 41 3 38 1 38 1 43 2 43 2
T22 (ms) 104 12 99 7 95 11 113 14 109 8
 J. Brndum et al. / Meat Science 55 (2000) 177185
Fig. 6. The relation between the interpulse spacing, , and the amplitude for the LF-NMR CPMG data.
Fig. 7. The T21 values for all 52 samples measured with LF-NMR
CPMG data. A dierence between the LD and ST muscles is observed.
seen in Table 3, all four reference parameters are pre-
dicted better with multivariate analysis than by using
quantitative parameters from the exponential t (M21
and M22). This is in good agreement with investigations
made by Jepsen, Pedersen and Engelsen, (1999) who
found signicantly better correlations of above 0.9 from
LF-NMR data to the WHC of cod meat. The WHC
method used in the sh trial involved centrifugation of
minced sh meat with measurement on the supernatant
uids, a more destructive method than either drip loss
and lter paper methods. Thus, the dierence in WHC
reference method could well explain the better correla-
183
tion in the sh study compared to the present pork meat
investigation.
The VIS data showed a good correlation (0.72; Table
3) with drip loss and we consider this relies on the
indirect relationship between the light scattering and the
extracellular space where some of the drip uid is con-
tained. Measurements from monochromatic reectance
and colour instruments have been investigated for their
relation to WHC by Swatland (1995a). Monochromatic
measurements are biased by the strong eect of myo-
globin, and 3-factor colour instruments are optimised
for reproduction of colour interpretation by the human
eye. This optimisation is justied when evaluating the
response of consumers to the meat quality, but is not
necessarily optimal for prediction of WHC. The results
obtained in this study indicate that whole spectra VIS
data indirectly can be related to WHC using multi-
variate calibration. Such data sets should be optimal in
describing both WHC and the visual colour of the meat.
The uorescence method has a slightly lower prediction
(r=0.68) of drip loss than VIS spectroscopy. It also
shows a slight signicance in predicting IMF. The spec-
tral instruments are not strictly comparable in this
investigation. The uorescence instrument measures an
area of 19 mm, while the VISNIR instrument mea-
sures an area 20 mm in diameter, the FOP an area with
a diameter of 2 mm and the LF-NMR instrument a
184 J. Brndum et al. / Meat Science 55 (2000) 177185

Table 3
Correlation coecients (r) and standard error of prediction (SEP) and number of principal components (PC) in the PLSR models between various
spectral techniques and compositional parameters of pork meat. SEP values are absolute (%)

Drip loss Filter paper IMF Water

r SEP #PC r SEP #PC r SEP #PC r SEP #PC

FOP 0.61 2.53 4 0.26 12.92 2 0.15 6.29 1 0.35 1.17 1

VIS 0.72 2.14 2 0.54 18.04 2 0.52 1.61 3 0.46 1.05 4
NIR 0.64 2.43 4 0.62 16.01 4 0.70 1.32 5 0.46 1.13 4
Fluorescence 0.68 2.27 5 0.43 10.41 3 0.57 1.09 4 0.16 0.93 1
LF-NMR (INVREC) 0.74 2.07 3 0.46 19.35 2 0.77 1.13 3 0.64 0.92 2
LF-NMR (CPMG) 0.75 2.03 3 0.53 18.55 3 0.68 1.27 3 0.67 0.83 1
LF-MNR Experimental ttinga 0.72 2.14 0.61 17.11 0.57 1.43 0.55 0.97
a
This row indicates simple PLSR (equal to MLR) models using only the two quantitative variables (M21 and M22) extracted from the CPMG
data using the exponential t procedure.

The FOP measurements are of interest since this

Fig. 8. Measured versus predicted drip loss with the LF-NMR CPMG
data. The middle diagonal line shows the regression line, and the
two dotted diagonal lines show the prediction errors. The lables indi-
cations are *: LD HAL-NN, : LD HAL-Nn, +: ST HAL-NN, *:
ST HAL-Nn.
cylinder with a diameter of 14 mm and a length of 20
mm. The uorescence instrument thus measures a far
smaller sample and the weak correlations should be
further investigated with an instrument with a larger
measurement area.
When the results for the HAL-Nn samples were
examined separately, the correlations were generally
higher due to the increased variation, as displayed in
Table 1; e.g. the correlation for the LF-NMR INVREC
data to WHC improved to r=0.81.
The LF-NMR data also gave the highest correlations
with the composition parameters IMF and total water
(Table 3) although they were not as high as those
reported by Jepsen et al. (1999) using the same NMR
equipment on sh meat. However, this dierence most
likely stems from the much higher variation in the sh
data (IMF varying from 2 to 42% and total water vary-
ing from 39 to 77%). The NIR data also showed good
predictions of IMF but failed to predict total water.
instrument could be used in industry online. The results
showed the highest correlation (0.62) with WHC but
only very weak correlations with the compositional
measures. This is perhaps due to the small sample areas
measured by the probe (dia. 2 mm) being unrepresenta-
tive despite three insertions into the meat. The range of
results presented are encouraging for further research in
spectroscopic methods for rapid measurement of func-
tional and compositional attributes of meat. The LF-
NMR and the VIS reectance predictions are especially
promising. Much eort is still required before use of the
techniques on process lines can be introduced. However,
in order to improve spectroscopic methods, it is neces-
sary also to improve the repeatability and to understand
in more detail the rather unspecic reference methods
for determination of the WHC characteristics, e.g. by
advancing the methods described by Trout (1988) for
complementary use in WHC characterization. The VIS
technique looks to be the more readily applicable for
online measurements, but a problem with the sampling
technique to ensure representative sampling is still to
be overcome. The LF-NMR sampling is advantageous
in being a volume-based rather than a surface-based
measurement, which was clearly observed from the
composition predictions. The CPMG technique pro-
vides a relatively rapid assessment of the T2 relaxation,
but the destructive sample preparation is a dis-
advantage. Ultimately, the spectroscopic technique
should be applicable early post mortem, preferably
3045 min before the carcasses are transported to the
cooling facilities.
4. Conclusion
The use of spectroscopic techniques for rapid assess-
ment of water-holding capacity (WHC) and composition
of two porcine muscles (LD and ST) has been outlined in
this paper. Visible reectance and uorescence provided
 J. Brndum et al. / Meat Science 55 (2000) 177185
signicant predictions of WHC. Near infrared reec-
tance predicted the intramuscular fat (IMF) reasonably
well. The bre optical probe and the uorescence mea-
surements suered critically from the small sample area
which could not be representative of such a hetero-
geneous material as meat. The most successful techni-
que for prediction of WHC, IMF and total water
content was 1H low-eld nuclear magnetic resonance.
We propose that the T21 exponential t parameter was
directly related to the diameter of the bres and further
work is in hand to verify this proposition. Further
comparative work between spectroscopic methods in
meat quality analysis should be done with instruments
which are capable of taking representative samples.
Acknowledgements
J. Brndum was supported by the Academy for Tech-
nical Sciences (ATV), project EF621. L. Munck and S.B.
Engelsen were supported by funds from The Danish
Research Councils (SJVF), the Danish Centre for
Advanced Studies (LMC) and by the center for Predictive
Multivariate Process analysis nanced by the Depart-
ments of Industry and Research of the Danish Govern-
ment. We are indebted to Professor Allan Bremner (DFU,
Lyngby, DK) for useful discussions of the manuscript and
to Ph.D. student Henrik T. Pedersen (KVL, DK) for
technical assistance on the benchtop NMR-equipment.
References
Andersen, J. R., Borggaard, C., & Nielsen, T. (1995). Recent advances
in real time measurements of meat quality and safety. Proceedings
ECCAMST meeting, Roskilde, Denmark.
Bechmann, I. E., Pedersen, H. T., Nrgaard, L., & Engelsen, S. B.
(1999). Comparative chemometric analysis of transverse low-eld
1
H NMR relaxation data. In P. S. Belton, B. P. Hills, & G. A.
Webb, Advances in magnetic resonance in food science (pp. 217225).
UK: The Royal Society of Chemistry.
Borggaard, C., Andersen, J. R., & Barton-Gade, P. A. (1989). Further
development of the MQM-equipment for measuring water-holding
capacity and intramuscular-fat on-line. Proc. Int. Congress Meat
Science and Technology, 35, 212219.
Borisova, M. A., & Oreshkin, E. F. (1992). On the water condition in
pork meat. Meat Science, 31, 257265.
Carr, H., & Purcell, E. M. (1954). Eects of diusion on free preces-
sion in nuclear magnetic resonance experiments. Physical Review,
94, 630638.
Decanniere, C, Van Hecke, P., Vanstapel, F., Ville, H., & Geers, R.
(1993). Metabolic response to halothane in piglets susceptible to
malignant hyperthermia; an in vivo 31P-NMR study. Journal of
Applied Physiology, 75, 955962.
Fjelkner-Modig, S., & Tornberg, E. (1986). Water distribution in
porcine M. longissimus dorsi in relation to sensory properties. Meat
Science, 17, 213231.
Forrest, J. C., Sheiss, E. B., Morgan, M., & Gerrard, D. E. (1997).
Pork quality measurement toolsnow and in the future. In NPPC
pork quality summit proceedings (pp. 7996). Des Moines, IA.
View publication stats
185
Garrido, M. D., Pedauye, J., Banon, S., Lopez, M. B., & Laencina, J.
(1995). On-line methods for pork quality detection. Food Control,
6(2), 111113.
Honikel, K. O. (1987). How to measure the water-holding capacity of
meat? Recommendation of standardized methods. In P. V. Tarrant,
G. Eikelenboom & G. Monin, Evaluation and control of meat quality
in pigs. pp. 129142. Dordrecht, NL: Martinus Nijhof.
Jensen, S. A., Reenberg, S., & Munck, L. (1989). Fluorescence analysis
in sh and meat technology. In: Fluorescence analysis in foods (pp.
181192). UK: Longman, Harlow.
Jepsen, S. M., Pedersen, H. T., & Engelsen, S. B. (1999). Application
of chemometrics to low-eld 1H NMR relaxation data of intact sh
esh. Journal of the Science of Food and Agriculture, 79(13), 1793
1802.
Karlsson, A., Henckel, P., Pedersen, J. S., & Andersson, M. (1997). Early
prediction of ultimate pig meat quality. Proceedings. 48th meeting
European Association for Animal Production, Vienna, 2528 August.
Kauman, R. G., Eikelenboom, G., Waal, P. G.van der, Engel, B., &
Zaar, M. (1986). A comparison of methods to estimate water-holding
capacity in post-rigor porcine muscle. Meat Science, 18(4), 307322.
Martens, H., & Dardenne, P. (1988). Validation and verication of
regression in small data sets. Chemometrics and Intelligent Labora-
tory Systems, 44, 99121.
Martens, H., & Ns, T. (1993). Multivariate calibration. New York:
Wiley.
Meiboom, S., & Gill, D. (1958). Modied spin-echo method for mea-
suring nuclear relaxation times. Review of Scientic Instruments, 29,
688691.
Petersen, J. S., Henckel, P., Oksbjerg, N., & Srensen, M. T. (1996).
Adaptions in muscle ber characteristics induced by physical activ-
ity in pigs. Animal Science, 66, 733740.
Renou, J. P., Monin, G., & Sellier, P. (1985). Nuclear magnetic resonance
measurements on pork of various qualities. Meat Science, 15, 225233.
Renou, J. P., Kopp, J., Gatellier, Ph., Monin, G., & Kozak-Reiss, G.
(1989). NMR-relaxation of water protons in normal and malignant
hypothermia-susceptible pig muscle. Meat Science, 26, 101114.
Stoldt, W. (1952). Vorslag zur Vereinheitlichung der Fettbestimmung
in Lebensmittel. Fette und Seifen, 54, 206207.
Swatland, H. J. (1995a). On-line evaluation of meat. USA: Technomic,
Lancaster.
Swatland, H. J., Gullett, E., Hore, T., & Buttenham, S. (1995b). UV
ber-optic probe measurements of connective tissue in beef corre-
lated with taste panel scores for chewiness. Food Research Interna-
tional, 28, 2330.
Swatland, H. J., Nielsen, T., & Andersen, J. R. (1995c). Correlations
of mature beef palatability with optical probing of raw meat. Food
Research International, 28, 403416.
Swatland, H. J., & Findlay, C. J. (1997). On-line probe prediction of
beef toughness, correlating sensory evaluation with uorescence
detection of connective tissue and dynamic analysis of overall
toughness. Food Quality and Preference, 8, 233239.
Swatland, H. J., Brooks, J. C., & Miller, M. F. (1998). Possibilities for
predicting taste and tenderness of broiled beef steaks using an opti-
cal-electromechanical probe. Meat Science, 50, 112.
Tornberg, E., Andersson, A., Goransson, A., & von Seth, G. (1993).
Water and fat distribution in pork in relation to sensory properties.
In E. Poulanne, & D. I. Demeyer, Pork quality: genetic and meta-
bolic factors. (pp. 239258). CAB International.
Tornberg, E., Wahlgren, M., Brndum, J., & Engelsen, S. B. (1999).
Instrumental analysis of pre-rigor conditions in beef at varying
temperature- and pH-falls. Submitted for publication.
Trout, G. R. (1988). Techniques for measuring water-binding capacity
in muscle foodsa review of methodology. Meat Science, 23, 235252.
Vold, R. L., Waugh, J. S., Klein, M. P., & Phelps, D. E. (1968).
Measurement of spin relaxation in complex systems. Journal of
Chemical Physics, 48, 38313832.