In vivo studies on goose liver development by means of computer tomography
In vivo studies on goose liver development by means of computer tomography
László LOCSMÁNDI 0
Róbert ROMVÁRI 0
Ferenc BOGENFÜRST 0
András SZABÓ 0
Marcell MOLNÁR 0
Gabriella ANDRÁSSY-BAKA 0
Péter HORN 0
0 University of Kaposvár, Faculty of Animal Science , 7400 Kaposvár, Guba Sándor 40 , Hungary
- Commercial geese (Gray Landes) were examined by means of high-resolution spiral computer tomography in order to follow their liver development in vivo. Three ganders were scanned six times before, within and after a force-feeding period. 3D images of the liver were reconstructed from the 2D transverse slices with segmentation and rendering methods. The changes in the external surface, the volume of the liver and also the characteristic Hounsfield values were determined. The livers of another 70 ganders were examined by CT, then by direct chemical analysis (days 0, 13, 18, 19, 20 and 21 of force-feeding). To estimate the chemical composition of the tissue, prediction equations were developed based on the pixel frequency distributions. With partial least squares (PLS) regression, the ether extract and crude protein content could be estimated with R2 = 0.97 and R2 = 0.96 accuracy, respectively. Data analysis was complemented with serial blood serum measurements characteristic of liver steatosis. The method applied may be a unique possibility to study the real geometrical relations of liver development and also to describe the qualitative changes of tissue composition during the force-feeding period in vivo, with special regards to selection purposes.
« in vivo » donne la possibilité unique de caractériser les corrélations géométriques réelles au cours
du développement du foie et de décrire les modifications qualitatives dans la composition du tissu
pendant l’engraissement par gavage. Cette méthode pourrait être utilisée en sélection.
foie gras / oie / tomographie / rendement / composition
One of the main objectives of goose
breeders is to maximize the accessible liver
volume during the force-feeding period and
also to keep its fat content at an optimum
level for the processing industry and
consumers. Under natural conditions some
degree of hepatic steatosis occurs in wild
waterfowl, as a consequence of energy
storage before migration. In poultry production,
this specific capacity is used for the
production of commercial fatty liver. The excess
of triglycerides is normally stored in the
cytoplasmic storage vesicles of the liver.
When overproduction of triglycerides occurs,
which is the case during force-feeding, the
liver responds in two ways: triglycerides are
secreted into plasma as VLDL increases
and since force-feeding does not allow the
birds to fast, the liver continues to
accumulate triglycerides [
]. It is well known that
geese breeds differ in their susceptibility to
liver steatosis, considering that the response
to force-feeding is partly under genetic
]. According to Rouvier et al. [
the direct genetic effects due to autosomal
and sex linked genes were high and positive
for fatty liver weight in selected strains of
geese. It is notable that the liver weight
could be increased by selection without a
great effect on “paletot” weight as Larzul
et al. [
] established. The Landes goose –
used in the present study – is among the best
in response to overfeeding as described by
Mourot et al. [
]. To a certain extent the
high susceptibility of the breed is
explainable by the high activity of malic enzyme
and also by the fact that hepatic lipogenesis
remains very active until the end of the
overfeeding period in order to improve the
quality and quantity of the liver.
At present our experiments are partly
focused on the preliminary step
(preparation) of force-feeding and the development
of new, non invasive, in vivo, CT imaging
based selection methods for liver yield [
With special regards to the expectable
severity in fatty liver production, the
improvement of the so-called force-feeding free
methods seems to be reasonable in the near
future. These developing methods are based
on voluntary feed intake that is improved by
the limited time to access the feed.
Conventionally, the lipid content of the
liver can be measured by direct chemical
analysis. According to Guy [
], the total
lipid content of the fatty goose liver is around
50–55%. Storage lipids are predominant,
with 95% triglycerides and 1–2%
cholesteryl esters. Structural (membrane) lipids,
such as phospholipids and free cholesterol,
account for only 1–2 and <1%, respectively
]. Force-feeding induces a large
hypertrophy of hepatic cells in relation to the
accumulation of triglycerides. Reaching high
liver weights, geese provide good
technological liver quality with a fat loss limited
to 13.9% during autoclaving, as established
by Guy [
In studies with chickens [
] and turkeys
] it has been shown that computer
tomography (CT) is a suitable non-invasive
technique to measure the volume or mass of
the pectoralis muscle and abdominal fat
permitting single or repetitive measurements.
Romvári et al.  published a new in vivo
3D method to estimate the volume or yield
and the geometrical conformation of the
breast muscle of broiler chickens by
In the present study different in vivo CT
methods were applied to follow the liver
morphologic development of geese and to
analyze the changes in the liver
composition during the force-feeding period.
2. MATERIALS AND METHODS
In the first methodological experiment
commercial type Landes ganders were
examined at 11, 15, 16, 17, 18, 20 weeks of
age, four weeks before (period 1), within
(periods 2–5) and two weeks after (6) the
force-feeding period. The three repeatedly
examined animals (A, B, C) were selected
from a larger population representing the
average weight at a given age. In the second
trial altogether 70 ganders were scanned at
the end of the preparation period, at 13, 18,
19, 20, 21 and 22 days of the force-feeding
period. The rearing conditions and the
nutrition procedure corresponded to the common
intensive management technology widely
practiced in Hungary. A commercial pelleted
goose feed (3.01 g EE (ether extract)·100 g–1
DM, 21.72 g CP·100 g–1 DM, 6.09 g C
Ash·100 g–1 DM, ME: 11.2 MJ·kg–1) was
fed ad libitum. The force-feeding diet
contained 4.91 g EE·100 g–1 DM, 12.78 g
CP·100 g–1 DM, 2.33 g C Ash·100 g–1 DM
and 12.4 MJ·kg–1 ME. A so-called
preparation period was started twenty-eight days
before force-feeding. During that interval,
the total daily feeding time was decreased
continuously. At the end of the preparation
period, birds only had access to feed for
30 minutes, two times per day. Later on, as
a result of this technological process, the
geese were able to consume the sufficient
amount of feed for fatty liver production.
2.2. CT procedure
The ganders were scanned in vivo by
means of a Siemens Somatom Plus S40
spiral CT scanner at the Institute of Diagnostic
and Oncological Radiation of the Kaposvár
University. The high resolution CT scans
were taken from the geese using a zoom
factor of 3.4. During the examination, the
animals were fixed with belts in a plastic cradle
without using anaesthetics (Fig. 1). In the
methodological experiment depending on
the size of the birds, 25 to 40 pictures – with
five mm slice thickness covering the whole
region of the liver-were acquired.
The picture-forming pixels (512 × 512 in
a single slice, 0.98 × 0.98 mm area of each)
are in fact small prisms with a definite
volume (pixel area multiplied by the slice
thickness, i.e. 5 or 10 mm) and are called voxels.
Each pixel or voxel is characterized by a
defined X-ray density value (expressed in
Hounsfield Units). We are able, therefore,
to determine the part of the total volume of
the examined scan that falls into the
Hounsfield (HU) interval of interest [
picture evaluation methods were applied. As a
first approach the liver surface and volume
data were determined from the series of
cross sectional images. In parallel to this,
the mean and also the most frequent
Hounsfield values of the liver tissue were
measured. In addition, a special imaging process
was applied to handle the 2D slices, and a
segmentation method was used to generate
the boundary data. Finally surface
rendering was undertaken to obtain a real 3D view
of the liver.
In the second experiment, the serial
scans of the liver were taken with 10 mm
thickness. The post processing of the images
was based on the frequency distribution of
the voxels. In the present work, the extreme
density values were excluded and only those
corresponding to the liver were retained, i.e.
the range from –80 to +100 on the
Hounsfield scale (water = 0).
2.3. Blood samples and chemical analysis
In the second experiment at every
scanning event, blood samples were drawn from
v. brachialis. Blood was allowed to clot at
room temperature and serum was
centrifuged on 5500 rpm for 10 minutes. Serum
metabolite (triglyceride, total and HDL
cholesterol, inorganic phosphorus, total
protein, albumin, uric acid, total bilirubin)
concentrations and enzyme activity (lactate
dehydrogenase, γ-glutamyl transpeptidase)
values were determined on a Konelab 20i
automatic equipment, using Konelab reagent
kits. Immediately after the CT examination,
experimental slaughter was performed after
electrical stunning; EE and CP content of
the liver were determined by direct
chemical analysis [
2.4. Data processing and statistical methods
Based on the similarity of the X-ray
frequency distribution diagrams with the
absorption spectra [
] Partial Least Squares
(PLS) regression was applied to diminish
the multi-co-linearity of the neighboring
HU values (HUv) [
]. The principal
components (PC) were calculated from the HU
variables of the liver tissue. Prediction
equations were developed by linear regression
from the previously calculated principal
components or factors.
3. RESULTS AND DISCUSSION
3.1. Methodological experiment
There was no negative effect of the
repeated CT scanning on liver
development, since the final liver volume at the end
of force-feeding (period 5) was similar to
the rest of the geese from the original stock
in the methodological experiment.
Substantial changes were observed in the liver
volume and surface area within the examined
period (Fig. 2).
At the time of preparation (period 1) and
the beginning of force-feeding (period 2),
the measured values were similar. At the
end of the 21-day period (5), the volume
data became three times higher compared to
the starting point. Two weeks later, as a
consequence of feed withdrawal, the birds B
and C reached their starting weight. A
similar observation was made by Prehn [
monitoring the fate of birds, which returned
to basic conditions within approximately
four weeks after reaching the terminal stage
of force-feeding. After the early rapid
increase (period 3), the A bird showed a
fallback in liver tissue development (period 5)
probably caused by its individual
susceptibility to force-feeding, which resulted in
fatal gastro enteritis.
Quantitative changes of the liver were
demonstrated by creating index values from
the surface and the volume data. Index
values (liver surface / liver volume) give
information about the liver surface falling on one
liver volume unit. At the time of preparation
(period 1) and the beginning of
force-feeding (period 2) the estimated values were
similar. Fast cross development of the liver
of the B and the C bird was observed from
the beginning of the force-feeding period to
the end of the 21st day (2–5 period). Index
values were about 1.4 and 0.7 at periods
2 and 5, respectively. The liver substance
became more compact. The birds marked B
and C were investigated after the
force-feeding period (5–6 period) for the regenerative
capacity of the pathologically developed
liver. According to the values (1.43 and
1.41), the liver substance restored its normal
physiological condition within two weeks.
The 3D reconstruction in Figure 3 shows
the real in vivo anatomic characteristics of
the lobes of a liver (B bird) from a lateral
view. This is not examinable by means of
slaughtering, caused by the loss of the
original geometrical conformation of the organ.
It can be remarked that the geometry of the
post preparation (period 2) and post
forcefed (period 6) states are very similar proving
the reversibility of the process. The large
proportional changes of the liver tissue within
the force-feeding period (3, 4, 5) are also
perceptible in the figure.
In addition to the quantitative and
morphological evaluations, a certain qualitative
analysis of the liver was also performed.
The goose liver tissue has a characteristic
(the most frequent HU value) X-ray density
value (around 80 HU) in its normal,
physiological status. Substantial changes can be
seen in Figure 4, in the HU values
throughout the period.
Similarly with the geometrical and
volumetric changes the measured density values
sensitively follow the force-feeding process.
The characteristic HU values within the
period of preparation (1–2) and post
forcefeeding (6) are nearly similar (55–80 HU).
In the course of force-feeding (2, 3, 4 and
5), the density values were 80, 80, –20 and
–50 HU respectively. The latter value
approximates the typical fat tissue density,
referring to the high fat content of the fatty liver.
In certain cases fat deposition becomes
irreversible together with the depreciation of
the product. According to Bogin et al. [
(cited in SCAHAW [
]), if force-feeding
is continued after three to four days, the
level of cell damage rises significantly. In
Hungary, 6–10% of processed fatty livers
show the so-called extreme fatty liver
3.2. Chemical content determination
Data of liver weight and the results of
chemical analysis of the livers are
summarized in Table I. It is clearly visible that after
day 18, practically no developmental changes
can be measured in the characteristic data
measured. Dominant compositional changes
of the liver occurred mainly before this
time, showing a likewise saturation process.
Different models were used to
characterize the relation between the liver weight and
the chemically determined ether extract
content. The highest correlation (R2 = 0.91)
resulted with the “S” curve fitting
procedure (Fig. 5). It can be remarked that the fat
(g·100 g–1 DM)
content remained constant above the liver
weight of 350 g.
For the better understanding of the applied
estimation method some basic relations can
be seen in Figure 6, where the curve of the
correlation coefficients between HU
variables and fat content values show a high,
blunt peak in the interval corresponding to
fat tissue. The second, slightly higher
positive peak is located in the interval of muscle
explaining the high accuracy of prediction of
the chemical content demonstrated hereafter.
First the fat content of the liver was
characterized with the most frequent HUv of the
tissue, based on the relation described in the
methodological part. Figure 7 shows a strong
linear relation between the most frequent
voxel density value and the fat content (R2 =
A slightly weaker correlation (R2 = 0.97)
was found between the most frequent voxel
density value and the crude protein content.
The result of the estimation of chemical
composition (R2 = 0.9604 and 0.9216, EE
and CP respectively) based on the average
HUv of the liver tissue was very similar to
that of the latter.
The use of prediction equations in CT
based experiments aimed at predicting the
chemical content of the total body is
general. In previous studies on rabbits, fish,
broiler chickens and turkeys [
2, 3, 17, 18
variables of the prediction equations
originated from the summation of the
neighboring 10 – 10 density values of the Hounsfield
intervals between –200 and +200,
corresponding to the fat and muscle tissue. These
values varied between –80 and 100 without
any reduction in the investigated ganders.
The reliability of the prediction of ether
extract and/or crude protein may be decreased
by the strong linear connection between the
neighboring HU variables. With the applied
PLS regression this effect has been
diminished, by finding latent variables that explain
most of the variability (Tab. II). In general,
the closer the validation variance is to the
calibration variance, the more reliable are
the model conclusions. It can be seen that
in both cases (CF and CP) above PC7
(principal component), the validation variances
are decreasing, and thus the so-called noise
level has been reached.
Figures 8 and 9 show the correlations
between the measured and estimated liver
fat and protein content (calculated for dry
matter) on the basis of voxel density data.
The MGLH equations were developed from
PC with a stepwise procedure of the
multivariate linear regression procedure. The
basic data of the two prediction equations
are presented in Table II.
The validation of the above mentioned
predictions resulted in a SEC (standard
error of calibration) value of 5.64 and 6.73
for EE and CP, respectively. In the authors’
experience in CT imaging, this was the first
case when the approach based on the most
frequent- and average HUv resulted in a
similar correlation with the chemical
content based on the use of prediction
equations. The liver tissue with its homogenous
texture seems to be an ideal object for CT
based analysis, since any changes in the
tissue composition coincide with considerable
changes in its X-ray density value.
While the liver substance is relatively
constant, a few CT scans seem to be
sufficient to grade the fatty liver. The large
variation in the characteristic HU values
corresponding to the liver tissue of the different
ganders encourages the development of this
in vivo method for selection purposes. It
seems worth screening the top breeder
candidates at least in order to get one’s own
performance data on liver development.
3.3. Blood serum parameters
Making serum measurements had a
double goal. The first was to characterize the
metabolic and physiological status of the
birds along the feeding regime, while the
second goal was to determine possible
relationships among in vivo determined liver
characteristics and blood serum data.
In blood lipids, triglyceride, total
cholesterol and HDL cholesterol were determined.
When plotted against the EE content of the
liver, all the above mentioned parameters
showed significant correlations (r = 0.681,
0.792 and 0.824, respectively, P < 0.05 in
Serum cholesterol concentration was
found to be highly indicative of the
physiological status of the birds. In general, when
sorting the birds according to their serum
cholesterol concentration within the every
day-category (i.e. slaughtered at the 18th,
19th, 20th, 21st or 22nd day of
force-feeding), the birds having the highest total
cholesterol values showed serum total protein,
triglyceride, uric acid, total bilirubin and
inorganic phosphorus concentrations, as well
as lactate dehydrogenase and γ-glutamyl
transpeptidase activities highly exceeding
the within group average. Furthermore, it
was found that birds with high metabolic
and enzymatic values did not show
averageexceeding liver weight values. Moreover,
these birds were characterized by liver EE
content values below the group mean. It was
thus supposed that geese with markedly
high serum lipid values are most likely
nonresponders to force-feeding, since lower
liver lipid accumulation was paralleled, in
these cases, to strong lipid traffic in the
Between total and HDL cholesterol, a
strong (r = 0.897, P < 0.001) correlation was
found (Fig. 10). However, no correlation
was found between liver weight and serum
parameters as well as serum enzymes,
which is consistent with the findings of
Davail et al. [
], in Landes geese.
The applied in vivo CT examination
method seems to be suitable to follow liver
development and to analyze the liver
composition of geese during the force-feeding
period. The procedure is non-invasive and
could be a powerful tool for the
development of new liver producing technologies
by genetic selection with or without
The financial support of the National Research
Fund (No. NKFP4/034) is gratefully
 Association of Official Analytical Chemists, Official Methods of Analysis , Washington, DC, AOAC, 1980 , pp. 69 - 82 .
 Andrássy-Baka G. , Romvári R. , Milisits G. , Sütö Z. , Szabó A. , Locsmándi L. , Horn P. , Non-invasive body composition measurement of broiler chickens between 4-18 weeks of age by computer tomography , Arch. Tierz . 46 ( 2003 ) 585 - 595 .
 Andrássy-Baka G. , Romvári R. , Sütö Z. , Szabó A. , Horn P. , Comparative study of the body competition of different turkey genotypes by means of CT, Arch . Tierz. 46 ( 2003 ) 285 - 292 .
 Bogenfürst F. , Handbook of the goose breeders (in Hungarian) , Új Nap Lap és Könyvkiadó , Budapest, 1992 .
 Bogin E. , Avidar Y. , Merom M. , Israeli B. , Malkinson M. , Soback S. , Kulder Y. , Biochemical changes associated with fatty liver in geese , Avian Pathol . 13 ( 1984 ) 683 - 701 .
 Brenoe U.T. , Kolstad K. , Body composition and development measured repeatedly by computer tomography during growth in two types of turkeys , Poultry Sci . 79 ( 2000 ) 546 - 552 .
 Dobrowolski A. , Branscheid W. , Romvari R. , Horn P. , Allen P. , X-ray computed tomography as possible reference for the pig carcass evaluation , Fleischwirtschaft 84 ( 2004 ) 109 - 112 .
 Davail S. , Guy G. , Andre J. , Hermier D. , HooParis R., Metabolism in two breeds of geese with moderate or large overfeeding induced liver-steatosis, Comp . Biochem. Physiol. A 126 ( 2000 ) 91 - 99 .
 Fournier E. , Peresson R. , Guy G. , Hermier D. , Relationships between storage and secretion of hepatic lipids in two breeds of geese with different susceptibility to liver steatosis , Poultry Sci . 76 ( 1997 ) 599 - 607 .
 Guy G. , Physiological terms of quality liver production , Proceeding of the 3rd International Poultry Breeding Symposium , Kaposvár, Hungary, 2000 , pp. 51 - 67 .
 Hermier D. , Saadoun A. , Salichon M.R. , Sellier N. , Rousselot-Paillet D. , Chapman M.J. , Plasma lipoproteins and liver lipids in two breeds of geese with different susceptibility to hepatic steatosis: changes induced by development and force-feeding , Lipids 5 ( 1991 ) 331 - 339 .
 Hounsfield G.N. , Computed medical imaging, Nobel Lecture , 1979 Dec. 8, J. Comput . Assist. Tomo. 4 ( 1980 ) 665 - 674 .
 Larzul C. , Rouvier R. , Rousselot-Pailley D. , Guy G. , Estimation of genetic parameters for growth, carcass and overfeeding traits in a white geese strain , Genet. Sel. Evol . 4 ( 2000 ) 415 - 427 .
 Lorber A. , Wangen L.E. , Kowalski B.R. , A theoretical foundation for the PLS algorithm , J. Chemometr . 1 ( 1987 ) 19 - 31 .
 Mourot J. , Guy G. , Lagarrigue S. , Peiniau P. , Hermier D. , Role of hepatic lipogenesis in the susceptibility to fatty liver in the goose (Anser anser) , Comp. Biochem. Physiol. B Biochem. Mol. 126 ( 2000 ) 81 - 87 .
 Prehn D. , Contribution à l'étude du bien-être du canard (Cairina moschata × Anas platyrhynchos) en gavage : études biométriques, cliniques, histologiques et fonctionnelles (1996), in: Welfare Aspects of the Production of Foie Gras in Ducks and Geese , Report of the Scientific Committee on Animal Health and Animal Welfare , 1998 .
 Romvári R. , Szendrö Zs ., Jensen J.F. , Sorensen P. , Milisits G. , Bogner P. , Horn P. , Csapó J. , Noninvasive measurement of body composition of two rabbit populations between 6-16 week of age by computer tomography , J. Anim. Breed. Genet . 115 ( 1998 ) 383 - 395 .
 Romvári R. , Hancz Cs., Petrási Zs ., Molnár T. , Horn P. , Non-invasive measurement of fillet composition of four freshwater fish species by computer tomography , Aquacult. Int . 10 ( 2002 ) 231 - 240 .
 Rouvier R. , Poujardieu B. , Rousselot-Pailley D. , Larrue P. , Esteve D. , Genetic effects on growth, force-feeding and fatty liver traits in crosses of two selected strains of geese (Anser anser) , Genet. Select. Evol . 1 ( 1992 ) 53 - 69 .
 SCAHAW ( The Scientific Committee on Animal Health and Animal Welfare) , Report of the Scientific Committee on Animal Health and Animal Welfare on Welfare Aspects of the Production of Foie Gras in Ducks and Geese , 1998 [Online] Available: http://europa.eu.int/ comm/food/fs/sc/scah/out17_en. html [1 Aug . 2004 ].