The mirror carp (
In order to predict the chemical fillet fat content, thickness measurements
and volume calculations of the back fat were carried out using CT. Compared to
the CT-based back-fat thickness measurement correlated with the results from
the chemical analysis (
The evaluation of the fillet yield resulted in a mean value of 42.89 %
with a standard deviation of
Aquaculture will play a major role in feeding the growing world population, which will mean
feeding nearly 10 billion people by 2050 – in this context aquaculture is of
particular interest when it comes to solving the upcoming global protein
deficiency (Evans, 2009; Searchinger et al., 2018). Worldwide, cyprinids are
the most important and largest fish family. The common carp (
Different imaging technologies such as ultrasound (US), magnetic resonance
imaging or computed tomography are used for performance testing in live
animals (Scholz et al., 2015). Computed tomography (CT) has already been
successfully used for determining carcass composition in fish species such
as salmon, rainbow trout, cod, common carp, grass carp and silver carp
(Gjerde, 1987; Rye, 1991; Romvári et al., 2002; Hancz et al., 2003b; Kolstad
et al., 2004, 2008). This imaging technology is based on the
density-related attenuation of x-rays by different tissues. The object to be
examined is positioned on a table and moved stepwise through the gantry of
the CT device. A measuring unit consisting of an x-ray tube and opposing
detectors rotates along the gantry. During rotation, the object is
irradiated and the remaining radiation after passing the object is detected.
From these measurements, a specific attenuation value is calculated for each
volume element (voxel). The attenuation is expressed in Hounsfield units
(HU) and can range from
Fat content and fillet yield are the main traits in the marketing of carp. Previous studies investigated the morphology (Cibert et al., 1999) and heritability estimates of the fillet yield (heritability of 0.38 for % fillet yield with skin; Kocour et al., 2007). Regarding the fat content, a high variation has been described in common carp: Zeitler et al. (1984) 6.7 %–17.6 %, Ljubojević et al. (2013) 6.3 %–15 %, Bauer and Schlott (2009) 2.7 % to 6.9 %.
Oberle and Aas (2015) described negative effects on the taste of carp flesh
exceeding a fat content of 15 %. In Germany, some areas are allowed to
produce carp under a European Quality Scheme (EU Regulation no. 1151/2012
article 7, 2012). According to the requirements laid down in the
product's specifications, the fat content including the skin
should not exceed 10 %. Due to the fact that carps are mostly traded
alive, fast and noninvasive methods are needed to determine carcass traits.
Currently the fat content is determined by a microwave-based Fish Fatmeter
(Kent, 1990). Oberle et al. (2015) reported a coefficient of determination of
The measurement of back-fat thickness with an ultrasound is widely used for other species, such as pigs and cattle, as a noninvasive method to determine the carcass quality (Brethour, 1992; Newcom et al., 2002; Müller and Polten, 2004). It is also possible to estimate various parameters in animals using linear measurements. In cattle, linear measurements are successfully used to predict body weight (Alderson, 1999; Ozkaya and Bozkurt, 2009).
The objective of the present study was to estimate the body composition and fillet yield of carps using computed tomography and different linear body measurements. Chemical analysis and dissection were used as reference methods. In addition, a validation of the back-fat thickness measurement by ultrasound was done.
During autumn 2014, 33 three-year-old mirror carps (
The Bavarian government was informed about this study. A notification as an animal experiment was not necessary, as the handling of the animals did not differ from the routine handling of the fish farmers. The handling of the fish was carried out with special care and without prolonged exposure to the outside air.
The non-sedated fish were put into narrow water-filled containers, and oxygen
was added to the water. Four fish fit into one container at the same time,
separated by thin walls. The carps were examined for their back-fat thickness
using a mobile ultrasound device (MicroMaxx, Fujifilm SonoSite, Frankfurt am
Main, Germany) and a 5
Linear measurement of carp including height, four lengths and four circumferences.
After slaughtering, the fish were chilled immediately using ice-cube-filled
boxes. Within 6
The CT images had a matrix of
In a next step, the fillet thickness was measured on single cross-sectional
CT images at the level of CT4 in order to predict the fillet yield. At an
angle of 90
Measurement of the back-fat thickness in four positions using
transversal CT images; 2.4, 3.0 and 3.6
CT cross section 2.1
After the CT scans were completed, the carps were filleted. The in vivo
examinations, the slaughtering, the CT examinations and the filleting of the
fish took place in 1 d. The fillets including the skin were weighed and
fillet yield was calculated. The sex was determined by adspection at the
opened carcass. Afterwards, the fillets were frozen at
Data were collected using Excel (version 16.15, Microsoft Corporation,
Redmond, USA) and analyzed using RStudio (Integrated Development for R –
RStudio, Inc., Boston, MA, USA) and MATLAB (MATLAB and Statistics Toolbox
Release 2012b, The MathWorks, Inc., Natick, MA, USA). A Kruskal–Wallis test
was performed to evaluate the difference in fat content between the ponds.
The Mann–Whitney
Data for body weight, fillet fat content, fillet weight and fillet yield are
summarized in Table 1. The fillet weight represents
the sum of the left and right fillet weight. The fillet yield was calculated
as a percentage of the fillet weight of the total body weight. The fillet
fat content was determined by chemical analysis of the left fillet. The
fillet fat content showed a very high range, which can be attributed to the
individual ponds. The average values of the fillet fat content for the carps
from individual ponds were 5.2 % (
Both sexes were approximately equally represented (15 females, 17 males). The sex of one fish could not be determined. No significant differences between the sexes regarding the fillet fat content, the fillet weight and the fillet yield were found.
Mean, minimum, maximum and standard deviation (
An uncertainty test of all following linear models was performed by testing the residuals using the diagnostic plots described in Sect. 2.6. The residuals showed no nonlinear patterns and were normally distributed. No major deviation was found regarding the assumption of equal variance of the residuals. One influential case was found regarding the multiple linear regression models for fillet fat content, CT-based back-fat thickness CT4_BF1 and linear measurements. However, the regression results without the influential case were only marginally better. Therefore, the fish causing the influential case was not removed in favor of the number of observations.
The back-fat layer and its influence on the carcass quality, i.e., the fillet fat content, were investigated using CT technology. In a first step, the relationship between the back-fat thickness measured on CT images and the fillet fat content determined by chemical analysis was analyzed. The results are presented in Table 2. The linear correlation of CT4_BF2 and the fillet fat content determined by chemical analysis is shown in Fig. 4.
Mean and standard deviation (
RMSE: root mean square error, dependent variable: fillet fat content.
Correlation between fillet fat content determined by chemical
analysis (%) and back-fat thickness CT4_BF2 (cm) determined
by CT image (
Single linear correlations of the analogous measuring positions of CT and US are shown in Table 3. In addition, a Bland–Altman analysis was performed to compare these two measurement methods (Fig. 5).
Relationship between back-fat thickness measured on a single CT image
and back-fat thickness measured by ultrasound (US) (
Bland–Altman analysis comparing two methods for measuring back-fat
thickness in carp: the ultrasound method (US) and the CT-based method
(CT_BF);
Next, the influence of the linear measurement was investigated. Single correlations of the results of fillet fat content determined by chemical analysis and the linear measurements are shown in Table 4.
Relationship between fillet fat content (%) measured by chemical
analysis and linear measurements;
In a next step, multiple regressions were performed using the results of the chemical analysis as dependent variables. As independent variables, the CT back-fat thicknesses were used and different linear measurements were added. The prediction of the fillet fat content was significantly improved with the addition of linear measurements. If more than one linear measurement was added, the prediction could not be improved decisively. The results of multiple regression analysis are shown in Table 5.
Results of multiple linear regression for fillet fat content (%)
and CT-based back-fat thickness (cm) with the addition of linear measurement
(
Furthermore, the volume of the back fat (
3-D model of a mirror carp created with 3D-Doctor software showing the back fat from the beginning of the back to the beginning of the dorsal fin.
Results of multiple linear regression for fillet fat content (%)
and back-fat volume (
The fillet yield was calculated from the fillet weight and the total body
weight of the fish. Mean fillet yield was 42.89 % (
The fillet thickness measurement was performed on single CT images
(Fig. 3). The mean of CT4_R was
2.74
In order to predict fillet yield, linear regression studies were done. In a first step, the total body weight and the linear measurements such as lengths, height and circumferences were used to predict the fillet yield. In a next step, linear regression studies were done for fillet yield and the CT measurement (Table 7). Multiple regression models of the CT measurement combined with linear measurements did not result in stronger predictions.
Results of single linear regression between fillet yield (%),
linear measurement and CT measurement;
The study aimed at the prediction of the body composition of mirror carps
using CT and linear measurements. A sample size of
However, carcass composition is not only determined by environmental factors; it also depends on the breed (Gela et al., 2003; Varga et al., 2013). Genetic improvement in common carp was investigated by several authors (Bakos and Gorda, 1995; Linhart et al., 2002; Kocour et al., 2005, 2007). The best results for carcass quality were achieved by cross-breeding. However, before genetic selection can be performed, a precise phenotyping must be carried out.
The fillet fat content of the examined carps ranged between 2.41 % and 26.60 %, measured by chemical analysis. In addition, the fat content of the carps varied considerably between the different ponds the fish originated from, which is mainly related to feeding, but also to other factors, e.g., stocking density and water temperature (Zeitler et al., 1984; Yamamoto et al., 2003; Schwarz et al., 2006).
The back-fat layer is known to be a good predictor of the fat content of carp
(Oberle et al., 2015). It is of constant thickness in the range 2.4 to
3.6
In the area of the dorsal fin the back fat is split. The spinous process of the spinal column protrudes from below into the fat layer. At its upper end a muscular tissue divides the fat layer into two sections. This separation made it difficult to measure the thickness of the back-fat layer accurately in the area of the dorsal fin using in vivo methods such as ultrasound.
The CT-based back-fat thickness measurement and the US measurement showed
higher correlation in the range of 2.4 to 3.6
Using single linear regression models to evaluate the correlation between
fillet fat content and linear measurement, moderate, nonsignificant results
were obtained. The RMSE ranged between 6.44 % and 7.11 %. Remarkably,
the best result was obtained by predicting the fat content using Length 1;
the Pearson correlation coefficient was negative (
The back-fat thickness measurement using CT images showed a correlation of
The volume calculation of the back fat provides better results with regard to the fillet fat content than the thickness measurement. The evaluated area ranged from the beginning of the back to the beginning of the dorsal fin. The prediction of the fillet fat content using the back-fat volume resulted in a Pearson coefficient of correlation of 0.92 (RMSE of 2.75 %). A Pearson coefficient of correlation of 0.94–0.96 was achieved using the chemical analysis results as a dependent variable and including back-fat volume and linear measurement.
Hancz et al. (2003a) achieved a similar result in calculating the back-fat area
(
Usually, carp is marketed as whole fish. The marketing of fillets as part of a new food trend is becoming increasingly important. Therefore, besides the fillet fat content, the fillet yield (%) is an important parameter in the marketing of carp. Compared to other studies, the fillet yield of 42.89 % found in our sample was very good. Fillet yields including the skin were examined in Austria with 34 %–35.9 % (Bauer and Schlott, 2009), in France with 34.6 % (Cibert et al., 1999) and in the Czech Republic with 41.1 % (Kocour et al., 2007).
Linear regressions between fillet yield and linear measurements resulted in
a Pearson coefficient of correlation of 0.33–0.68 and RSME of 1.78 %–2.29 %. The best correlation was achieved between fillet yield and the
circumference of the fish on the level of the pectoral fin (CF1). With the
exception of Length 1 and Circumference 3, all linear measurements achieved
a significant correlation with the fillet yield (
Linear regressions between fillet yield and CT measurements of fillet
thickness resulted in a Pearson coefficient of correlation of around 0.67
with RMSE of 1.79 % to 1.85 %. Kolstad et al. (2004) found correlations of
0.53 to 0.95 by calculating the area (%) of lean tissue in different CT scan
positions in Atlantic halibut (
In summary, it can be concluded that a volume calculation of the back fat based on three-dimensional CT images provides a more accurate prediction with regard to the fillet fat content than two-dimensional measurements of the back fat using single CT images or ultrasound.
The back fat in carp has proven to be a significant area regarding the fillet fat content and therefore the carcass quality. Multiple linear regression models including linear measurements can be used in both 2-D and 3-D measurements to improve the Pearson coefficient of correlation.
The fillet yield can be predicted with moderate results by measuring the thickness of the fillet on single transversal CT images. A prediction on a similar level is provided by some selected linear measurements and by total body weight.
In principle, CT technology, combined with linear measurements, offers great potential for phenotyping carp. In addition, the CT images can be used in the long term to evaluate further parameters for predicting body composition in mirror carp.
The results of our study should be verified with a larger number of animals. Next, the best predictive model could be established for in vivo measurements and will help to select suitable fish for breeding. In this way, a system of quality-oriented production can be established that leads to a high-quality product and thus to a high level of consumer acceptance.
The original data are available upon request to the corresponding authors.
PVKR conceived and designed the research idea and supervised the study; PM, BG and PK performed the experiments; MO provided the experimental environment and coordination; MJ performed the CT examinations; PM and BG analyzed the data; PM wrote the paper.
The authors declare that they have no conflict of interest.
This research has been supported by COST (European Cooperation in Science and Technology) (Farm Animal Imaging (FAIM) (grant no. FA1102)).
This paper was edited by Steffen Maak and reviewed by two anonymous referees.