Prediction of Intramuscular Fat in Live Swine Using Real-time Ultrasound

Kevin Ragland, Research Assistant;

Jesper Brondum, Visiting Scientist; Tom Baas, Assistant Professor;

and Lauren Christian, Professor

Department of Animal Science

Iowa State University, Ames, IA


Enhancements in muscle quality as well as quantity of muscle are essential for improved consumer acceptance of pork. Specific factors influencing pork quality, therefore must be addressed.

The presence of marbling or lipids is said to be of importance to meat quality and sensory properties (Garcia et al., 1968; and Essen-Gustavsson et al., 1994). Numerous studies have suggested that marbling influences the palatability of cooked meat (Judge et al., 1959; Blumer, 1963; Field et al., 1966; Romans et al., 1969; and Cameron and Enser, 1991).

The trend in modern pork production has been toward producing leaner pork. Increased selection for lean has caused considerable concern regarding the possibility of a reduction of intramuscular fat in pigs (Schworer et al., 1995). Lower intramuscular fat levels would have detrimental effects on eating quality of pork (Barton-Gade, 1990).

Heritability estimates for intramuscular fat range from 0.49 to 0.81 (Duniec et al., 1961; Barton-Gade, 1990; and Schworer et al., 1995). The genetic correlation between intramuscular fat and meat content is low (rg = -0.30). A breeding program emphasizing simultaneous selection for higher levels of intramuscular fat and a higher meat content, therefore should be possible (Barton-Gade, 1990).

At present breeding programs for improving muscle quality require the collection of measurements on the carcass. Direct selection, therefore cannot be performed, and breeding programs for these traits must be based on progeny tests.

Real-time ultrasound offers an alternative to traditional swine breeding programs for muscle quality, because it allows noninvasive measurement of a trait on the live animal at a reasonable cost.

Ultrasound has been used to predict tenth - eleventh rib subcutaneous fat and loin muscle area of live swine with a correlation near .9 and .8, respectively (Moeller, 1994). Few studies have been conducted to evaluate the accuracy of ultrasound in predicting intramuscular fat. Dion et al. (1996) predicted marbling score with real-time ultrasonic cross-sectional and longitudinal scans. The accuracy of prediction was essentially zero.

The objective of this study, therefore, was to investigate the feasibility of predicting intramuscular fat in the longissimus muscle of live swine from a single real-time ultrasonic longitudinal image, and to assess the merit of using ultrasound predicted IMF values to classify animals into specific IMF groups.


Data utilized for this project (N=756) were collected as a part of the 1995 National Barrow Show® (NBS) Progeny Test (N=348) and the 1996 Livestock Producer Assistance Program (LPAP) test (N=408) conducted at the Northeast Iowa Swine Improvement Association Station located near New Hampton, Iowa. Animals represented in the NBS test were of the eight major breeds and crossbreds, while only crossbreds were included in the LPAP test.

Pigs were weighed and scanned off-test on an individual basis at weekly intervals upon reaching a weight > 240 lbs. Scanning was 24 hours prior to slaughter and accomplished with an ALOKA 500V (Corometrics Medical Systems, Wallingford, Connecticut) real-time ultrasonic machine fitted with a 12.5 cm, 3.5 MHz linear array transducer. Ultrasonic images were digitized on-site using a personal computer equipped with a frame-grabber board and controlling software. The images were stored as digitized files for later interpretation.

Two ultrasonic images were taken on each pig. A cross sectional image of the loin muscle and subcutaneous fat overlying the loin muscle on the right hand side of the pig at the tenth rib was acquired using a sound emitting transducer guide which fitted the natural contour of the pig's back. A longitudinal image was taken approximately 2.5 in. off the midline and parallel to the spine. The image included the 9th, 10th and 11th ribs.

Digitized cross-sectional images were interpreted using Quality Evaluation and Prediction (Iowa State University, Ames, Iowa), a computer software package developed specifically to measure linear distance and area of digitized images and matriculate these to a data file. BF10 was measured as the distance from the outer edge of the skin to the start of the fascia layer in the center of the longissimus muscle at a point approximately 2.5 in. lateral to the spine.

Longitudinal images were subjectively evaluated for image quality. Images with less than ideal image quality were removed from this analysis. Two 40*80 pixel regions of interest (ROI) were then selected from the image above the 10th and 11th ribs. Image-processing parameters were then determined for the selected ROI. Image-processing parameters were calculated using texture analysis. Texture parameters provide information about the image patterns generated in part by ultrasound scattering.

Upon completion of the test, pigs were transported to Hormel Co. in Austin, Minnesota, for carcass evaluation. Carcass measurements were taken by Iowa State University personnel following a 2-hour rapid chill. Standard carcass collection procedures were followed, as outlined in Procedures to Evaluate Market Hogs (NPPC, 1991, 3rd ed.). A sample was excised from the loin muscle at the 10th rib. The intramuscular lipids from the excised sample were measured according to the total lipid extraction method of Bligh and Dyer (1959).

Of the 756 images collected, 300 images had ideal image quality and were used for this analysis. Image variables, live animal measures including 10th rib backfat and sex of the animal, IMF values and marbling score were statistically analyzed to select a set of parameters for regression model development. Stepwise regression procedures were used for selection of variables to determine the prediction models.

Linear regression (PROC REG; SAS, 1985) was used to predict IMF and marbling score. Of the 300 images used, 200 were randomly chosen for model development and the remaining 100 were used for validation of the model.

Bias and mean absolute difference were calculated between predicted IMF and marbling score and actual IMF and marbling score. Regression equations containing eight ultrasound image variables, sex and BF10 were utilized to obtain predicted IMF and marbling score values. Bias was defined as:

The mean absolute difference was defined as:

= ( ô Chemical - Ultrasound ô )

Standard errors of prediction (SEP), widely considered as the standard measure of the ability of RTU to precisely evaluate differences between carcass and ultrasound measurements, were computed using the formula:

Mean absolute differences were sorted into one of five groups: 0-.25%, .26-.50%, .51-.75%, .76-1.0% and > 1.0%. The frequency of observations in each group was determined.

Actual IMF and predicted IMF values were grouped in one of three classes: < 1%, 1-2%, 2-3%, 3-4%, 4-5% and > 5%. The frequency of observations in each class was determined.


The pearson and spearman correlations between actual and predicted IMF were .70 and .69, respectively. The mean absolute difference was .66 between actual and predicted IMF and .50 between actual and predicted marbling score. The SEP was .83 for IMF and .61 for marbling score. Bias for IMF prediction was -0.005, while bias for prediction of marbling score was 0.004.

Figure 1 shows the classification of mean absolute differences by group. Over 77% of the mean absolute differences were less than 1%.

Figure 2 illustrates the grouping of actual and predicted IMF values. Over 70% of the predicted values were correctly classified. All misclassifications fell into an adjacent class.

Figure 1. Classification of mean absolute differences between actual and predicted intramuscular fat.

Figure 2. Classification of actual and predicted intramuscular fat values.


The results of this study indicate that objective measurement of IMF of the live animal with real-time ultrasound is feasible. This will allow the identification of breeding animals that are superior for IMF. Selection based on muscle quality measurements taken on the live animal will, therefore, be possible. The use of real-time technology to predict IMF may also be applicable for carcass evaluation. Image collection and procedures for image analysis must be further refined to enhance accuracy of IMF prediction.


Barton-Gade, P.A. 1990. Pork quality in genetic improvement programmes - the Danish experience. Proc. of the National Swine Improvement Federation Annual Meeting. Des Moines, IA.

Bligh, E.G. and W.J. Dyer. 1959. A rapid method for total lipid extraction and purification. Canadian Journal of Biochemistry and Physiology. 37:911.

Blumer, T.N. 1963. Relationship of marbling to the palatability of beef. Journal of Animal Science. 22:771.

Cameron, N.D. and M.B. Enser. 1991. Fatty acid composition of lipid in Longissimus Dorsi muscle of Duroc and British Landrace pigs and its relationship with eating quality. Meat Science. 29:295.

Dion, N., D. Pettigrew and J.P. Daigle. 1996. Efficiency of various live measurements for the prediction of lean yield and marbling. Proc. of the National Swine Improvement Federation Meeting. Ontario, Canada.

Duniec, H. 1961. Heritability of chemical fat content in the loin muscle of baconers. Animal Production. 3:195.

Essen-Gustavson, B., A. Karlsson, K. Lundstrom and A.C. Enfalt. 1994. Intramuscular fat and muscle fibre lipid contents in halothane-gene free pigs fed high or low protein diets and its relation to meat quality. Meat Science. 38:269.

Field, R.A., G.E. Nelms and C.O. Schoonover. 1966. Effect of age, marbling and sex on palatability of beef. Journal of Animal Science. 25:360.

Garcia, P.T., J.J. Casal, C. Olsen and G. Berra. 1968. A comparison of distribution and composition of intramuscular fat in Duroc Jersy and Hampshire pigs at 100 kg liveweight. Meat Science. 16:283.

Judge, M.D., V.R. Cahill, L.E. Kunkle and W.H. Brunner. 1959. Pork quality. I. Influences of some factors on pork characteristics. Journal of Animal Science. 18:448.

Moeller, S.J. 1994. Evaluation of genetic parameters for fat and muscle deposition in swine utilizing serial real-time ultrasound measurements. Ph.D. dissertation. Iowa State University, Ames, Iowa.

Romans, J.R., H.J. Tuma and W.L. Tucker. 1965. Influence of carcass maturity and marbling on the physical and chemical characteristics of beef. I. Palatability, fiber diameter, and proximate analysis. Journal of Animal Science. 24:681.

SAS Institute, Inc. 1985. SAS User's Guide: Statistics, 1985 edition. SAS Institute, Inc., Cary, NC.

Schworer, D.A., A. Rebsamen and D. Lorenz. 1995. Selection of intramuscular fat in Swiss pig breeds and the importance of fatty tissue quality. 2nd Dummerstorf Muscle Workshop on Growth and Meat Quality. Rostock.