Pannon Agricultural University, Faculty of Animal Science, 7401 Kaposvŕr, Hungary
Just one hundred years ago. Wilhelm Conrad Rontgen published his paper in Wurzburg describing X-rays and their effects, earning him the first Nobel Prize in Physics, presented in 1901.
X-ray based diagnostic equipment revolutionized human diagnostics.
A new milestone in the development of radiology was when Godfrey Hotmufield proposed to get additional anatomical information from a cross-sectional plane of the body by computer aided mathematical synthesis of an image from X-ray transmission data obtained from many different angles through the plane in consideration. The idea of the X-ray Computer Aided Tomography (CAT) was born, and the dramatic development of X-ray CAT imaging technology began, enhancing efficiency and scope of human diagnostics in human medicine.
The numbers used to characterize the X-ray absorption in each picture element (pixel) of the CAT image are called "CT" or Hounsfield numbers (Hounsfield, 1979). The Nobel Prize was awarded to Hounsfield in 1979.
X-ray Computer Aided Tomography system and software used were developed for humen medical purposes - mainly to detect anatomical physiological disorders - by all leading high-tech manufacturers in the world.
As early as 1980, the potential of the new technique to be utilized in animal science was first recognized in Norway (Skjervold et al., 1981). In 1982, the Agricultural University of Norway acquired a Siemens Somatom 2 CAT system, and pioneering work started to apply X-ray CAT techniques in animal science. Based on the promising results obtained and published by the "Norwegian school" (Sehested, 1 984 Vangen, 1984 Allen and Vangen, 1984: Vangen and Standal, 1984), a research project proposal was prepared for the Hungarian Ministry of Agriculture and Food and the World Bank in 1985 to set up a digital imaging center at our Institute. The project was approved in 1986 (Horn, 1991a).
The new digital imaging and diagnostic center started its operation in 1990, equipped first with a Siemens Somatom DR type X-ray CAT and later with MRI and Spiral X-ray CAT and ultrasound scanner systems to serve both animal science, and human diagnostics and research.
THE CAT TECHNIQUE
CAT scanning is providing information on anatomical and tissue composition from a cross- sectional plane of the body. Most techniques employ an X-ray tube positioned opposite an array of a large number of detectors (512). Both rotate around the "patient," and X-ray projections are emitted either at full- or half-degree intervals. With each projection the detectors measure X-ray absorption by the patient. The computers compute the detector information to produce there constructed cross-sectional image of the subject. The tomogram consists of 256 x 256 picture elements. Each element is called a pixel and is characterized by a CT value or density value. Most CAT scanners are able to differentiate between 2000 CT values. Water has a typical density of zero air has a value of -1000. The different body tissues can be differentiated accurately, as they have no overlapping CT frequencies. For example, for fat the average range is -180 to -20 for muscle tissues the range is +20 to +180: and hard bone has a density around +1000.
The CT values are reconstructed into a black-gray-white color scale image appearing on the monitor of the scanner
IMAGE ANALYSIS AND ACCURACY OF CAT IN MEASURING BODY COMPOSITION
Technically, almost any number of scans can be taken from an animal. In the Norwegian programs, at first summarized by Vangen, 1992) 11 scans were taken at fixed anatomical positions of the pig from the 9th thoracic vertebrae to the head of the femur. It was found that information from more than 7 scans was marginal in terms of describing amount of carcass protein, fat and water. The statistical approach taken by Vangen and Allen (1986) was the use of prediction equations to describe the body composition traits.
The amount of the data was too large without summarization. Therefore, from every image 40 variables were created as the sum of pixels within a 10 CT value range. Only the range between-200 to +200 was of interest when describing fat, protein and water, 400 CT values/l 0 = 40 independent variables were created from each image. Based on cross-validation technique after stepwise regression analyses with the stop criteria "the best n-variable model" procedures to select variables, the authors concluded that around 30 independent variables in the regression model gave the lowest prediction errors. Prediction equations were developed for separate sexes, and were used the next years in a number of experiments to predict in viva weight of fat, protein and water, meat percentage in the body. The described method was found most practical when scanning large numbers of animals for selection purposes, as it required a limited number of scans, more simple image analysis, and less time per animal.
In the experiments of Vangen and Allen (1986), summarized by Vangen (1992), 208 pigs of two sexes were included in the range of normal slaughter weight. All animals were dissected after CAT scanning and carcass protein, fat and energy measured. It was found that fat and energy showed R2 values of 0.98 and protein R2 values of 0.93 from CAT variables live weight and sex. In addition to live weight and ultrasonic back fat measurements, the CAT variables in boars described 45 percent of the rest variation in carcass protein and 87 percent in carcass fat. The prediction errors were as low as 20-25 percent of the standard deviations for fat in carcass and between 27 and 46 percent for protein in carcass (Table 1).
Prediction equations for
|SD||SEP||SEP x 100 SD|
|Boars and Gilts||1.11||.506||45.6%|
|Boars and Gilts||3.17||.585||18.5%|
Vangen and Kolstad (1986). summarizing the Norwegian CAT studies, conclude:
It was observed that regression coefficients of the prediction equations may show low robustness over time (Storlien and Sehested, 1992). Vangen (1992) mentions several methodical aspects to explain the reasons.
The Hungarian program and approach was different from the Norwegian one, insofar as it had to focus to solve problems related to practical applicability of CAT technology - and other digital imaging methods as MRI - in meat animal selection.
The first question to be answered was whether CAT scanning of live pigs (105 kg live weight)could substitute slaughterhouse evaluations in progeny tests conducted in central stations.
We had to compare tile in viva CAT information data with the data gathered for the same pigs at the slaughterhouse regarding all parameters listed in the Hungarian standard defining carcass quality and contributing to tile progeny test index of the boars or gilts tested, The two parameters taken into consideration in the index are ratio of valuable meat cuts in the carcass(weight of barns + shoulders + neck without fat with bone, related to carcass weight) and ratio of fat in the carcass.
Two hundred and forty pigs from the central test station were CAT scanned close to 105 kg slaughter weight. Two genotypes (Large White and Landrace) and two sexes (gilts and castrates)were included. Breeds and sexes were equally represented.
After CAT spanning, all pigs were dissected in the slaughterhouse according to the Hungarian progeny testing standard.
The CAT scans were taken at the following positions:
|Image 1.||IX. Thoracal vertebra.|
|Image 2.||XI. Thoracal vertebra|
|Image 3.||XIII. Thoracal vertebra|
|Image 4.||II. Lumbar vertebra|
|Image 5.||IV. Lumbar vertebra|
|Image 6.||VI. Lumbar vertebra|
|Image 7.||Head of Femur|
The density ranges selected and used in quantitative analysis for the various tissues were:
|Lower Limit||Upper Limit|
After CAT scanning, the pigs were slaughtered and the following data were collected on their carcasses: weight of the carcass, back fat thickness at three positions, weight of hams, weight of the valuable meat cuts. weight of lean, weight of fat.
Tile CAT (in viva) data set and the quantitative information gathered at the slaughterhouse for the same pigs were analyzed with multiple regression methods to construct prediction equations.
As several analyses showed, special prediction equations are needed for different type of pigs and sexes.
The principles of prediction equation construction are: they are based on the 7 scans taken on each pig. The equations utilize only one fraction of the huge CHAT information available Using the CTPC software developed by Kövér and Berényi, exact distances, areas can be automatically measured for the various tissues of each scan.
The most frequently used CAT information to be included in constructing type and sex specific prediction equations regarding scans 1-7 is the following:
The letters and numbers listed mean tile parameters below:
|DF1P1..7||Backfat thickness in the cutting line|
|DF2P1..7||Backfat thickness at B cm from the cutting line|
|DMP1..5||Diameter of the musculus longissimus dorsi (Mill))|
|AMP1..5||Area of the MLD|
|AFP1..5||Fat area around and intramuscular the MLD|
|AMAP1..5||Muscle area inside the bacon|
|AFAP1..5||Fat area inside the bacon|
|APP3..5||Area of the psoas major|
|MHP5..7||Muscle area of the ham|
|FHP5..7||Fat area of the ham|
The indicated distances and surface areas represent several hundred or thousand pixels (quantitative unit information) on each scan indicated (1....7).
As an example, a simple equation is presented for estimating the ratio of valuable meat cuts by CAT in the carcass of a Large White gilt:
R2 = 0.9362
F = 205.4
sign F = 0.00
AFP3 = Fat area around and intramuscular of MLD at scan position
APP5 = Area of psoas major, scan position Image 5.
DMP2 = Diameter of MLD scan position Image 2.
In Table 2, the R2 values are summarized for several traits influencing carcass composition of pigs predicted in viva by CAT.
|Ratio of valuable meat cuts in the carcass||0.94||0.93||0.94||0.96|
|Ratio of fat in the carcass||0.92||0.92||0.88||0.94|
|Weight of hams||0.94||0.89||0.84||0.89|
|Weight of valuable meat cuts in the carcass||0.86||0.94||0.86||0.88|
|Weight of lean in the carcass||0.81||0.88||0.76||0.90|
|Ratio of lean in the carcass||.0.92||0.92||0.88||0.93|
The data presented in Table 2 show that CAT is a powerful method to predict body composition of live pigs.
Recent data indicate that with traditional CAT systems we can predict body composition of live boars with the same or higher accuracy as presented for gilts and barrows in Table 2.
With the help of further developed software underway, we are not far from being able to predict carcass composition of live pigs with close to 100% accuracy with traditional type CAT systems. The latest generation of Spiral X-ray CATs will enable us to measure tissue volumes in the whole body of the pig automatically within a few minutes (Reps, 1995).
In this way, the CATMAN program developed for traditional CATs and for research purposes by Thompson and Kinghorn (1992) may be applicable in routine selection work.
Some of the new possibilities using Spiral CATs in animal selection will be discussed by my colleagues (Reps et at.) at this meeting, along with the demonstration of how MRI systems can be utilized to screen top breeder candidate pigs to predict predisposition for cardiovascular disorders, using dynamic digital imaging techniques.
Vangen (1992) summarized several fields of CAT, Scholz et at. (1993) of MRI application in animal science, and Fuller et at. (1994) gives a very useful survey regarding ultrasound, CAT and MRI application in human and animal sciences.
SPECIAL REQUIREMENTS NECESSARY TO APPLY CAT SYSTEMS IN ANIMAL SELECTION PROGRAMS
The practical application of CAT systems in animal selection programs necessitates special conditions to be met that are technical, operational and infrastructural in nature.
For the efficient use of CAT in animal selection, special technical modifications are necessary compared to CAT systems used in human diagnostic centers. The most important are: special changeable "patient" tables to reduce scanning time of animals: easily manageable containers for individual transportation of animals: isolated units to be able to keep groups of animals of the same herds together: and very effective disinfecting system throughout all facilities. High health standards are essential in all pedigree herds using the central CAT.
The CAT system, to be economical, should have a large-scale breeding program to serve (pig, sheep, or other species), and it reduces costs - if the adequate infrastructure is provided - to use the facilities for several breeding programs, rotating the various species in time according to the actual needs. It is of great economic and technical advantage if performance testing stations or large nucleus herds are located close to the CAT center (Horn, 1991 a,b).
At present, CAT is used in several areas of animal selection in Hungary: selection of pig breeds and lines (Kövér et at., 1993), terminal sire breed selection in sheep, selection of males lines in geese and meat-type rabbits (Szendr et at., 1992), and red deer selection (Horn et at., 1994). Some of the programs are national improvement projects (as sheep and red deer): others are projects which have influence on a high proportion of the improvement of national genetic resources (30-40%).
To utilize CAT in animal selection programs, it seems vitally important to develop a system enabling us to use the expensive central CAT system only to take the necessary number of scans for each potential breeder candidate: and after the technical termination of scanning, all data should be made transmittable immediately to be further processed using a PC network. In this case, the scanning capacity of the CAT could be increased to a threat extent, and costs could be reduced significantly.
Most of the up-to-date PCS became capable for further processing the tomograph scans efficiently. Software programs were developed for this purpose in our lab and are used now by several major nucleus breeders and companies, and because of their many-fold advantages and cost effectiveness in medical centers (CTPC program developed by Kövér and Berényi) all over the country.
Data Transmission and Connections
The CAT images represent a very large amount of data, exceeding 60,000 pixels in most cases for only one scan. Forwarding the image (picture) data from the tomograph to the PC requires a flexible, high speed data connection. For example. the RS-232 serial data channel with a speed of 1200-9600 baud cannot be used, since forwarding a single scan by that speed would take 4-8 minutes. The main computer of the CAT supposed to be able to be connected to high speed networks (Ethernet, Token Ring, etc.).
The Graphic Workstation
To evaluate the scans, the PC should be of the latest 80 486 or Pentium models with 8 Mbyte memory. The hard disk supposed to be large enough to store the picture data the capacity of the hard disk has to be about 1 Byte. The graphic screens are adequate supplied to every PC today. The possibility to connect the PC to high speed networks is essential. The general outlay of the network attached to CAT (and MR.) systems is shown in Fig. 1.
Services of the Computer Network
All the computers connected to the network reach each other's data without delay. The images Pictures) taken by the CAT can be analyzed at the same time by any number of computers in other buildings, or other locations, without loss of information. In most cases, the scans are stored on servers with large data capacity and can be analyzed in detail at any time. The image data sari be stored for a long time by servers on replaceable media such as CD-ROMs, tape cartridges, magneto-optical disks.
The servers ensure the security of the collected data as well - a very important factor requested often by breeders or breeding companies using the services of the CAT (or MRI) center.
The whole network system and the great database is now of great value to researchers and students as well, interested in many fields of animal sciences (genetics. nutrition, physiology, anatomy, etc.) and human medicine.
The contributions to this manuscript of my colleagues Gy. Kvr, Gy. Pszthy, and I. Reps are greatly appreciated. The research project was financed by OTKA Nr 1626 project and funds provided by the Ministry of Agriculture and the Faculty of animal Science Kaposvar.
Allen, P., and K. Leymaster. 1985. Machine error in X-ray computer tomography and its relevance to prediction of in viva body composition. Livestock Prod. Science 13 :383-398.
Allen, P., and O. Vangen. 1984. X-ray tomography in pigs, some preliminary results. In: Lister, Ed.) In Vivo Measurement of Body Composition in Meat Animals, Elsevier, London, p. 52-66.
Fuller, M.F., P.A. Fowler, G. McNeil, and M.A. Foster. 1994. Imaging techniques for the assessment of body composition. J. Nutr. 124:1546-1550.
Horn, P. 1991a. New methods of in viva body composition evaluations in the selection of species bred for meat production with special reference to X-ray computerized tomography. Magyar Állatorvosok Lapja. 46. 3. 133-137.
Horn, P. 1991 b. The basic principals of X-ray computerized tomography and the special requirements of its practical application in animal selection programs. Állattenyésztés és Takarmnyozás. 40. 1. p. 61-68.
Horn, P., L. Sugar, Gy. Paszthy, and E. Bernyi. 1994. X-ray computerized tomography to measure body composition in red deer. In: (W. vat. Haven, H. Ebedes, and A. Convoy, Eds.) Wildlife Ranching. A Celebration of Diversity, Pretoria, p. 173- 174
Hounsfield, GEN. 1979. Computed medical imaging. Nobel lecture, Dec. 8, 1979. J. Computer Assisted Tomography 4(5):665-674. Raven Press, New York.
Kövér, Gy., P. Horn, G. Kovach, Gy. Paszthy. 1993. Computer tomografiaval nyert adatok es a vagoertek adatok osszefuggese serteseknel. Vagollat es Hustermeles 4. 235-237.
Reps, I. 1995. Spiral CAT systems and their possible application in animal research (Unpublished).
Rontgen, W.C. 1895. User eine neue Art von Strahlen. Sitzungsberichten der Physikalisch- Medizinischen Gesellschaft in Wurzburg. Band 137 132-141.
Scholz, A., U. Ban lain, and E. Kallweit. 1993. Quantitative Analyse van Schnittbildern lebender Schweine aus der Magnet-Resonanz-Tomographie. Zilchtungskunde, 65. 3. 206-215.
Sehested, E. 19S4. Evaluation of carcass composition of live lambs based on computed tomography. 35th Annual Meeting of the EAAP, The Hague, August 6-9, 1984.
Sehested, E. 1984. Computerized tomography of sheep In: (I). Lister, Ed.) In Vivo Measurement of Body Composition in Meat Animals, Elsevier, London, p. 67-74.
Skjelvold, H., K. Groenseth, O. Vangen, and A. Evens en. 1981. In viva estimation of body composition by computerized tomography. Z. Tiers. Zuchtungsbiol. 98. 77-79.
Storlien, H., and E. Sehested. 1992. Slakteen genkaper I avlsarbeidet pa svin (Slaughter traits in pigs). Norvinreport no 1/92 Hamar, Norway. 37 pp. (Cit: Vangen, O., 1992).
Szendro, Zs.. P. Horn, Gy. Kover, E. Berenyi, I. Radnay, B. Nemeth, E. 1992. In viva measurement of carcass traits of meat type rabbits by X-ray computerized tomography. J. Appl. Rabbit Research 15:799-809.
Thompson, JAM., and B.P. Kinghorn. 1992. CATMAN A program to measure CAT scans for prediction of body components in live animals. Aust. Asses. of Animal Breeding and Genetics. Proc. 10th Conf., Rockhampton, Australia, 5 pp.
Vangen, O. 1984. Evaluation of carcass composition of live pigs based on computed tomography. 35th Annual Meeting of the EAAP, The Hague. August 6-9, 1984.
Vangen, O. 1992. Estimation of body composition of pigs using computer assisted tomography. Pig News and Informatione, Vol. 13,4,p. 159-162.
Vangen O., and P. Allen. 1986a. Computed tomography in pig breeding I. Evaluation of in vivo carcass composition (Unpublished) Cit. Vangen, O. (1992).
Vangen, O., and P. Allen. 1986b. Computed tomography in pig breeding II Prediction errors of carcass composition (Unpublished) Cit. Vangen, O. (1992).
Vangen, O., and N. Scandal. 1984. Tissue deposition rate in genetically lean and fat pigs estimated by computed tomography. 35th Annual Meeting of the EAAP. The Hague, August 6- 9, 1984.
Vangen, O, and N. Kolstad. 1986. Genetic control of growth, composition, appetite and feed utilization in pigs and poultry. 3rd World Congr. Gen. Appl. Livestock Prod. XI, 367. Lincoln. Nebraska.