MULTIVARIATE STATISTICAL-ANALYSIS OF IMAGES FROM MAGNETIC-RESONANCE-TOMOGRAPHY ON LIVING PIGS

One important aim of Magnetic-Resonance-Tomography (MRT) in animal bre eding research is the in vivo estimation of body composition (proporti on of fat and lean tissue). 1500 images of 75 German Landrace gilts an d castrates (90 kg live weight) from a growth experiment at the Instit ute of Animal Breeding and Animal Behavior in Mariensee were evaluated for the measuring position 13th/14th thoracic vertebra. Five transver sal slices with four measurements (4 echoes) on each slice were perfor med by a Multi-Slice-Multi- Echo-(MSME) imaging sequence. For referenc e, the pigs were slaughtered and the cuts were dissected into lean, fa t and bone. After a Cluster Analysis based on the distribution of imag e grey values and a Stepwise Regression Analysis the combination best fitted in order to estimate lean tissue percentage and fat tissue perc entage on living pigs was got by including echo informations 2 to 4 in regression equations (R2 = 0,76 and 0,74 respectively).