Estimating total body fat using computed tomography in two dog breeds

Abstract

Background: Use of computed tomography (CT) in canine obesity research is limited and accuracy of estimated fat is still questioned. Using latest image analysis programs, previous studies in sheep showed a significant correlation between CT-derived and hot carcass measurements. Hence we tested the method in canine research. Objective: To develop an improved research method to estimate body composition including total body fat, lean tissue and bone in two morphologically diverse breeds of dogs. Design: A whole body scan was performed twice (four weeks apart) on six beagles and six greyhounds using Picker UltraZ 2000 CT scanner. They were sedated to the prone position with fore/hind limbs pointing forward and tails tucked in. The individual CT images were taken at 10 mm interval and edited using latest software (OSIRIX) in diacom format to remove extraneous objects within the image. Images were then converted from 16 bit CT image to 8 bit binary image using Image J and rescaled to 256 greyscales using AutoCAT. Boundaries for fat, lean and bone were set to the scales of 20-130, 131-220, and 221-255 respectively. The AutoCAT program then computed the total weight of fat, lean and bone for each dog. CT-derived total body weight was calculated from the sum of fat, lean and bone. Percentages of fat, lean and bone were also calculated. Pearson correlation analysis between CT-derived and measured total body weight were performed separately in two breeds for two scans (CT1, CT2). Significant differences of fat% between breeds were determined using paired t-test. Outcomes: When CT-derived total body weight was compared with measured total bodyweight, both beagles and greyhounds showed a significant correlation (R²=0.99; P=0.000) for both scans. The method also detected a significant breed difference (P=0.000) in the mean fat% for both scans. Conclusion: The proposed method using latest image analysis software programs showed significant correlation between CT-derived and measured total body weight for both the breeds, and the results were reproducible in two scans. The body compositions of different breeds measured by the CT scan were significantly different. Future studies should compare the proposed method with other existing methods used for detecting body composition such as dual-energy X-ray absorptiometry and deuterium oxide dilution.