Comparison of volume estimation in CT scans using mixture models

Summary

​This project used data and CT scans from a range of MLA funded projects, including CT images taken from the Sheep CRC Information Nucleus Flock lambs, to compare the estimation of whole body and carcase tissue volume using three approaches to computing a Bayesian spatial mixture model.  Approaches used included the standard MCMC method for spatial mixtures (AIston, Gardner & Mengersen, 2009). This base model was compared with computation using a) variational Bayes and b)  Random scan Gibbs sampling to assess the relative speed and accuracy of the alternate methods.  It is anticipated that these two methods will be substantially faster to compute than the standard MCMC base model but will be less accurate.  However, if the comparison shows close agreement in volume estimates, then this will be of interest in bringing mixture model analysis into the main stream CT scanning program.