Routines for the estimation of model parameters.
hfit-scale executes a simplex minimisation to estimate the parameters of a composite density model for the grey level frquency distribution (histogram). It operates on the image currently displayed in the Sequence Tool within the specified region of interest (ROI). The model comprises Gaussian peaks for pure tissues and uniform distributions for partial volume mixture pixels. Relative mean grey level and distribution widths are taken from the specified Model File. Optimisation then determines an overall scale and initial normalisation parameters. Other parameters are to be determined via use of the Expectation maximisation algorithm via the algorithms described below. The user is advised to histogram the image in order to ensure that fitting has been sucessful.
E-step computes the probability that each voxel is consistent with the fitted model based on grey level values only (Expectation step of the EM algorithm).
E-step grad computes the probability that each voxel is consistent with the fitted model based on grey level and image slope values (Alternative Expectation step of the EM algorithm).
M-mean computes new estimates for the mean grey level parameters of the model.
M-cov computes new estimates of the covariance terms for the model.
M-prior computes new estimates of the density parameters for the model.
M-k grad calc computes new estimates of the slope parameters for the model (Tina memo 2004-009).