Abstract
Estimation of the effective half-life, i.e. the time during which the activity within a region of interest falls to half of its original value, is a task often met in nuclear medicine applications. Usually, the estimation is based on the least squares (LS) fit of a straight line to the observed data expressed in semi-logarithmic coordinates.
Such a solution is susceptible to measurement errors and provides little information on the estimate reliability. The Bayesian solution presented treats this estimation problem in a more efficient way by respecting the well defined probabilistic problem structure and exploiting all information sources available.
The efficiency claimed has been verified on an extensive set of real life data related to diagnostics/therapy of thyroid diseases by I-131. An illustrative account of the experiments performed is presented.