(eng): A model of a weighted mixture of regression models is applied to nuclear medicine data collected in treatment of thyroid carcinoma using 131I. Parameters of the mixture are estimated by means of the Bayesian methodology. The relationship between (i) the marker of the therapy success, and (ii) the activity of 131I to be supplied to a patient, a decision on which is made by the physician, is discovered. This decision is supoported by proposing the estimate of the patient-specific 131I activity to be supplied.