The purpose of this study was to establish which level of geometrical accuracy can be obtained during radiotherapy, using portal image analysis, with a minimum number of patient set-up measurements and corrections. A set-up verification and correction procedure using decision rules for improving the set-up of a patient during radiotherapy was investigated by means of a computer simulation. In this simulation study, set-up deviations were assumed to be the sum of random and systematic deviations and varying ratios of random and systematic deviations were studied. The distribution of random deviations (SD equal to sigma) was assumed to be equal for all patients of a specific treatment site. Set-up deviations are measured during the first N consecutive fractions after the start of the treatment or after a patient set-up correction. A set-up is corrected when the deviation averaged over these measurements is larger than an N-dependent action level. This action level is specified by alpha/square root of N, in which alpha is a variable initial action level parameter. After the start of the treatment or after each correction, Nmax measurements are made to decide on a possible (further) correction. By varying alpha and Nmax, the relation between the overall accuracy and the workload has been analyzed. It was possible to obtain a resulting overall accuracy level which is almost independent of the initial distribution of systematic deviations.(ABSTRACT TRUNCATED AT 250 WORDS)