In calculating the relative and absolute renal uptake of technetium-99m mercaptoacetyltriglycine (MAG3), inter-operator variability in the assignment of the renal region of interest (ROI) is a critical factor. Our goal was to develop a semi-automated method of assigning the renal ROI and then to compare the inter-operator variability in calculating the percent injected dose (%ID) in the kidney at 1-2 min, using semi-automated versus manual ROIs. The manual ROIs were drawn independently by three operators (A, B and C). Operator A had about 20 years, experience in nuclear medicine, while operators B and C respectively had 3 years and 1 year of experience. In the semi-automated renal ROI selection method using the double-threshold technique, the operators only click around the centre of each kidney. The same three operators processed the ROIs using this double-threshold method on 1-2 min images. The semi-automated method failed in three kidneys with very markedly reduced function owing to superimposition by liver or spleen. Inter-operator reproducibility in the remaining 59 kidneys was estimated using manual and semi-automated ROIs. With manual ROIs, the %ID (mean+/-standard error of mean) was 4.32+/-0.167 for A, 4. 14+/-0.165 for B and 3.28+/-0.139 for C. Although there was good correlation among them, these values were significantly different (P<0.0001). Using semi-automated ROIs, the %ID was 4.38+/-0.160 for three operators. No significant difference was observed. Complete reproducibility was shown in 58 of 59 kidneys; the %ID difference of the remaining kidney was only 1.2%. The lowest %ID of all the kidneys successfully detected using the semi-automated method was 0. 77%. The semi-automated renal ROI selection method using the double-threshold technique displays good detectability of the renal contour. The renal uptake calculated using this method is reproducible and acceptable in routine clinical practice.