Impact of ROI delineation methods on the repeatability of quantitative indices in 99mTc-MAG3 diuretic renography

To assess how two different region-of-interest (ROI) delineation strategies - whole-kidney ROI and cortical ROI - affect the intra- and interobserver repeatability of quantitative indices in ^99mTc-MAG3 diuretic renography.

We retrospectively analyzed 85 patients (mean age 52.3±17.2 years) who underwent standardized ^99mTc-MAG3 diuretic renography for renal function assessment. Whole-kidney and cortical ROIs were generated using manual and semi-automated techniques. Intraobserver repeatability was assessed by repeated analysis by an experienced radiographer with an interval of one month between processing sessions. Interobserver repeatability was assessed by independent analysis performed by an experienced radiographer and a junior radiographer. Extracted parameters included differential renal function (DRF), time to maximum counts (Tmax), and time to half-peak counts (T1/2) per side (TmaxR, TmaxL and T1/2R, T1/2L, respectively). Repeatability was assessed using Bland-Altman analysis, Lin's concordance correlation coefficient (CCC), and Fleiss' kappa for ordinal T1/2 classification. The Pitman-Morgan test was applied to compare intraobserver variance between whole-kidney and cortical ROI-derived parameters.

Intraobserver repeatability was high across both ROI strategies (whole-kidney ROI, cortical ROI) and delineation methods (manual, semi-automated). DRF demonstrated minimal bias and narrow 95% LoA (whole-kidney ROI: 0.02±1.40% and 0.04±3.29% for manual and semi-automated delineation; cortical ROI: 0.15±1.32% and 0.38±3.06%, respectively), with high CCC values (whole-kidney ROI: 0.991 and 0.955; cortical ROI: 0.992 and 0.961, respectively). Tmax also showed substantial agreement, with small bias and high CCC values. T1/2 classification demonstrated almost perfect intraobserver repeatability, with weighted kappa values for T1/2R of 0.906 and 0.867 for whole-kidney ROIs and 0.870 and 0.850 for cortical ROIs using manual and semi-automated delineation, respectively. Interobserver repeatability was likewise high overall, although agreement was consistently slightly lower for cortical compared with whole-kidney ROIs. For DRF, interobserver CCC values were 0.978 versus 0.971 for manual delineation and 0.944 versus 0.916 for semi-automated delineation for whole-kidney and cortical ROIs, respectively. For manual T1/2R classification, weighted kappa was 0.902 for whole-kidney ROIs and 0.717 for cortical ROIs.

Quantitative indices in ^99mTc-MAG3 diuretic renography showed high intra- and interobserver repeatability irrespective of ROI strategy. Among renal function metrics, repeatability was highest for DRF, while Tmax and T1/2 also demonstrated high repeatability with only minor sensitivity to ROI selection. Both whole-kidney and cortical ROI approaches yielded reliable measurements, although cortical ROI delineation showed slightly lower interobserver agreement, suggesting greater sensitivity to operator-dependent variability.