MR-based semi-automated quantification of renal functional parameters with a two-compartment model—An interobserver analysis

Abstract Purpose To assess the interobserver agreement in a semi-automated quantification approach of MR-renal perfusion and filtration parameters with a two-compartment model analysis. Materials and methods Twelve consecutive patients underwent renal perfusion measurements after intravenous injection of 7 ml Gd-BOPTA at 4 ml/s at 3.0 T. Two independent observers placed two regions of interest (ROI) manually on the axial slice, one in the abdominal aorta to determine the arterial input function (AIF), and one at the tissue–air interface for retrospective triggering. The data were fitted on a pixel-by-pixel basis to the two-compartment model, producing maps of the perfusion parameters F P (plasma flow), T P (plasma mean transit time) and of the tubular filtration parameters F T (tubular flow) and T T (tubular mean transit time). A cortical ROI was segmented by selecting those pixels with plasma volume V P  > 10 ml/100 ml, and the model fit was repeated on a ROI basis to produce the cortical averages. Results The average values (observer 1/observer 2) were F P (226.2/187.3 ml/100 ml/min), T P (9.0/9.1 s), F T (23.5/20.8 ml/100 ml/min), T T (142.1/140.0 s). The correlation coefficients between both observers were 0.90 ( F P ), 0.80 ( T P ), 0.80 ( F T ), 0.78 ( T T ). Correlations of all values were significant ( p t -test yielded significant differences for F P ( p  = 0.004). Discussion/conclusion The data demonstrate a significant systematic difference for the parameter F P , while T P seems to be most stable. Further decrease of the residual variability of all parameters seems desirable to improve the robustness of the method for clinical routine.