Accuracy of Quantification for Renography with 99mTc-MAG3: A Study on Virtual and Clinical Data

Sammanfattning: Aim: Renography with 99mTc-MAG3 is one of the most common nuclear medicine examinations. It’s extensive use justifies a thorough investigation of the methods used to accurately determine the split renal function, for instance with regard to background subtraction and correction for renal depth differ- ences. Previous studies on this subject have been made, mainly on patient image data, thus the ”true” values have not been possible to determine. In this study, both patient images and images from virtual patients obtained from Monte Carlo simulations have been used in the analysis in order to determine the most accurate method for quantitating the split renal function. Method: Initially, a transmission measurement was performed with a 57Co flood source and an anthropomorfic phantom (KYOTO KAGAKU) filled with water. The phantom was then filled with an activity concentration correspond- ing to the uptake in a real patient and an activity quantification was performed. Additionaly, twenty sets of simulated dynamic renography images obtained from a gamma-camera simulation of an anthropomorphic digital XCAT phantom to- gether with 25 dynamic patient studies were used for analysis. The split renal function was calculated from the images using three different types of ROIs for background correction; (a) distant perirenal ROI (2 pixels wide and 4 pixels outside the kidney ROI) (b) adjacent perirenal ROI (2 pixels wide and 1 pixel outside the kidney ROI) and (c) adjacent lateral ROI (2 pixels wide and 1 pixel outside the kidney ROI). Distant and adjacent perirenal ROIs were applied with an automated program, adjacent lateral ROI was both automatically and manually drawn. To determine which background correction method that gave the most accurate results, the kidney and the background ROIs were drawn in all simulated images with split renal functions ranging from 50-50% to 90-10%. The background ROI that gave the most accurate result in the simulated images was then used in a separate study to investigate the improvement of applying the geometrical-mean (GM) correction on anterior/posterior images from pa- tient studies. The operator dependence was also studied by comparing results obtained with manually drawn and with automatically generated ROIs. Results: The deviation from the true uptake values was smallest when us- ing the distant perirenal ROI. The largest error was associated with the lateral ROI. The GM method, in contrast to assessment with solely posterior images, corrects for kidneys which are placed at different depths in the body. The use of automatically placed ROIs showed more reproducible results without the intra- operator variability. Discussion: The use of virtual images in conjunction with patient studies was found to be very useful for estimating uncertainties associated with back- ground correction, operator dependence and the use of the GM method. The method which was most accurate and hence suitable to use in clinical routine was the GM method with automatically drawn, distant perirenal background ROIs. Conclusions: The result of this study supports the use of distant ROIs when calculating the split renal function, preferably applied with an automated ROI program. The GM method is recommended as a standard method in the clinic.