Comparison of two methods to determine single kidney GFR from MR renography: A multicompartmental model and Patlak- Rutland analysis

L. Bokacheva, A. Huang, H. Rusinek, M. Kaur, Q. Chen, E. Kramer, E. Leonard, V. S. Lee Department of Radiology, New York University School of Medicine, New York, New York, United States, Department of Chemical Engineering, Columbia University, New York, New York, United States Introduction Dynamic gadolinium-enhanced MR images of the kidneys have been investigated by several groups for the noninvasive measurement of renal function. However, methods of analysis vary across groups. We compared two approaches, a multicompartmental model and Patlak-Rutland method for the determination of single kidney glomerular filtration rates (GFRs) against same day reference nuclear medicine measurements. Methods Dynamic renal perfusion images were acquired on 1.5 T scanner (Avanto, Siemens) in 9 patients using 3D FLASH sequence (TR/TE/flip angle=2.84/1.05/12°). Total observation time was 20 min. Patients received 4 ml (2 mmol) injection of Gd-DTPA over 2 s followed by 20 ml saline flush at the beginning of the scan. Resulting kidney images were segmented into cortex and medulla and the corresponding signal intensity curves were extracted. Blood signal was measured in an ROI placed in the aorta at the level of the kidneys. Tissue and blood signal intensities were converted into concentration of Gd using a reference phantom curve [1]. Pre-contrast T1 values required for the conversion were measured by low-flip angle inversionrecovery prepared TrueFISP method [2]. Resulting Gd concentrations as functions of time were analyzed with a multicompartmental renal model [3] and Patlak-Rutland method [4]. The multicompartmental model uses measured aortic concentration as an input function and models the renal cortex as comprised of vascular and proximal tubular compartments, while the medulla consists of vascular and loop of Henle compartments. Cortical and medullary Gd concentration-time curves are fit, using measured regional volumes as fixed parameters. Glomerular filtration rate was determined as a free parameter of the medullary curve fitting. The Patlak-Rutland method based on a simplified two-compartment model [4]. GFR was determined as the slope of the plot of K(t)/b(t) vs b(τ)dτ/b(t), where K(t) is the amount of Gd in the kidney and b(t) is the concentration of contrast in the aorta, multiplied by (1-Hct), where Hct is the hematocrit (Hct=0.4). Actual concentrations obtained as described above were used to create Patlak-Rutland plots. Time intervals between 45 and 150 s were selected for determination of GFR in most cases, as these were shown to provide the most accurate GFR values [4]. Reference GFRs were determined from Tc-DTPA blood clearance values (for global GFR), combined with gamma camera based methods to determine split renal function (scintigraphy), and hence single kidney GFR; MR and nuclear medicine studies were performed on the same morning. Results