Diffusion parameters derived from multi b-value DWI-data as surrogate marker for kidney function
Oliver J. Gurney-Champion1,2, René van der Bel3, Martijn Froeling1,4, C.T. Paul Krediet3, and Aart J. Nederveen1

1Radiology, Academic Medical Center, Amsterdam, Netherlands, 2Radiation Oncology, Academic Medical Center, Amsterdam, Netherlands, 3Internal Medicine, Academic Medical Center, Amsterdam, Netherlands, 4Radiology, University Medical Center Utrecht, Utrecht, Netherlands


In this study, we show the potential of the tri-exponential intravoxel incoherent motion (IVIM) model parameters as surrogates for kidney function. We show (n=8) that f1, f2 and D (fit parameters) changed as function of angiotensin-II dose (0, 0.3, 0.9 and 3 ng/kg/min). The changes in f1 and f2 correlated (correlation coefficient r=0.42 and -0.62; p=0.01 and p<0.001) to changes in effective renal plasma flow (ERPF) and glomerular filtration rate (GFR) deduced from a 125I-thalamate and 131I-hippuran clearing test. Finally, we showed (n=6) high inter-session and intra-session repeatability with coefficients of variation below 15% for all fit parameters.


The currently available gold standard tests to determine kidney glomerular filtration rate (GFR) and effective renal plasma flow (ERPF) are invasive, costly and time-consuming (6 hours). Recently, it was suggested that blood fraction, urine fraction and tissue fraction can also be characterized by the tri-exponential intravoxel incoherent motion (IVIM) model for diffusion weighted MR-images (DWI).1 Potentially, this model can provide a cheaper, faster and non-invasive method to assess similar information as the standard kidney function tests.

In this study, we hypothesize that renal function can be reliably assessed by perfusion fractions derived from a tri-exponential IVIM model. Our aim was to evaluate the sensitivity of these parameters to changes in kidney function and to test the repeatability of the method.


Our DW-scan consisted of twenty 2D coronal slices acquired with an EPI read-out during free breathing. Further sequence parameters were: TE=45ms, TR=1344ms, field of view=336x204mm2, voxel size 3.5x3.5mm2, slice thickness=3.5mm, BW=35Hz/voxel, b-values=0,2,4,8,12,18,24,32,40,50,75,110,200,300,450 and 600s/mm2 with 9 gradient directions.

To compensate for in-plane respiratory motion, we used a 2D rigid registration algorithm (elastix2) to register each slice to the mean b=0 s/mm2 image. For each voxel, we fitted a tri-exponential model which returned the diffusion coefficient (D), pseudo-diffusion coefficients (D*1, D*2) and signal fractions with D*1, D*2 and D (f1, f2, f3). D*1 and D*2 were fixed to values obtained from a fit to data from all subjects at baseline. To assess kidney function, we took the mean parameter values from the voxel-wise fits to data from the kidney medulla and cortex.

To investigate how the model parameters change as a function of kidney function, we acquired DW images in 8 healthy volunteers (aged 18-24 years) on a 3T (Philips, Ingenia) scanner. The volunteers all were subjected to a continuous angiotensin-II infusion (0, 0.3, 0.9 and 3.0 ng/kg/min) during acquisition. During a second visit, the effect of the angiotensin-II on the kidney function, namely GFR and ERPF, were assessed using the gold standard 125I-thalamate and 131I-hippuran clearing tests, during a similar infusion protocol. We applied linear regression fits between the fit parameters and angiotensin-II dose to test if there was any response. In addition, we used a Pearson correlation test to test for correlations between the perfusion fractions, f1 and f2, and the GFR and ERPF.

To study potential challenge unrelated trends and in order to test intra-session repeatability, we obtained time-controlled data by repeating the protocol on 6 healthy volunteers (aged 23-28), without injecting angiotensin-II. To study the inter-session repeatability, we also acquired DW images in 6 healthy volunteers (aged 19–22 years) in two separate sessions. We calculated the inter-session and intra-session within-subject coefficients of variation (CVs).


We fixed D*1 and D*2 to 9.7x10-3mm2/s and 551x10-3mm2/s, respectively.

The mean parameter values at baseline were 1.93±0.08×10-3mm2/s, 11.3±2.4%, 7.8±1.9% and 80.8±2.2% for D, f1, f2 and f3, respectively. The perfusion fractions in the kidneys changed as a function of angiotensin-II dose throughout the kidney (Fig 1). There was a significant response in f1, f2 and D, which changed by -4.7 (p=0.021), 6.4 (p<0.001) and -1.2 (p<0.001) % per ng/kg/min of angiotensin-II (Fig 2). The correlation coefficient, r, between changes in f1 and ERPF was 0.42 (p=0.01), and f2 and GFR was -0.62 (p<0.001, Fig 3). These correlations were higher than those previously reported between BOLD measurements and estimated GFR, for which |r|<0.24.3

The IVIM parameters from the time-controlled data had no significant changes between the first (time-controlled to baseline) and last (time-controlled to max dose) acquisition (Fig 4). For D, f1, f2, f3 the intra-session CVs were 3.5%, 14.4%, 6.5% and 1.9%, and the inter-session CVs were 1.7%, 10.4%, 13.8% and 2.2%, respectively (Fig 5).


We showed that pseudo diffusion signal fraction parameters f1 and f2 derived from a tri-exponential model for DWI-data presented has potential as a surrogate marker for kidney function as: (1) they change with angiotensin-II dose, (2) they showed a stronger correlation with the golden standard kidney function tests than other MR-methods and (3) they are highly repeatable.


No acknowledgement found.


1S. van Baalen et al. ISMRM (2014), p. 2193.

2S. Klein et al. IEEE Trans. Med. Imaging 29, 196–205 (2010).

3H.J. Michaely et al. Kidney Int. 81, 684–689 (2012).


Figure 1: The f2-maps as function of angiotensin-II dose for all patients.

Figure 2: f1 f2, f3 and D as a function as a function of angiotensin-II dose.

Figure 3: f1 f2, f3 and D as a function of time-controlled data from acquisitions without injection of angiotensin.

Figure 4: Correlation plots between f1 and ERPF and between f2 and GFR. Values are given as % difference with respect to the baseline measurements.

Figure 5: Bland-Altman plots of inter-session repeatability for the tri-exponential fit parameters.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)