Introduction

Renal tissue hypoxia is considered to be an important factor in the development of numerous acute and chronic kidney diseases. Blood oxygenation sensitized MRI can provide quantitative information about changes in renal blood oxygenation via mapping of T₂*. Simultaneous MRI and invasive physiological measurements in rat kidneys demonstrated that changes in renal T₂* do not accurately reflect renal tissue oxygenation under pathophysiological conditions.^1,2 Confounding factors that should be taken into account for the interpretation of renal T₂* include renal blood volume fraction and tubular volume fraction.^1,2 Tubuli represent a unique structural and functional component of renal parenchyma, whose volume fraction may rapidly change due to alterations in filtration, tubular water reabsorption, or tubular outflow.^3,4 Diffusion-weighted imaging (DWI) provides a method for in-vivo evaluation of water mobility. In the kidneys water mobility may be linked to three different sources: i) tissue water diffusion, ii) blood flow, and iii) tubular fluid flow. To account for the tubular volume fraction in renal diffusion assessment the commonly used bi-exponential IVIM modelling,⁵ was recently extended to a tri-exponential approach.⁶ Nevertheless, starting values are crucial and fixing some model coefficients is common practice in order to increase fit stability and improve the sensitivity of the model to physiological changes.⁷ This applies particularly to the tri-exponential model (6 fit parameters), but runs the risk of introducing a bias. Here we explored the feasibility of assessing tubular volume fraction changes using the non-negative least squares (NNLS) approach that is data-driven and requires no a priori knowledge (number of exponential components, starting values, fixed coefficients).^8,9

Methods

In-vivo experiments with adult female Wistar rats were performed on a 9.4T small animal-scanner (Bruker Biospec, Ettlingen, Germany). A bolus of glucose solution was administered i.v. to induce changes in the tubular volume fraction. We employed a diffusion-sensitized split-echo RARE variant to ensure renal-DWI free of geometric distortion. The b-values were: 0,4,8,12,18,24,34,43,52,75,115,201,300,460,600 and 800s/mm². To account for the non-isotropic diffusion, images of three orthogonal diffusion directions were averaged. Images were de-noised with a spatially-adaptive-non-local-means filter. ROIs were defined in the cortex(COR), outer medulla(OM) and inner medulla(IM) using semi-automated kidney segmentation (Fig.1)¹⁰. The NNLS method⁸ was implemented by adapting an open-source toolbox⁹ and used to analyze the measured signal decay of each ROI. The NNLS analysis yields a spectrum of detected exponential components, where each peak represents a (pseudo)diffusion compartment with a mean diffusivity (MD; geometric-mean of peak) and volume fraction (area-under-the-curve). Monte-Carlo-simulations were performed to assess the impact of SNR on the NNLS results.

Results

Split-RARE DWI provided excellent image quality with anatomic fidelity and ample diffusion contrast in the rat kidney (Fig.2). Monte Carlo simulations showed that the estimated MD and fraction values deviate from the generated decay signal by <10% for SNRs higher than 400 (Fig.3). SNR levels of the measured in-vivo data always exceeded this threshold, after averaging over each ROI. NNLS revealed three distinct components for all renal layers, at baseline (Fig.4) as well as during hyperglycemia (Fig.5). At baseline these consisted of a slow component (MD_slow= 1.78–2.34x10^-3mm²/s, f_slow= 0.76–0.83), intermediate component (MD_intermediate= 9.15–9.68x10^-3mm²/s, f_intermediate= 0.14–0.22) and fast component (MD_fast= 181–184x10^-3mm²/s, f_fast= 0.02–0.03). During hyperglycemia significant alterations in the MDs and fractions were observed in all renal layers. While the f_slow= 0.14–28 decreased, f_intermediate= 0.40–0.73 and f_fast= 0.11–013 increased. Furthermore, the MD_slow= 0.56–0.87 x10^-3mm²/s decreased in all regions, as did the MD_intermediate= 3.93–4.46x10^-3mm²/s in the medulla but not in the cortex MD_intermediate= 9.38x10^-3mm²/s

Discussion and Conclusion

This work demonstrates the feasibility of renal tubular volume fraction assessment by DWI in combination with a data-driven IVIM analysis. These unbiased model-free results confirm the existence of the three distinct exponential components in renal DWI data. Baseline MD_slow and MD_intermediate are in the range of diffusivities reported in previous IVIM studies.^7,11 The obtained diffusivities, fractions and their changes during hyperglycemia support the interpretations of slow, intermediate, and fast components as representing tissue diffusion, tubular flow, and blood flow respectively. Even so, it has been suggested that capillary blood flow might also contribute to the intermediate component.⁷ Observed alterations during hyperglycemia may be explained by expected changes in physiological parameters. Acute hyperglycemia induces osmotic diuresis and increases glomerular filtration rate and renal blood flow,¹² which is reflected by the increases in f_intermediate and f_fast, respectively.¹² Probing vascular and tubular volume fractions in the kidney is essential for detailing and interpreting the relations between changes in renal hemodynamics, tissue-oxygenation, vascular and tubular volume fraction under (patho)physiological conditions. Our novel approach is a promising refinement of the common IVIM analysis for the unbiased MRI assessment of renal tubular volume fraction.

Acknowledgements

We thank Stefanie Münchberg for technical assistance. This work was supported in part by the Bundesministerium für Bildung und Forschung (BMBF, German Federal Ministry for Education and Research; grants VIP+ 03P00081, VIP+ 03P00082, VIP+ 03P00083).

References

[1] Pohlmann et al., Acta Physiologica (2013) 8(2):673-689; [2] Pohlmann et al., Invest. Radiol. (2014) 49(8):547-60; [3] Niendorf et al. Acta Physiologica (2014) 213(1):19-38; [4] Knepper et al. (1977) 12(5) 313-323; [5] Le Bihan et al. Radiology (1988) 168(2):497-505; [6] van Baalen et al., Magn Reson Med. (2017) 46(1):228-239; [7] van der Bel et al., European Journal of Radiology (2017) 91(3) 168-174; [8] Lawson et al., 1974; [9] Bjarnason et al., Journal of Magnetic Resonance 206(2):200-204; [10] Pohlmann et al., Tomography (2017) 3(4):188-200; [11] Zhang et al., Abdominal Radiology (2018) 43(3):655-662; [12] Woods et al., The American journal of physiology (1987) 252(1):65-73