Multiparametric Magnetic Resonance Imaging of the Kidneys: Effects of Regional, Side, and Hydration Variations on Functional Quantifications

Abstract

Backgroud:

To standardize renal functional MRI, it is important to understand the influence of side-to-side variation, regional variation within the organ, and hydration states in MRI and to search for variables that are not affected by those variations.

Purpose:

To assess MRI-based biomarkers for characterizing the kidney in healthy volunteers while considering variations in anatomic factors and hydration states.

Study type:

Prospective.

Subjects:

25 healthy volunteers (15 females and 10 males, median age 25 years).

Field Strength/Sequence:

3.0T intravoxel incoherent motion diffusion-weighted imaging, arterial spinning labeling imaging, blood oxygenation level dependent imaging, and three-dimensional MR elastography.

Assessment:

Functional variables were measured before and after water challenge. Regions of interest were manually drawn by two investigators (JC and ZZ, with 8- and 5- years experiences in abdominal radiology) in the cortex, the medulla, and the entire kidney. The medulla/cortex ratio was calculated.

Statistical Tests:

Paired t-test or Wilcoxon signed rank test; interobserver correlation coefficient; repeatability coefficients; Spearman’s correlation; Significance level: P < 0.05.

Results:

Diffusion parameters were only subject to regional variation. R2*, RBF and renal stiffness (RS) showed regional variation, side variation, and dependence on hydration states. For each side and hydration state, the cortex showed significantly higher standard apparent diffusion coefficient (sADC), higher true diffusion (D), lower R2*, and lower renal stiffness (RS) than the medulla. For each region at baseline, the left kidney showed significantly higher R2*, higher RS, and lower renal blood flow (RBF) than the right kidney. For each region and side, RS and RBF increased significantly while R2* decreased significantly after water intake. After introducing the intrinsic regional difference, significantly higher medulla/cortex ratio of RS remained after water intake except for RS@90Hz in the right kidney.

Data Conclusion:

Renal multiparametric MRI quantifications were affected by regional variation, side variation, and hydration states.

INTRODUCTION

Multiparametric magnetic resonance imaging (mpMRI) is a promising tool for noninvasive diagnosis and treatment monitoring in patients with kidney disease, especially patients with chronic kidney disease. It allows quantitative measures of renal perfusion, diffusion, and oxygenation (1). Emerging techniques also enable assessments of biomechanical properties in vivo, such as renal stiffness (RS) by magnetic resonance elastography (MRE) (2). These variables may change under different physiological and pathological conditions, and therefore could be sensitive measures of physiology and pathology, such as renal function and interstitial fibrosis(3–5).

It has been recognized that, as a highly vascularized and heterogeneous organ, selection of regions of interest (cortex vs. medulla, and right vs. left) and standardization of hydration states are important factors affecting the validity of MRI-based functional renal biomarkers (5–9). However, these effects have not been fully accounted-for in many previous clinical studies, which might be one of the reasons for inconsistent results or limited diagnostic performance (4, 10–12).

To standardize renal functional MRI, multiple consensus-based technique recommendations have been published (8, 9, 13). Among those recommendations, a standardized hydration state is recommended for patient preparation. Regional and side variations are also considered, and separate analysis and report of the cortex and medulla for each kidney are recommended. Variables not affected by those variations are also desired for generalized clinical application and comparison between multiple sites. As the first step toward a standardized imaging and analysis protocol, it is important to clearly understand the influence of side-to-side variation, regional variation within the organ, and hydration states in quantitative mpMRI and to search for variables that are not affected by those variations.

MATERIALS AND METHODS

Study population

This prospective study was approved by the Institutional Review Board and written informed consent was obtained from each participant. From June 2021 to October 2021, 25 healthy volunteers were included. Inclusion criteria for volunteers were: (a) no history of any kidney disease; (b) no contraindication for MRI examination; (c) no abnormal findings at target MR examination, which was evaluated by a certificated radiologist with 8 years’ experience (JC). The following data were recorded: age, gender, and body to mass index (BMI).

MRI acquisition

All examinations were performed on a 3T scanner (Discovery MR 750, GE Healthcare, Waukesha, WI) using a 32-channel body coil. All participants were fasted at least 4 hours before the MRI scan, and water-restriction of two hours at least before baseline MRE was required. Two soft, flexible acoustic passive drivers (13cm × 13cm), which were connected to the same active drivers (Resoundant, Inc., Rochester, MN) using a Y connector, were used for MRE, as previously reported (5). The two drivers were placed posteriorly under the left and right back, and were centered at the left and right kidney, respectively.

All participants completed three sessions: baseline, baseline retest, and 30 minutes after water intake. Baseline examination was conducted first with anatomic in and out of phaseT1-weighted imaging and T2-weighted imaging for participants screening, followed by the intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI), blood oxygen level dependent (BOLD) imaging, arterial spinning label (ASL), MRE at 60Hz, and MRE at 90Hz, in that order. A 10 minutes off-scanner rest was applied between the baseline and the baseline retest acquisition. After baseline retest acquisition, participants were escorted out of the scanner room, and were given 8 to 10 minutes to consume one liter of water. After water consumption, participants were given 10 minutes to prepare for the follow-up scan and then repositioned in the scanner. The second session started exactly 30 minutes after the start of water consumption. All functional sequences were scanned in the same order in three sessions.

All sequences were acquired at coronal direction. Participants were instructed to hold their breath at the end of expiration and image acquisition was triggered during each breath hold for all sequences except for the IVIM-DWI, which was acquired at free-breath. Imaging parameters were TR=8000ms, TE=88MS, 13 b-values (0, 10, 30, 50, 100, 150, 200, 400, 600, 800, 1000, 1500, and 2000 s/mm²), FOV=360mm×360mm, matrix=128×128, slice thickness=6mm, and acceleration factor=2 for the IVIM-DWI using an echo-planar imaging sequence with respiratory gating; TR=75ms, 16 TEs (ranging from 2.5ms to 49.9ms), FOV=480mm×480mm, matrix=160×160, slice thickness=3.5mm, and acceleration factor=2 for the BOLD imaging; TR=6000ms, TE=17.6ms, number of delays=1, perfusion labeling time=4000ms, perfusion labeling delay=800ms, FOV=380mm×380mm, matrix=96×128, and slice thickness=8mm for the ASL imaging; TR=1600ms, TE=57.8ms, FOV=440mm×440mm, matrix=96×96, slice thickness=3.6mm, and acceleration factor=2 for the three dimensional MRE at the vibration frequency of 60Hz and 90Hz, respectively. The arterial labeling plane of ASL was 10cm above the center of the left kidney, oriented perpendicular to the aorta, to label the inflow blood in the descending aorta. Detailed imaging parameters are also summarized in Supplementary Table 1.

Post-processing and image analysis

All acquired data were post-processed online or transferred to a workstation (Advanced Workstation 4.6, GE Medical Systems, Milwaukee, WI) for post-processing. For IVIM-DWI, a least squares minimization was performed with the Levenberg-Marquardt algorithm for the fitting of the signal data to derive the true diffusion coefficient (D), the pseudo-diffusion coefficient (D*), and the pseudo-diffusion component fraction (F), using the equation of $\frac{S\left(b\right)}{S\left(0\right)}=F\exp \left(-b\times D^{*}\right)+\left(1-F\right)\mathrm{e}\mathrm{x}\mathrm{p}(-b\times D)$, where S(0) and S(b) were the signal intensity in the pixel with diffusion coefficient 0 and b, respectively. A standard apparent diffusion coefficient (sADC) that incorporates both the true diffusion and the pseudo-diffusion components was also calculated from all b-values by using the mono-exponential model using the equation of $\frac{s\left(b\right)}{S\left(0\right)}=\mathrm{e}\mathrm{x}\mathrm{p}(-b\times sADC)$(14). BOLD images were post-processed with a vendor-provided T2* fitting software using a mono-exponential model, and the apparent relaxation rate R2* map was generated. ASL was post-processed online using a single-compartment model, and renal blood flow (RBF) was calculated (9). For MRE at each frequency, a 3D multiscale direct inversion algorithm was used to generate corresponding wave images, phase images, magnitude images, and shear stiffness maps (15).

Quantitative assessment of mpMRI was conducted using an in-house developed tool based in Python. For each kidney, free-hand regions of interests (ROIs) were manually drawn by two independent investigators (JC and ZZ, with 8- and 5- years’ experience in abdominal radiology, respectively) on the central four slices using anatomic images as reference (e.g., b0 images for IVIM-DWI, and magnitude images and wave images for MRE). Regions with severe artifacts or poor wave propagation were excluded using the wave image and the magnitude image of MRE as references. ROIs were then copied to corresponding parameter maps with adjustments to generate the mean value of each variable. The whole kidney was traced along the border with special attention to avoid the major vascular and renal sinus. One continuous cortical stripe ROI or 4 to 6 separate medullary free-hands ROIs were drawn on each of four slices, and the mean value was calculated separately for the cortex and medulla on a per-kidney basis. RBF from the ASL imaging was only measured for the cortex of each kidney. The medulla/cortex ratio of each variable was calculated. The mean value of two measurements was calculated for final comparison

Statistical analysis

Quantitative values were summarized as mean ± standard deviation. The within-subject coefficient of variation (CV) and interobserver correlation coefficient (ICC, two-way mixed effect, absolute agreement, single measurement) were calculated to assess repeatability and interobserver agreement. The interobserver agreement was defined as excellent, good, moderate, and poor with ICC≥0.8, 0.6≤ICC<0.8, 0.4≤ICC<0.6, and ICC<0.4, respectively. Repeatability coefficients (RC) were also calculated to assess the same-day test-retest repeatability. Variables with poor interobserver agreement or RC>40% were excluded from further analysis. Differences in variables between baseline and after water intake, cortex and medulla, then right and left kidney were compared using a paired t-test or Wilcoxon signed rank test, as appropriate. Spearman’s correlation was used to assess the relationship between variables and the variation of variables. A strong correlation was considered only when the absolute correlation coefficient>0.70 with a p-value <0.05. All statistical analyses were performed with SPSS (IBM SPSS Statistics 25, Armonk, NY), and results were considered significant if p<0.05.

RESULTS

Baseline characteristics

Fifteen female participants and ten male participants were included. The median age of the included 25 volunteers was 25 years old (ranging from 20 to 56 years old). The median body mass index was 20.96 kg/m² (ranging from 16.90 to 26.37 kg/m²). The median duration of water restriction was 3.5 hours (ranging from 2 to 6 hours). The total acquisition time of the third session was controlled within 20 minutes in all participants.

Test-retest repeatability and interobserver agreement

All examinations were included in the analysis. The RC was 16%, 19%, 59%, 42%, 29%, 20%, 12%, and 12% for sADC, D, D*, F, RBF, R2*, RS at 60Hz (RS@60Hz), and RS at 90Hz (RS@90Hz), respectively, in the cortex. The RC was 20%, 21%, 75%, 50%, 17%, 13%, and 15% for sADC, D, D*, F, R2*, RS@60Hz, and RS@90Hz, respectively, in the medulla. The RC was 12%, 14%, 68%, 52%, 14%, 14%, and 15% for sADC, D, D*, F, R2*, RS@60Hz, and RS@90Hz, respectively, in the entire kidney. The RC was 16%, 20%, 81%, 47%, 16%, 9%, and 11% for the medulla/cortex ratio of sADC, D, D*, F, R2*, RS@60Hz, and RS@90Hz, respectively. All biomarkers showed moderate to excellent (ICC=0.42-0.99) interobserver agreement except for the medulla/cortex ratio of the D* (ICC=0.37). Consequently, only sADC, D, RBF, R2*, RS@60Hz, RS@90Hz, and the medulla/cortex ratio of each corresponding biomarker were included in further analyses. Results of test-retest repeatability are summarized in Table 1, and results of interobserver agreement are summarized in Table 2. Diffusion images with different b-values and corresponding signal to noise ratio was shown in Supplementary Figure 1.

Table 1.

The test-retest repeatability of mpMRI on a per-kidney basis.

Variable	Region	Unit	Baseline value#	Retest value#	Bias (%)	CV (%)	RC (%)	ICC (95% CI)
sADC	Cortex	mm2/s	1.89±0.13	1.91±0.15	0.57	5.62	16	0.42 (0.16, 0.62)
	Medulla	mm2/s	1.63±0.14	1.69±0.17	1.61	7.13	20	0.43 (0.18, 0.63)
	Entire kidney	mm2/s	1.85±0.10	1.87±0.14	0.37	4.40	12	0.52 (0.15, 0.63)
	Medulla/cortex	/	0.86±0.05	0.88±0.06	1.04	5.70	16	0.25 (0.03, 0.49)
D	Cortex	mm2/s	1.30±0.11	1.36±0.12	2.09	6.98	19	0.42 (0.08, 0.56)
	Medulla	mm2/s	1.16±0.12	1.18±0.13	0.50	7.57	21	0.49 (0.24, 0.67)
	Entire kidney	mm2/s	1.26±0.09	1.31±0.11	1.70	5.22	14	0.59 (0.31, 0.76)
	Medulla/cortex	/	0.90±0.08	0.87±0.07	−1.59	7.26	20	0.34 (0.07, 0.56)
D*	Cortex	mm2/s	0.67±0.13	0.67±0.22	1.13	21.22	59	0.21 (0.07, 0.46)
	Medulla	mm2/s	0.57±0.15	0.61±0.22	2.12	27.23	75	0.02 (0.00, 0.26)
	Entire kidney	mm2/s	0.74±0.16	0.70±0.22	−3.87	24.53	68	0.07 (0.01, 0.11)
	Medulla/cortex		0.87±0.29	0.93±0.30	3.14	29.33	81	0.04 (0.00, 0.24)
F	Cortex	%	54.18±8.04	52.64±8.97	1.54	13.46	42	0.28 (0.003, 0.51)
	Medulla	%	50.68±9.76	51.02±10.5	0.23	17.93	50	0.14 (0.01, 0.40)
	Entire kidney	%	52.23±5.74	50.55±10.00	−2.57	18.89	52	0.21 (0.07, 0.46)
	Medulla/cortex	/	0.94±0.17	0.98±0.18	1.74	16.83	47	0.15 (0.08,0.41)
RBF	Cortex	ml/100g/min	141.94±34.56	147.13±31.01	2.14	10.62	29	0.81 (0.64, 0.90)
R2*	Cortex	sec−1	18.33±1.86	17.33±1.97	−1.73	3.46	20	0.66 (0.20, 0.89)
	Medulla	sec−1	20.00±2.19	20.00±2.90	−0.01	3.03	17	0.84 (0.54, 0.95)
	Entire kidney	sec−1	19.33±1.21	18.83±2.32	−0.38	2.52	14	0.83 (0.52, 0.95)
	Medulla/cortex	/	1.13±0.19	1.16±0.13	1.71	2.83	16	0.87 (0.60, 0.96)
RS@60Hz	Cortex	kPa	3.49±0.25	3.44±0.29	−0.83	3.97	12	0.71 (0.53, 0.83)
	Medulla	kPa	4.22±0.48	4.22±0.50	0.02	4.30	13	0.84 (0.72, 0.91)
	Entire kidney	kPa	4.22±0.49	4.24±0.53	0.12	4.58	14	0.83 (0.71, 0.90)
	Medulla/cortex	/	1.21±0.09	1.23±0.09	0.84	3.21	9	0.81 (0.67, 0.89)
RS@90Hz	Cortex	kPa	5.92±0.64	5.87±0.60	−0.34	4.18	12	0.82 (0.69, 0.89)
	Medulla	kPa	6.65±0.99	6.54±0.95	−0.75	5.36	15	0.84 (0.73, 0.91)
	Entire kidney	kPa	6.66±0.98	6.55±0.92	−0.80	5.49	15	0.84 (0.72, 0.91)
	Medulla/cortex	/	1.12±0.06	1.11±0.08	0.48	3.71	11	0.59 (0.35, 0.75)

Note:

^#value is mean±standard deviation;

CV, coe1fficient of variation; RC, repeatability coefficient; ICC, interobserver correlation coefficient; sADC, standard apparent diffusion coefficient; D, true diffusion coefficient; D*, pseudo-diffusion coefficient; F, pseudo-diffusion component fraction; RBF, renal blood flow; R2*, the apparent relaxation rate; RS@60Hz, renal stiffness at 60Hz; RS@90Hz, renal stiffness at 90Hz.

Table 2.

The interobserver agreement of mpMRI measurement on a per-kidney basis.

Variable	Region	Unit	Baseline value#	Retest value#	Bias (%)	CV (%)	ICC (95% CI)
sADC	Cortex	mm2/s	1.85±0.14	1.91±0.14	−1.64	3.09	0.85 (0.22, 0.95)
	Medulla	mm2/s	1.67±0.16	1.64±0.18	−1.15	4.81	0.84 (0.67, 0.93)
	Entire kidney	mm2/s	1.87±0.12	1.86±0.13	−0.26	1.52	0.96 (0.90, 0.98)
	Medulla/cortex	/	0.87±0.05	0.88±0.06	0.49	4.60	0.52 (0.15, 0.75)
D	Cortex	mm2/s	1.33±0.12	1.31±0.10	−0.97	3.67	0.81 (0.59, 0.91)
	Medulla	mm2/s	1.19±0.12	1.19±0.11	0.27	5.68	0.67 (0.37, 0.84)
	Entire kidney	mm2/s	1.26±0.10	1.27±0.10	0.01	0.47	0.98 (0.97, 0.99)
	Medulla/cortex	/	0.89±0.07	0.91±0.06	1.24	5.46	0.45 (0.10, 0.71)
D*	Cortex	mm2/s	0.64±0.09	0.68±0.14	2.05	10.92	0.56 (0.23, 0.77)
	Medulla	mm2/s	0.58±0.16	0.62±0.24	2.13	22.66	0.42 (0.04, 0.69)
	Entire kidney	mm2/s	0.70±0.13	0.69±0.13	−0.68	2.49	0.78 (0.65, 0.97)
	Medulla/cortex	/	0.92±0.31	0.91±0.25	0.04	24.05	0.37 (0.04, 0.66)
F	Cortex	%	52.71±7.92	53.15±6.12	0.63	7.11	0.73 (0.47, 0.87)
	Medulla	%	50.33±8.50	49.73±7.89	−0.55	8.94	0.71 (0.44, 0.86)
	Entire kidney	%	52.80±6.95	52.91±6.69	0.15	1.12	0.89 (0.78, 0.99)
	Medulla/cortex	/	0.97±0.17	0.93±0.10	−1.18	10.30	0.45 (0.07, 0.71)
RBF	Cortex	ml/100g/min	135.44±44.80	128.51±44.12	4.97	6.23	0.97 (0.87, 0.99)
R2*	Cortex	sec−1	19.58±3.11	19.26±2.62	0.91	6.88	0.78 (0.65, 0.87)
	Medulla	sec−1	30.92±6.26	31.02±5.21	−0.84	14.34	0.66 (0.47, 0.79)
	Entire kidney	sec−1	20.67±2.29	20.19±2.33	2.21	4.36	0.85 (0.73, 0.92)
	Medulla/cortex	/	1.65±0.50	1.62±0.25	−2.80	15.77	0.45 (0.20, 0.65)
SS@60Hz	Cortex	kPa	3.47±0.25	3.49±0.26	−0.64	2.63	0.93 (0.88, 0.96)
	Medulla	kPa	4.19±0.48	4.23±0.47	−1.04	2.80	0.96 (0.93, 0.98)
	Entire kidney	kPa	4.20±0.49	4.00±0.41	3.50	4.55	0.93 (0.85, 0.97)
	Medulla/cortex	/	1.20±0.09	1.21±0.08	−0.44	2.72	0.91 (0.84, 0.95)
SS@90Hz	Cortex	kPa	5.90±0.63	5.92±0.64	−0.32	1.87	0.98 (0.97, 0.99)
	Medulla	kPa	6.63±0.95	6.60±0.96	0.36	2.35	0.99 (0.98, 0.99)
	Entire kidney	kPa	6.64±0.96	6.32±0.74	1.81	2.95	0.95 (0.90, 0.97)
	Medulla/cortex	/	1.12±0.06	1.11±0.05	0.65	2.57	0.86 (0.76, 0.92)

Note:

^#values are mean±standard deviation;

CV, coefficient of variation; RC, repeatability coefficient; ICC, interobserver correlation coefficient; sADC, standard apparent diffusion coefficient; D, true diffusion coefficient; D*, pseudo-diffusion coefficient; F, pseudo-diffusion component fraction; RBF, renal blood flow; R2*, the apparent relaxation rate; RS@60Hz, renal stiffness at 60Hz; RS@90Hz, renal stiffness at 90Hz.

Regional variation

There were significant regional variations in all variables at both kidney sides and hydration states. sADC and D were significantly lower in the medulla than in the cortex. Although RBF was only measured in the cortex, it was markedly lower in the medulla at visual evaluation (Figures 1 and 2). R2*, RS@60Hz, and RS@90Hz were significantly higher in the medulla than in the cortex.

Figure 1.

Examples of the segmentation of regions of interest in the cortex (red, right kidney), in the medulla (green, right kidney), and in the whole kidney (green, left kidney). IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging, sADC, standard apparent diffusion coefficient; D, true diffusion; D*, pseudo-diffusion coefficient; F, pseudo-diffusion component fraction; BOLD, blood oxygen level dependent; R2*, the apparent relaxation rate; RBF, renal blood flow; RS@60Hz, renal stiffness at 60Hz; RS@90Hz, renal stiffness at 90Hz.

Figure 2.

Multiparametric magnetic resonance imaging of both kidneys at baseline and 30 minutes after 1-liter water intake. Delineated was the contour of the whole kidney. sADC, standard apparent diffusion coefficient; D, true diffusion; R2*, the apparent relaxation rate; RBF, renal blood flow; RS@60Hz, renal stiffness at 60Hz; RS@90Hz, renal stiffness at 90Hz. For both hydration states, sADC, D, and RBF were higher in cortex than in medulla, and the R2*, RS@60Hz, and RS@90Hz were lower in cortex than in medulla. For both hydration state, the R2*, RS@60Hz, and RS@90Hz were higher in the left kidney that in the right kidney, the RBF was lower in the left kidney than in the right kidney, and the sADC and D showed no difference between left and right kidney. For both kidneys, the R2* was decreased after water intake, the RBF, RS@60Hz and RS@90Hz was increased after water intake, and the sADC and D showed no difference between baseline and after water intake.

Side variation

No significant difference was observed for sADC and D in all regions (cortex, medulla, and entire kidney) between the right and the left kidney, both at baseline and after water intake (p value for paired comparison in the cortex, medulla, and entire kidney was 0.76, 0.24, and 0.36 for sADC at baseline, 0.83, 0.36, and 0.54 for sADC after water intake, 0.07, 0.06, and 0.60 for D at baseline, and 0.11, 0.74, and 0.32 for D after water intake). R2*, RS@60Hz, and RS@90Hz were significantly higher in the left kidney than in the right kidney in all regions at baseline. RBF of the cortex was higher in the right kidney than in the left kidney.

Effect of hydration states

No significant difference was observed for sADC and D in all regions (cortex, medulla, and entire kidney) between baseline and after water intake (p value for paired comparison in the cortex, medulla, and entire kidney was 0.22, 0.12, and 0.49 for sADC in the right kidney, 0.10, 0.13, and 0.45 for sADC in the left kidney, 0.98, 0.35, and 0.60 for D in the right kidney, and 0.51, 0.20, and 0.31 for D in the left kidney). R2* demonstrated a slight but statistically significant decrease in all regions after water intake. RBF, RS@60Hz, and RS@90Hz demonstrate significant increase in all regions (cortex for RBF) after water intake. The effect of water intake on mpMRI was observed in both kidneys. Results of mpMRI between different regions and sides of the kidney at different hydration states are summarized in Table 3. An example case of the regional variation, side variation, and hydration effect is provided in Figure 2.

Table 3.

Paired comparison of mpMRI quantifications between different anatomic regions, sides, and hydration states.

Hydration status	Regions	sADC	D	R2*	RBF	RS@60Hz	RS@90Hz
Baseline	RC	1.88±0.14	1.32±0.10	18.48±2.60	145.5±34.76	3.41±0.27	5.63±0.53
	RM	1.65±0.17&	1.19±0.10&	28.81±5.40&	NA	4.06±0.47&	6.27±0.80&
	RE	1.86±0.12	1.26±0.10	19.54±2.08	NA	4.07±0.47	6.24±0.78
	LC	1.87±0.11	1.27±0.09	20.68±3.24*	129.56±31.16*	3.54±0.22*	6.18±0.61*
	LM	1.61±0.14&	1.13±0.10&	33.02±6.46*&	NA	4.32±0.46*&	6.98±0.97*&
	LE	1.84±0.08	1.26±0.08	21.80±1.94*	NA	4.33±0.48*	7.04±0.96*
After water intake	RC	1.93±0.26	1.32±0.14	17.08±2.23*	188.2±54.98*	3.86±0.37*	6.72±0.64*
	RM	1.72±0.23&	1.16±0.13&	26.22±4.73*&	NA	4.81±0.57*&	7.74±0.89*&
	RE	1.86±0.12	1.31±0.10	18.20±2.09*	NA	4.84±0.62*	7.76±0.93*
	LC	1.92±0.14	1.38±0.13	18.75±1.45@#	177.05±55.83#@	3.78±0.30#	6.64±0.58#
	LM	1.68±0.17&	1.17±0.13&	30.94±4.28@#&	NA	4.75±0.48#&	7.77±0.94#&
	LE	1.86±0.12	1.29±0.11	20.30±1.38@#	NA	4.82±0.58#	7.79±0.99#

Note: mpMRI, multiparametric magnetic resonance imaging; sADC, standard apparent diffusion coefficient; D, true diffusion; R2*, the apparent relaxation rate; RBF, renal blood flow; RS@60Hz, renal stiffness at 60Hz; RS@90Hz, renal stiffness at 90Hz; RC, right cortex; RM, right medulla; RE, right entire kidney; LC, left cortex; LM, left medulla; LE, left entire kidney; NA, data not applicable.

^*p<0.05 compared with the measurement in the corresponding region of the right kidney at baseline.

^#p<0.05 compared with the measurement in the corresponding region of the left kidney at baseline.

^@p<0.05 compared with the measurement in the corresponding region of the right kidney after water intake.

^&p<0.05 compared with the measurement in the cortex of the same kidney in the same scan.

Medulla/cortex ratio

The medulla/cortex ratio of sADC, D, and R2* showed no significant difference between different sides and hydration states (right versus left kidney: p value for paired comparison at baseline and after water intake was 0.31 and 0.30 for the medulla/cortex ratio of sADC, 0.38 and 0.06 for the medulla/cortex ratio of D, 0.49 and 0.06 for the medulla/cortex ratio of R2*. Baseline versus after water intake: p value for paired comparison in the right and left kidney was 0.43 and 0.46 for the medulla/cortex ratio of sADC, 0.08 and 0.09 for the medulla/cortex ratio of D, and 0.10 and 0.22 for the medulla/cortex ratio R2*). For RS@60Hz and RS@90Hz, the medulla/cortex ratio showed no significant difference between the right and the left kidney (p value for paired comparison was 0.11 and 0.67 for the medulla/cortex ratio of RS@60Hz at baseline and after water intake, and 0.53 and 0.33 for the medulla/cortex ratio of RS@90Hz at baseline and after water intake). However, significant differences still existed for the medulla/cortex ratio of RS@60Hz and RS@90Hz between baseline and after water intake, except for the RS@90Hz in the right kidney (p=0.06). Results of the medulla/cortex ratio of corresponding variables between different sides and hydration states are summarized in Table 4.

Table 4.

Paired comparison of the medulla/cortex ratio of mpMRI quantifications between different sides and hydration states.

	Baseline		After water intake
	Right kidney	Left kidney	Right kidney	Left kidney	P1	P2	P3	P4
sADC	0.88±0.05	0.86±0.08	0.89±0.05	0.88±0.06	0.43	0.46	0.31	0.30
D	0.90±0.05	0.89±0.09	0.88±0.05	0.86±0.08	0.08	0.09	0.38	0.06
R2*	1.58±0.28	1.62±0.36	1.54±0.25	1.66±0.24	0.10	0.22	0.49	0.06
RS@60Hz	1.19±0.09	1.22±0.08	1.25±0.09	1.26±0.08	<0.01	0.01	0.11	0.67
RS@90Hz	1.12±0.06	1.13±0.06	1.15±0.06	1.17±0.08	0.06	0.02	0.53	0.33

Note: sADC, standard apparent diffusion coefficient; D, true diffusion; R2*, the apparent relaxation rate; RS@60Hz, renal stiffness at 60Hz; RS@90Hz, renal stiffness at 90Hz; P1, the p value of paired comparison for the right kidney between baseline and after water intake; P2, the p value of paired comparison for the left kidney between baseline and after water intake; P3, the p value of paired comparison between the right and the left kidney at baseline; P4, the p value of paired comparison between the right and the left kidney after water intake.

Correlations between variables and the variation of variables

Spearman’s correlation between variables at baseline and variation of each variable are detailed in Supplementary Tables 2 to 5. No strong correlation was found in quantitative variables between IVIM-DWI, BOLD, ASL, and MRE at baseline (absolute rho<0.7 for all). No strong correlation was found in the regional variation, side variation, and hydration variation of each variable between IVIM-DWI, BOLD, ASL, and MRE (absolute rho<0.7 for all).

DISCUSSION

This study demonstrated that variations exist in BOLD, ASL, and MRE quantification between different anatomic regions, sides, and hydration states of the native kidney. Whereas IVIM-DWI was only subject to regional variations. After introducing the intrinsic regional variation, the medulla/cortex ratio of R2* was not affected by hydration states and side variations. The medulla/cortex ratio of RS, either at 60Hz or at 90Hz, is not affected by side variation, but the hydration effect still exists. Additionally, no strong correlation was found in quantitative variables between techniques, nor in the regional variation, side variation, and hydration variation of variables between techniques. This study may provide references for patient preparation and data analysis for future studies using mpMRI to assess kidney diseases.

The IVIM-DWI is an advanced diffusion technique which provides separate estimation of the true water diffusion between structures (quantified as D) and the micro-perfusion in the microvasculature or renal tubules (quantified as the D* and F), both may decrease in patients with impaired renal function or interstitial fibrosis(13, 16, 17). The R2* of the BOLD imaging is directly proportional to tissue deoxyhemoglobin levels(18), and the RBF by the ASL technique is a direct measurement of renal perfusion using arterial blood as endogenous tracer(9). Under pathological conditions such as renal fibrosis or hypoxia, the RBF decreases and the R2* may increases(17, 19). The MRE is a novel technique that assesses the biomechanical properties of tissue in vivo, such as the stiffness. The accumulation of extracellular matrix components during development of renal fibrosis typically hardens affected tissue and causes increased RS(19).

Among all explored techniques, IVIM-DWI yields the most stable quantification that are not affected by hydration states and side variations. The results that both the IVIM-derived D* and F have poor repeatability agree with prior studies (3, 20). Sigmund et al explored the effect of water intake on IVIM-DWI using a breath hold echo-planar readout sequence with 8 b-values (highest b-value=800 s/mm²), and reported a significant increase in sADC and D after water intake (20). However, in a recent study by Kemeiene et al, which used a free-breathing echo-planar readout sequence with 11 b-values (highest b-value=1500 s/mm²), no significant difference was found for IVIM parameters in the renal parenchyma after water intake (21). Water diffusion is consistently lower in the medulla than in the cortex. Consequently, the consensus-based technique recommendation for renal DWI recommends a normal hydration states and separate quantification of the cortex and medulla. However, it remains undetermined regarding the number of b-values and the highest b-values for IVIM imaging of kidneys (13). Result of this study are similar with that of Kemeiene’s, as is the imaging protocol. It is speculated that the discrepancy in the effect of hydration states on diffusion parameters might be due to the difference in number of b-value and the highest b-value. Diffusion with more b-values and the highest b-value >800 s/mm² might be less affected by hydration states. Further studies are needed to verify the hypotheses.

BOLD, ASL, and MRE quantifications all demonstrated regional variation, side variation, and hydration state variation. The regional variation in mpMRI between different regions can easily be explained by the intrinsic anatomic and functional heterogeneity, where the cortex is rich in glomeruli that are soft and highly perfused, and the medulla is mainly composed of tubulars and fibrotic interstitium that is stiffer and less perfused(22). Low perfusion and high oxygen-consuming tubular sodium reabsorption together contributed to the low tissue oxygenation in the medulla, reflected in BOLD as higher R2*(6). Side variation and the hydration effect of R2* can both be explained by the difference in perfusion, reflected in ASL as higher RBF in the right kidney and after water intake in our study. Higher perfusion or the water-induced reduction in oxygen-consuming tubular transportation and the local deoxyhemoglobin volume fraction due to increased tubular volume after water intake resulted in the decreased R2* in both the cortex and the medulla(6). The increase in RS after water intake can also be explained by increased perfusion. The side variation of RS, however, cannot be explained by the difference in perfusion. Although perfusion was higher in the right kidney, stiffness was consistently higher in the left kidney. Gandhi et al also observed higher stiffness in the left kidney than in the right kidney and speculated that the higher position of left kidney may have contributed to this (5). Potential reasons of higher stiffness in the left kidney could also be (a) the left kidney is further from the inferior vena cava than the right kidney and has a longer renal vein, which tends to be compressed by structures like the superior mesenteric artery or jejunum, causing stenosis of the left renal vein in extreme cases known as the nutcracker syndrome(23). Therefore, the higher average stiffness of the left kidney may be due to modestly higher venous pressure (similar situation with increased liver stiffness when patient has heart failure(24)). (b) the spleen on the left has higher mobility than the liver on the right and tends to create a static pressure on the left kidney (supported by the observation of more frequent deformation in the left kidney due to splenomegaly or adjacent location with the normal spleen). However, those speculation also needs to be verified.

Regardless of the cause of the side and hydration variation, it affected both cortex and medulla. After introducing the intrinsic regional difference, the medulla/cortex ratio of R2* was no longer affected by side and hydration variations. Similar methods in renal MRI analysis include the recently named 12 layers concentric objects method and the cortical-medullary difference ratio calculated as (cortex-medulla)/cortex, both considered the regional difference (3, 25). The 12 layers concentric objects method measures the gradient of R2* from outer layers to the inner layers and has low interobserver variability; however, it is presently limited by availability (25). The cortical-medullary difference ratio can be calculated for any parameters that are measurable in both the cortex and the medulla. Bane et al found that compared with the T1 value and other parameters, the cortical-medullary difference ratio of T1 demonstrated the highest diagnostic performance in distinguishing stable function from chronic dysfunction with fibrosis (3).

Although the medulla/cortex ratio of RS was not affected by the side variation, it was affected by hydration states. Therefore, standardization of hydration states is still warranted for renal MRE studies. For instance, a normal hydration of 100ml water intake before renal MRI was suggested for patient preparation in current recommendations (8, 9, 13). The potential disadvantage of normal hydration includes the washout of the regional difference. This study also demonstrated that even under the extreme condition of one-liter water intake within 10 minutes, the regional difference still exists. Due to high heterogeneity of renal tissue, a high frequency of 90Hz was also explored for renal MRE in this study. Higher frequency means shorter wavelength and consequently better spatial resolution for MRE (26). Results showed that both RS@60Hz and RS@90Hz have excellent test-retest repeatability and interobserver agreement. Further studies are needed to verify the clinical application of MRE at 90Hz.

Limitations

First, to accurately illustrate the regional variation, this study was conducted in healthy volunteers, and 84% (21/25) of them were under the age of 30. Accurate measurement of the cortex and medulla might be challenging in patients with advanced kidney diseases due to decreased cortico-medullary differentiation. Second, the repeatability of mpMRI after water intake and the duration of hydration effect on mpMRI were not explored in this study. Third, the individual duration of water restriction ranges from 2 to 6 hours. Due to inaccurate recall of the water restriction time, we did not explore the effect of the duration of water restriction on mpMRI quantifications. Fourth, due to the spatial kernel used in the inversion algorithm, RS measurements near boundaries are subject to partial volume effects which may be especially relevant to measurements of cortical stiffness.

Conclusion

For multiparametric magnetic resonance imaging of the kidneys, the quantification of BOLD, ASL, and MRE is affected by regional variation, side variation, and hydration states. IVIM-DWI is only subject to regional variations. For future studies using mpMRI to assess kidney diseases, separate measurements of these biomarkers in cortex and medulla for each kidney may be useful in some quantitative studies in recognition of these systematic differences. Variables that consider the intrinsic regional difference, such as the medulla/cortex ratio, may also useful, especially for BOLD and MRE imaging.

Supplementary Material

Supp

Supplementary Figure 1. Diffusion images with different b-values and corresponding signal to noise ratio (SNR). SI, signal intensity; SD, the standard deviation of the signal intensity; ROI, region of interest; WL, window level; WW, window width.