- Email: [email protected]
Quantitative renal function assessment of atheroembolic renal disease using view-shared compressed sensing based dynamic-contrast enhanced MR imaging: An in vivo study
Journal Pre-proof Quantitative renal function assessment of atheroembolic renal disease using view-shared compressed sensing based dynamiccontrast enhanced MR imaging: An in vivo study
Hanjing Kong, Bin Chen, Xiaodong Zhang, Chengyan Wang, Min Yang, Li Yang, Xiaoying Wang, Jue Zhang PII:
S0730-725X(19)30078-5
DOI:
https://doi.org/10.1016/j.mri.2019.10.007
Reference:
MRI 9329
To appear in:
Magnetic Resonance Imaging
Received date:
3 February 2019
Revised date:
9 October 2019
Accepted date:
14 October 2019
Please cite this article as: H. Kong, B. Chen, X. Zhang, et al., Quantitative renal function assessment of atheroembolic renal disease using view-shared compressed sensing based dynamic-contrast enhanced MR imaging: An in vivo study, Magnetic Resonance Imaging(2018), https://doi.org/10.1016/j.mri.2019.10.007
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2018 Published by Elsevier.
Journal Pre-proof
1
Quantitative Renal Function Assessment of Atheroembolic Renal Disease using View-shared Compressed Sensing based Dynamic-contrast Enhanced MR Imaging: An in Vivo Study
Hanjing Kong1, Ph.D., Bin Chen1,2, Ph.D., Xiaodong Zhang3, Ph.D., Chengyan Wang1, Ph.D., Min Yang4, M.D. Li Yang5, M.D. Xiaoying Wang1,3, M.D. Jue Zhang1,6, Ph.D.
Academy for Advanced Interdisciplinary Studies, Peking University, 100871, Beijing, China
2.
Department of technical research and development, Instrumentation Technology and
oo
f
1.
pr
Economy Institute, 100088, Beijing, China
Department of Radiology, Peking University First Hospital, 100034, Beijing, China
4.
Department of interventional radiology and vascular surgery, Peking University First Hospital,
e-
3.
Pr
100034, Beijing, China
Renal Division, Peking University First Hospital,100034, Beijing, China
6.
College of Engineering, Peking University, 100871, Beijing, China
rn
al
5.
* Corresponding Author: Jue Zhang
Jo u
Address: 5 Yiheyuan Rd, Peking University, College of Engineering, Beijing, 100871, P.R.CHINA Tel: 8610-62755036, Fax: 8610-62753562 Email: [email protected]
Journal Pre-proof
2
Quantitative Renal Function Assessment of Atheroembolic Renal Disease using View-shared Compressed Sensing based Dynamic-contrast Enhanced MR Imaging: An in Vivo Study
Abbreviations: view-shared compressed sensing (VCS)
oo
f
Atheroembolic Renal Disease (AERD) glomerular filtration rate (GFR)
pr
Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI)
e-
digital subtraction angiography (DSA)
Pr
VCS reconstructed data (VCSD)
standard deviation (SD)
Jo u
rn
coefficient of variation (CV)
al
region of interest (ROI)
Journal Pre-proof
3
Abstract Atheroembolic renal disease (AERD) is the major cause of renal insufficiency in the elderly, and particularly, the diagnose of AERD is often delayed and even missed due to its nonspecific presentation and the sudden occurrence of an embolic event. To investigate the feasibility of the view-shared compressed sensing (VCS) based dynamic
oo
f
contrast enhanced magnetic resonance imaging (DCE-MRI) in the assessment of AERD in animal models. The reproducibility of VCS DCE-MRI based glomerular filtration rate (GFR)
pr
estimation was first evaluated using the three healthy rabbits. Animal models of unilateral AERD
e-
were then conducted. All the rabbits underwent VCS DCE-MRI and the GFR maps were
Pr
estimated by a commonly used cortical-compartment model. The whole kidney and suspicious lesion region GFR values of embolized kidneys were then compared with the corresponding
al
values of normal kidneys. Finally, the suspicious lesion regions were confirmed by the
rn
corresponding renal specimens and histological findings. The reproducibility of GFR
Jo u
measurements was analyzed using the coefficient of variation and Bland-Altman analysis. The GFR values of normal and embolized kidneys were compared using the Student t-test. Contrast-enhanced images with sufficient diagnostic quality and reduced motion artifacts are obtained at a temporal resolution of 2.5s. The Bland-Altman plot indicated close agreement between the GFR values estimated from between-day scans in healthy rabbits. Besides, there existed significant differences between the pixel-wise GFR values of normal and AERD kidneys in region-based comparison(P<0.0001). The suspicious lesions are consistent well with the renal specimen and histological findings. The preliminary animal study verified the feasibility of VCS DCE-MRI for renal function
Journal Pre-proof
4
evaluation, and the strategy could potentially provide a valuable tool to identify AERD.
Keywords: Atheroembolic Renal Disease, Glomerular Filtration Rate, DCE-MRI, View-sharing,
Jo u
rn
al
Pr
e-
pr
oo
f
Compressed Sensing
Journal Pre-proof
5
Introduction Atheroembolic Renal Disease (AERD) is part of a multisystemic disease and has attracted enhanced attention in recent years for its increasing incidence in the elderly[1]. It is a significant cause of renal loss in patients who suffer from valvular cardiopathy, aortic atheromatosis, and hypercoagulable states[2-5]. AERD is caused by showers of cholesterol emboli from the atherosclerotic aorta to many organs, and because of its variable clinical
renal damage in a significant number of patients [6-8].
oo
f
presentation, the diagnosis may be late or missing and cause more severe or less irreversible
pr
In fact, the emboli may dislodge from the aorta or other major arteries spontaneously or during
e-
or after intravascular trauma with angiographic catheters or after the use of anticoagulants and
Pr
thrombolytic agents. And the kidney is usually involved because of the proximity of the renal arteries to abdominal aorta, wherein the erosion of atheromatous plaque is most likely to occur.
rn
clinic.
al
Therefore, the diagnosis and renal function assessment of AERD is of great importance in the
Jo u
The definitive diagnosis of AERD is made by renal biopsy[9]. However, many patients are too acutely ill to tolerate renal biopsy. In recent years, several imaging methods including radioisotope renogram, renal angiogram, and contrast-enhanced CT have been applied to aid in the diagnosis of renal diseases. However, radioisotope renogram and renal angiogram are invasive and not employed as a first line study[5]. Contrast-enhanced CT has relatively poor sensitivity and patients are exposed to iodinated contrast during their diagnostic workup, placing them at high risk for contrast-associated nephropathy. MR imaging skirts these handicaps by offering anatomic and functional information simultaneously without exposure to ionizing radiation and could provide unilateral renal
Journal Pre-proof
6
functional information. Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) involving temporally resolved imaging and contrast agent injection has been established as a reliable method for estimating of single kidney glomerular filtration rate (GFR) and assessment of renal disease[10]. High spatial resolution and high temporal resolution in DCE-MRI are desirable to obtain
oo
f
accurate measurements of renal functional parameters[11, 12]. To balance the tradeoff between spatial and temporal resolution, strategies employing k-space undersampling have been proposed,
pr
including view sharing methods[13-20], Reduced-encoding Imaging by Generalized-series
e-
Reconstruction[21], k-t Broad-use Linear Acquisition Speed-up technique[22], parallel imaging
Pr
methods[23, 24], and compressed sensing methods[23, 25, 26]. Recently, Levine et al. proposed a method termed view shared compressed sensing (VCS) that combined view-sharing and
al
compressed sensing technique together and showed promising results in breast imaging[11]. The
rn
combination of compressed sensing and view sharing, can be used to trade off between temporal
Jo u
footprint and k-space sampling density. Thus this technique is particularly useful for abdomen imaging for its high temporal resolution and robustness to motion. The high temporal and spatial resolution, less invasive, and function evaluation characteristics of VCS DCE-MRI implying that it may be ideally suited for AERD detection and evaluation. Thus, in this study, we aim to assess the diagnostic performance of unilateral AERD by using a VCS DCE-MRI in a rabbit model. A commonly used cortical compartment model was served to DCE-MRI examinations of 15 rabbits and the discrimination power of the GFR of the healthy kidneys with respect to AERD kidneys was analyzed.
Journal Pre-proof
7
Materials and Methods Animals The in vivo animal experiments were approved by the Hospital Ethics Committee for Animal Research. Experiments were performed on 15 male New Zealand White rabbits (weighing 2.8– 3.3 kg). All the rabbits were free fed with standard animal food and tap water at room temperature and stopped feeding 6 hours before experiments.
oo
f
MRI
In the experiment of DCE-MRI, a dose of 0.5 ml/kg of pentobarbital sodium was injected
pr
through the marginal ear vein in the rabbit with a 24G catheter for anesthetization before
e-
scanning. All the animals were placed in a fixed device in a supine position to limit abdominal
Pr
motion during scans. Heart rate was continuously monitored during the scan by using a pulse sensor of the MR scanner.
al
DCE-MRI data were acquired using a GE 3.0T MR scanner (Signa ExciteTM; General Electric
rn
Medical Systems, Milwaukee, WI, USA) with 8-channel TORSOPA coil. Before DCE-MRI
Jo u
acquisition, data for constructing T1 map were acquired with the identical parameters using conventional GRE imaging sequence with variable flip angles method[27]. Subsequently, the coronal VCS DCE-MRI scan was performed with the following parameters: TR = 3.2 msec / TE = 1.3 msec, flip angle = 12°, acquisition matrix = 256 × 256 × 12, NEX=1, slice thickness = 4 mm, field of view = 180mm, VCS acceleration factor = 4. In dynamic contrast-enhanced experiments, five frames of non-enhanced volumes of the kidney were acquired before the bolus administration. Then, 0.1 mmol/kg body weight of Gd-DTPA (Magnevist, Bayer Schering Pharma AG, Berlin, Germany) was injected, followed by a 5.0 ml saline flush. Images were acquired immediately and totally scan time is about 8 minutes for each rabbit.
Journal Pre-proof
8
To test the between-day repeatability of the VCS method, a rescan was implemented on the second day under the same conditions for 3 healthy rabbits. GFR was calculated based on VCS-DCE images by the cortical-compartment model. The scanning position and protocols stayed the same in all the repetitive measurements. In the experiment of embolized kidneys, 12 rabbits were included. Anesthesia was induced by an intravenous injection of pentobarbital sodium before surgery. Unilateral renal embolization was
oo
f
induced by injection of a 50000-microsphere dose (acryl beads, 40-120 um in diameter) into the right renal artery. A 4F catheter was first injected into the right femoral artery and moved to the
pr
ostium of the right renal artery under the guidance of digital subtraction angiography (DSA).
e-
Then microspheres suspended in 2.0 ml of physiological saline were injected slowly. The
Pr
animals’ body temperature was maintained about 38° by using a hotplate during surgery and
experiments were implemented.
rn
VCS undersampling scheme
al
exposed to an infrared light in the box before MR scans. One week later, the DCE-MRI
Jo u
In this study, the view-shared compressed sensing sampling approach, which integrated the CS undersampling technique into the view-sharing strategy, was adopted. As demonstrated in Fig.1a, the ky-kz plane in typical k-space is divided into two independent regions: the center region 𝐴, which performs fully sampling, and the outer peripheral region B, which performs pseudo-random partial k-space sampling. First, the region 𝐵 is randomly separated into several nonoverlapping sub-regions, 𝐵𝑖 (𝑖 = 1,2 … 𝑁). Then, each sub-region 𝐵𝑖 combines a region 𝐴 to build an initial undersampled k-space, 𝐾𝑖 = 𝐴 + 𝐵𝑖 (𝑖 = 1,2 … 𝑁) for each frame, seen in Fig.1b. All the consecutive 𝐾𝑖 datasets could add up to the full k-space by sharing the data points in the 𝐵𝑖 regions. Subsequently, a two-fold CS undersampling matrix Φ𝐶𝑆 is generated according to
Journal Pre-proof
9
CS theory[28]. The CS matrix is selected uniformly at random and kept largely incoherent with some sparse transformation, used for the peripheral k-space undersampling on the basis of 𝐾𝑖 . Then, in our 3D k-space trajectory, the 𝐾𝑖 matrix is processed with the CS matrix Φ𝐶𝑆 to produce the view-shared CS measurement (MVCS), that is, 𝑀𝑉𝐶𝑆𝑖 = 𝐵𝑖 ∙ Φ𝐶𝑆 (𝑖 = 1,2 … 𝑁) and is used for view-shared CS undersampling scheme in each continuous frame, seen in Fig.1c.Here, ∙ stands for element-wise product of two matrix. The CS matrix has variable density sampling with
oo
f
denser sampling near the center region A, which matches the energy distribution in k-space. Thus, the order of acquisition is also progressively undersampled away from the center k-space
pr
according to the MVCS matrix. Here, N is set to be 3 in this study, thus, 𝑀𝑉𝐶𝑆1 , 𝑀𝑉𝐶𝑆2 , and
e-
𝑀𝑉𝐶𝑆3 are successively used for each undersampling measurement in DCE-MRI, achieving
Pr
much shorter acquisition time of each frame, and the corresponding raw data from each accelerated measurement is presented in terms of 𝐷𝑖 (𝑖 = 1,2 … 𝑁).
al
Subsequently, for image reconstruction, the view-shared compressed sensing algorithm was
rn
performed. The k-space points in MVCS are recombined by view sharing the adjacent 𝐷𝑖 (𝑖 = view-shared CS reconstructed data (𝐷𝑉𝐶𝑆 ), which equals the
Jo u
1,2 … 𝑁) data sets to generate the
CS trajectory according to the proposed acquisition scheme, seen in Fig.1d, that is, the missing portions in region 𝐵 of each frame are shared from the adjacent undersampled k-space data in sub-regions 𝐵𝑖 . Here, three subsequent partial k-space data sets in 𝐵𝑖 are recombined into the current k-space to produce the corresponding 𝐷𝑉𝐶𝑆 for further CS reconstruction,
𝐷𝑉𝐶𝑆𝑖 =
𝐷𝑖−1 ∙ 𝐵𝑖−1 + 𝐷𝑖 + 𝐷𝑖+1 ∙ 𝐵𝑖+1 The k-space samples in the new recombined 𝐷𝑉𝐶𝑆 datasets are corresponding to the points of the identified CS undersampling matrix, and CS reconstruction is then carried out according to the equation as follows: minimize λ1 ‖𝛹𝑆𝑖 ‖1 + λ2 𝑇𝑉(𝑆𝑖 ),
(1)
Journal Pre-proof
10
subject to ‖𝐹(𝑆𝑖 ) − 𝐷𝑉𝐶𝑆𝑖 ‖2 ≤ 𝜀,
(2)
where 𝑆𝑖 is the 𝑖-th image to be recovered, and Ψ denotes the linear operator that transforms from pixel representation into a sparse representation, λ1 and λ2 trade Ψ sparsifying transform with finite-differences sparsity. The typical initial values of λ1 and λ2 were both set to be 0.02 and optimized according to a previous study [29]. Ψ was set to identity transform and TV represents total variation of 𝑆. F is the Fourier transform corresponding to the given 𝑀𝑉𝐶𝑆𝑖 undersampling
oo
f
scheme, and 𝐷𝑉𝐶𝑆𝑖 is the current undersampled k-space of each frame, 𝑖 indicates the frame number.,. Additionally, ε controls the fidelity of the reconstruction which is usually set below the
pr
expected noise level[28].
e-
In this study, the data points percentage in A region were set to be 25%, and data points
Pr
percentage in 𝐵 region were set to be 25%. On the other hand, the CS matrix samples 50% of the total data points. Thus, (25%+25%)*50%=25% data points were finally sampled from the
al
full k-space of each frame, achieving a 4-fold acceleration for VCS measurement. During the
rn
image reconstruction, view-shared compressed sensing reconstruction was successfully
GFR Analysis
Jo u
performed as described in detail above.
GFR was quantitatively evaluated by using the cortical compartment model[30]. A 3×3 voxel region of interest (ROI) within the abdominal aorta distal to the branch of the renal artery was drawn to generate the arterial input function (AIF). To further reduce the noise effect related to respiratory motion, the tail of AIF was fit to a biexponential decay according to previous studies [31, 32]. The tissue signal intensity curves were generated from the DCE-MR images. T1 mapping was obtained by the acquired multi-flip angle data with a proposed method [33]. The contrast agent concentration curve in aorta 𝐶𝑎 (𝑡) and the contrast agent concentration curve in
Journal Pre-proof
11
renal cortex 𝐶𝑅𝑂𝐼 (𝑡) were calculated from the signal intensity-time curves according to the following equation: △ 𝑅1 = 𝑅1𝑝𝑜𝑠𝑡 − 𝑅1𝑝𝑟𝑒 = 𝑟1 𝐶,
(3)
Where △ 𝑅1 is the longitudinal relaxation rate change, 𝑅1𝑝𝑜𝑠𝑡 is the postcontrast longitudinal relaxation rate, 𝑅1𝑝𝑟𝑒 is the precontrast longitudinal relaxation rate, 𝑟1 is the longitudinal
f
relaxivity, and C is the concentration of the contrast agent.
oo
The cortical compartment model was adopted, the contrast agent concentration curve in the ′
pr
glomeruli 𝐶𝑎 (𝑡) could be estimated from 𝐶𝑎 (𝑡) according to the following equation: ′
1
1
e-
𝐶𝑎 (𝑡) = 𝐶𝑎 (𝑡 − 𝜏) ⊗ (𝑑) 𝑒 −𝑑
(4)
Pr
τ is a delay, d is a dispersion constant, and ⊗ is the convolution operation.
rn
al
The contrast agent concentration in the renal cortex is described as following:
′
𝑡
′
Jo u
𝐶𝑅𝑂𝐼 (𝑡) = 𝑓𝑎 𝐶𝑎 (𝑡) + 𝑘21 ∫0 𝐶𝑎 (𝜏)𝑒 −𝑘12 (𝑡−𝜏) 𝑑𝜏.
(5)
where 𝑓𝑎 is the fractional plasma volume，𝑘12 is the output of renal tubules, 𝑘21 is the GFR, The parameters 𝑘21 , 𝑘21 , 𝑓𝑎 , and 𝜏 of Eqs. (4) were fitted with the Levenberg-Marquardt nonlinear leastsquares algorithm. Considering the robustness of Levenberg-Marquardt algorithm (LMA) in the pharmacokinetic modeling of DCE-MRI, it was used to perform the non-linear least squares regression for pixel-wised analysis[34]
Journal Pre-proof
12
Histology Under sterile conditions, kidneys were taken off for pathological examinations one month after the embolization surgery. Kidneys tissues were fixed in 10% formalin and embedded in paraffin for light microscopic observation. Kidneys were sectioned into 2-mm slides and stained with hematoxylin-eosin-saffron. A pathologist specializing in kidney diseases reviewed histological findings. The pathologist was blinded to the imaging findings and the results of pathology and
oo
f
renal specimen were considered as a reference standard for assessing the kidney injury. Data Analysis
pr
DCE-MR images with the best visualization were selected by two radiologists for analysis. For
e-
the reproducibility study, the kidney was segmented by the radiologists and the whole kidney
Pr
GFR was calculated. For the AERD rabbit’s analysis, the lesion ROIs were carefully drawn by a radiologist with thirteen years’ experience and corresponding normal regions were also selected
al
on contra-lateral normal kidney images. Then, the lesion size and the mean value of pixel-wise
rn
GFR in each ROI were calculated for further comparison. All calculations were performed by
Jo u
MATLAB (MathWorks, Natick, MA). Statistical Analysis
Results were represented in mean ± standard deviation (SD). Between-day reproducibility of the GFR measurement was evaluated by coefficient of variation (CV), which was calculated as the standard deviation divided by the mean of the GFR from two scans. Also, Bland–Altman analysis of the scan and rescan (between-day) whole kidney GFR was implemented to validate the reproducibility of VCS method. Student t-test was performed to compare the value of pixel-wise GFR between normal tissue and contra-lateral embolized lesion. P value < 0.05 was considered statistically significant.
Journal Pre-proof
13
Results A representative VCS DCE-MR image and contrast agent uptake curve in normal kidneys are shown in Fig 2. The detailed spatial resolution enables display of the renal parenchyma and renal artery without obvious artifacts. Cortical ROIs were manually drawn (blue and green circle in Fig 2b) and slice that with a branch of the renal artery (red circle in Fig 2d) was selected for aortic ROI drawing. Representative concentration curves at a temporal resolution of 2.5s derived
oo
f
from corresponding ROIs are shown in Fig 2 e,f.
Typical GFR maps of a healthy rabbit derived from VCS DCE-MRI are showed in Fig 3a. The
pr
maps are generally smooth and differentiate between central and subcortical regions, but isolated
e-
pixels or smaller clusters with deviating values can be observed in all maps. Reproducibility
Pr
results of VCS DCE-MRI based GFR are summarized in Fig 3b. The between-day CV of whole kidney GFR measurements of VCS methods was 8.96%. Bland-Altman plot of VCS DCE-MRI
al
derived whole kidney GFR in healthy rabbits are shown in Fig 3b, and all the whole kidney GFR
rn
values lay within the limit of agreement (d-1.96s and d+1.96s).
Jo u
Fig 4 shows the representative twelve frames VCS DCE-MRI of an embolized kidney. As pointed by yellow arrows, the cortical lesion was seen in 10 VCS images and medullary lesion could be observed in 3 VCS images (Red arrows). Dark areas in DCE-MR images showed close spatial correspondence to the embolized regions in the renal specimen (Fig 5c) and is confirmed by histology result (Fig 5d). Representative GFR map and corresponding concentration curves derived from cortical and medullary ROIs are shown in Fig 5 a,b. GFR reduction regions on the GFR map (Fig 5a) matched closely to the heavily embolized regions in the renal specimen (Fig 5c), but the lesion region in GFR map showed a certain degree of underestimation compared with specimen. In
Journal Pre-proof
14
histological findings (Fig 5d), the glomeruli showed ischemic and wrinkled features with thickening change of Bowman’s capsule, and necrosis of the renal tubular epithelial cells was observed, the basement membrane was exposed. Meanwhile, the brush border of some tubular epithelial cells fell off and the tubular epithelial cells became flat and tubular lumen expanded. A total of 17 lesions were found in all rabbits by the renal specimen and confirmed by histological findings. 14 lesions were found by VCS DCE-MRI and the lesion size is 0.14 ± 0.07
oo
f
cm2. For the GFR comparison, the pixel-wise GFR of the embolized lesion was significantly lower than GFR of normal tissue (0.0038 ± 0.005 ml/min vs 0.0075 ± 0.0008 ml/min, P<0.0001)
Jo u
rn
al
Pr
e-
pr
in Fig 6.
Journal Pre-proof
15
Discussion In this in vivo study, we explored the ability of VCS DCE-MRI to detect the impaired renal function in AERD kidney of rabbits. The results confirmed that VCS DCE-MRI could obtain not only diagnostic image quality with reduced motion artifacts and high temporal resolution but reliable GFR estimation with high reproducibility. In addition, compared with normal kidneys, the significant lower GFR values in AERD kidneys corroborated the hypothesis that VCS
oo
f
DCE-MRI was capable to distinguish between AERD and normal kidneys, and thus provide a valuable tool for detection patients with AERD.
pr
AERD involves the acute occlusion of the renal artery or its branches because of an embolus
e-
from a distant source[5]. In this study, we use microsphere that injected directly into the renal
Pr
artery under the guidance of DSA to mimic the embolism process[35]. Also, the catheter was first injected into the right femoral artery and moved to the ostium of the right renal artery, which
rn
and reliable.
al
greatly reduces the surgery trauma, and make the AERD model in rabbits more effective, safe
Jo u
Once the emboli enter the blood circulation, it will stay in a small artery about 150-200 μm in diameter[36] and further causing small artery occlusion and inflammation. The focal feature of AERD, placing greater demand on the spatial resolution of imaging modality. In this study, the spatial resolution of VCS DCE-MRI is 0.7mm*0.7mm. The high resolution and reduced partial volume effect enabled precise visualization of lesion and detection of a majority of lesions (14/17). For GFR quantification, besides spatial resolution, the temporal resolution also plays an important role. Theoretically, a higher temporal resolution yields a higher quality of AIF measurement, and therefore improve GFR estimation. In recent years, view-sharing and
Journal Pre-proof
16
compressed sensing strategies have already exhibited potential contribution in DCE-MRI to improve the temporal resolution[37-39]. By further integrating CS into view-shared peripheral k-space in each 3D volume, the temporal resolution of this VCS acquisition can be significantly improved according to the k-space sampling percentage using the VCS matrix in k-space. We recommend using VCS strategy instead of TWIST or CS based method alone for two reasons.
oo
f
First, all view sharing schemes would introduce temporal blurring of the high spatial frequencies,
pr
making them sensitive to both motion and signal changes due to contrast enhancement[40].
e-
Although it could achieve high resolution, it may also be unstable for clinical abdomen DCE MR imaging. Second, although CS sampling schemes showed promising results in many studies, they
Pr
suffer from image quality degradation under low k-space sampling density[41]. The combination
rn
k-space sampling density.
al
of compressed sensing and view sharing, can be used to trade off between temporal footprint and
Jo u
Theoretically, the VCS strategy could achieve higher resolution compared with CS or view sharing strategy alone. In this study, considering the image quality, we are conservative in choosing the acceleration factor. In this study, a temporal resolution of 2.5s was achieved by VCS strategy and was adequate for renal DCE-MRI. By utilizing a well-established cortical compartment model[30, 42], the GFR values estimated of normal kidneys with high reproducibility was close to literature reports[27, 30], which suggested that VCS was able to provide reliable GFR measurements. The reproducibility is evaluated by CV and Bland-Altman plot. Due to the complexity of DCE
Journal Pre-proof
17
experiments, there are many factors that may affect the consistency of experimental results, including health and metabolism condition, effects of anesthesia on blood pressure and renal perfusion during MRI[43, 44], ROI selection for arterial input function[45]. Many impactful Study reported the CV value of GFR reproducibility in the range of 19.4 to 28 (CV=22.4%[46], CV=19.4%-26.4%[47], CV=28%[48]) and the width of the limits of agreement is within clinical
oo
f
acceptability[49, 50]. Based on these results, we believe that VCS is a relatively stable method for GFR estimation.
pr
In our study, GFR of AERD lesion was significantly lower than that of normal tissue. The
e-
changes in embolized kidneys are also confirmed by renal specimen and histology, which shows
Pr
typical hallmarks of AERD. Altogether, these support our conclusion that VCS MRI is feasible to distinguish AERD from normal kidney, and the derived GFR can be served as a biomarker to
al
differentiate kidneys with AERD from healthy controls.
rn
There are still some limitations. First, only animal experiments were conducted in this study.
Jo u
Additional evaluation of the VCS DCE-MRI is desirable in patients with confirmed pathologies to confirm the clinical relevance of the sequence ultimately. Second, we chose to examine only VCS DCE-MRI in renal imaging for its superior performance in abnormal imaging, and other extended or alternative strategies would also be deserved to further explore. For example, parallel imaging is widely used in clinical scanners and can be further combined with VCS to provide a more practical imaging method for clinic using. Third, we only followed the animals at one week after surgery, it would be interesting to assess at other times to investigate the disease development process. Fourth, the thickness in this study is 4mm, which may cause partial volume effect, so high isotropic‐resolution imaging should be used for further study.
Journal Pre-proof
18
No accepted gold standard for GFR measurements such as creatine or inulin clearance was employed. Though these indicators are probably more sensitive, they are highly invasive and might bias the results by factors like additional anaesthesia[51-53] and repeated measurements[54, 55]. Moreover, imaging the animals by such gold standard only allows for estimation of global GFR[54, 55], and resulting in a potential decrease of sensitivity in the detection and accuracy of the unilateral injury. Besides, recent studies showed that MRI based
oo
f
GFR measurement correlated well to GFR measured by FITC-sinistrin clearance which could be considered as gold standard for measuring the GFR[56, 57]. We believed that the above
pr
limitation might be minor.
e-
In conclusion, this preliminary animal study demonstrated the feasibility of using the VCS
Pr
DCE-MRI in quantitative renal function evaluation, with satisfactory reproducibility. The efficacy of the VCS strategy was carefully verified by histology findings. To the best of our
al
knowledge, this is the first study to perform the function evaluation of AERD using DCE-MRI.
rn
This VCS DCE-MRI could potentially provide a valuable tool to identify AERD and monitor the
Jo u
success of therapeutic methods in clinical practice.
Journal Pre-proof
19
References [1]
Scolari F, Ravani P. Atheroembolic renal disease. The Lancet 2010;375(9726):1650-60.
[2]
Fort J. Renal artery embolism. Renal Failure 1997;19(6):R7-R8.
[3]
Nicholas GG, Demuth WE. Treatment of Renal-Artery Embolism. Arch Surg-Chicago
[4]
oo
f
1984;119(3):278-81. Korzets Z, Plotkin E, Bernheim J, Zissin R. The clinical spectrum of acute renal infarction. Isr Med
Gen Intern Med 2008;23(5):644-7.
Lessman RK, Johnson SF, Coburn JW, Kaufman JJ. Renal-Artery Embolism - Clinical Features
al
[6]
e-
Kansal S, Feldman M, Cooksey S, Patel S. Renal artery embolism - A case report and review. J
Pr
[5]
pr
Assoc J 2002;4(10):781-4.
Mugge A, Gulba DC, Frei U, Wagenbreth I, Grote R, Daniel WG, et al. Renal-Artery Embolism -
Jo u
[7]
rn
and Long-Term Follow-up of 17 Cases. Ann Intern Med 1978;89(4):477-82.
Thrombolysis
with
Recombinant
Tissue-Type
Plasminogen-Activator.
J
Intern
Med
1990;228(3):279-86. [8]
Modi KS, Rao VK. Atheroembolic renal disease. J Am Soc Nephrol 2001;12(8):1781-7.
[9]
Wilson DM, Salazer TL, Farkouh ME. Eosinophiluria in atheroembolic renal disease. The American journal of medicine 1991;91(2):186-9.
[10]
Hahn OM, Yang C, Medved M, Karczmar G, Kistner E, Karrison T, et al. Dynamic Contrast-Enhanced Magnetic Resonance Imaging Pharmacodynamic Biomarker Study of Sorafenib in Metastatic Renal Carcinoma. Journal of Clinical Oncology 2008;26(28):4572-8.
Journal Pre-proof
[11]
20
Levine E, Daniel B, Vasanawala S, Hargreaves B, Saranathan M. 3D Cartesian MRI with compressed sensing and variable view sharing using complementary poisson‐disc sampling. Magnetic resonance in medicine 2017;77(5):1774-85.
[12]
Heisen M, Fan XB, Buurman J, van Riel NAW, Karczmar GS, Romeny BMT. The Influence of Temporal Resolution in Determining Pharmacokinetic Parameters From DCE-MRI Data. Magnetic
oo
Herrmann KH, Baltzer PA, Dietzel M, Krumbein I, Geppert C, Kaiser WA, et al. Resolving arterial
pr
[13]
f
Resonance in Medicine 2010;63(3):811-6.
phase and temporal enhancement characteristics in DCE MRM at high spatial resolution with
Le Y, Kroeker R, Kipfer HD, Lin C. Development and evaluation of TWIST Dixon for dynamic
Pr
[14]
e-
TWIST acquisition. Journal of Magnetic Resonance Imaging 2011;34(4):973-82.
al
contrast‐enhanced (DCE) MRI with improved acquisition efficiency and fat suppression. Journal of
Song T, Laine AF, Chen Q, Rusinek H, Bokacheva L, Lim RP, et al. Optimal k‐space sampling for
Jo u
[15]
rn
Magnetic Resonance Imaging 2012;36(2):483-91.
dynamic contrast‐enhanced MRI with an application to MR renography. Magnetic resonance in medicine 2009;61(5):1242-8. [16]
Lim R, Shapiro M, Wang E, Law M, Babb J, Rueff L, et al. 3D time-resolved MR angiography (MRA) of the carotid arteries with time-resolved imaging with stochastic trajectories: comparison with 3D contrast-enhanced Bolus-Chase MRA and 3D time-of-flight MRA. American Journal of Neuroradiology 2008;29(10):1847-54.
[17]
Riederer SJ, Haider CR, Borisch EA, Weavers PT, Young PM. Recent advances in 3D
Journal Pre-proof
21
time-resolved contrast-enhanced MR angiography. Journal of Magnetic Resonance Imaging 2015;42(1):3-22. [18]
Haider CR, Hu HH, Campeau NG, John Huston I, Riederer SJ. 3D High Temporal and Spatial Resolution Contrast-Enhanced MR Angiography of the Whole Brain. Magnetic Resonance in Medicine 2008;60(3):749.
f
Saranathan M, Dan WR, Hargreaves BA, Clarke SE, Vasanawala SS. DIfferential Subsampling
oo
[19]
pr
with Cartesian Ordering (DISCO): a High Spatio-temporal Resolution Dixon Imaging Sequence for Multiphasic Contrast Enhanced Abdominal Imaging. Journal of Magnetic Resonance Imaging
Zimmer F, Zöllner FG, Hoeger S, Klotz S, Tsagogiorgas C, Krämer BK, et al. Quantitative renal
Pr
[20]
e-
2012;35(6):1484–92.
al
perfusion measurements in a rat model of acute kidney injury at 3T: testing inter-and
Webb AG, Liang ZP, Magin RL, Lauterbur PC. Applications of reduced‐encoding MR imaging with
Jo u
[21]
rn
intramethodical significance of ASL and DCE-MRI. 2013;8(1):e53849.
generalized ‐ series
reconstruction
(RIGR).
Journal
of
Magnetic
Resonance
Imaging
1993;3(6):925-8. [22]
Tsao J, Boesiger P, Pruessmann KP. k‐t BLAST and k‐t SENSE: Dynamic MRI with high frame rate exploiting spatiotemporal correlations. Magnetic resonance in medicine 2003;50(5):1031-42.
[23]
Rosenkrantz AB, Geppert C, Grimm R, Block TK, Glielmi C, Feng L, et al. Dynamic contrast‐ enhanced MRI of the prostate with high spatiotemporal resolution using compressed sensing, parallel imaging, and continuous golden‐angle radial sampling: Preliminary experience. Journal of
Journal Pre-proof
22
Magnetic Resonance Imaging 2015;41(5):1365-73. [24]
Feng L, Grimm R, Block KT, Chandarana H, Kim S, Xu J, et al. Golden‐angle radial sparse parallel MRI: Combination of compressed sensing, parallel imaging, and golden‐angle radial sampling for fast and flexible dynamic volumetric MRI. Magnetic resonance in medicine 2014;72(3):707-17.
[25]
Zhang T, Chowdhury S, Lustig M, Barth RA, Alley MT, Grafendorfer T, et al. Clinical performance
oo
f
of contrast enhanced abdominal pediatric MRI with fast combined parallel imaging compressed
[26]
pr
sensing reconstruction. Journal of Magnetic Resonance Imaging 2014;40(1):13-25. Zhang T, Cheng JY, Potnick AG, Barth RA, Alley MT, Uecker M, et al. Fast pediatric 3D free‐
e-
breathing abdominal dynamic contrast enhanced MRI with high spatiotemporal resolution. Journal
Chen B, Zhang Y, Song X, Wang X, Zhang J, Fang J. Quantitative estimation of renal function with
al
[27]
Pr
of Magnetic Resonance Imaging 2015;41(2):460-73.
[28]
Jo u
2014;9(8):e105087.
rn
dynamic contrast-enhanced MRI using a modified two-compartment model. PloS one
Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magnetic resonance in medicine 2007;58(6):1182-95.
[29]
Li B, Dong L, Chen B, Ji S, Cai W, Wang Y, et al. Turbo fast three‐dimensional carotid artery black‐blood MRI by combining three‐dimensional MERGE sequence with compressed sensing. Magnetic resonance in medicine 2013;70(5):1347-52.
[30]
Annet L, Hermoye L, Peeters F, Jamar F, Dehoux J-P, Van Beers BE. Glomerular filtration rate: Assessment with dynamic contrast-enhanced MRI and a cortical-compartment model in the rabbit
Journal Pre-proof
23
kidney. Journal of Magnetic Resonance Imaging 2004;20(5):843-9. [31]
Tofts PS, Kermode AG. Measurement of the blood‐brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magnetic resonance in medicine 1991;17(2):357-67.
[32]
Misselwitz B, Schmitt-Willich H, Ebert W, Frenzel T, Weinmann H-J. Pharmacokinetics of
[33]
pr
in Physics, Biology and Medicine 2001;12(2):128-34.
oo
f
Gadomer-17, a new dendritic magnetic resonance contrast agent. Magnetic Resonance Materials
Cheng HLM, Wright GA. Rapid high‐resolution T1 mapping by variable flip angles: Accurate and
e-
precise measurements in the presence of radiofrequency field inhomogeneity. Magnetic
Marquardt DWJJotsfI, Mathematics A. An algorithm for least-squares estimation of nonlinear
al
[34]
Pr
resonance in medicine 2006;55(3):566-74.
rn
parameters. 1963;11(2):431-41.
Kassirer JP. Atheroembolic renal disease. New England Journal of Medicine 1969;280(15):812-8.
[36]
Eliot RS, Edwards JE, Kanjuh VI. Atheromatous Embolism. Circulation 1964;30(4):611-&.
[37]
Rapacchi S, Han F, Natsuaki Y, Kroeker R, Plotnik A, Lehrman E, et al. High spatial and temporal
Jo u
[35]
resolution dynamic contrast ‐ enhanced magnetic resonance angiography using compressed sensing with magnitude image subtraction. Magnetic resonance in medicine 2014;71(5):1771-83. [38]
Wang D, Shao J, Rapacchi S, Middione MJ, Ennis DB, Hu P. Phase contrast MRI with flow compensation view sharing. Magnetic resonance in medicine 2015;73(2):505-13.
[39]
Chandarana H, Feng L, Block TK, Rosenkrantz AB, Lim RP, Babb JS, et al. Free-breathing
Journal Pre-proof
24
contrast-enhanced multiphase MRI of the liver using a combination of compressed sensing, parallel imaging, and golden-angle radial sampling. Investigative radiology 2013;48(1). [40]
Levine E, Daniel B, Vasanawala S, Hargreaves B, Saranathan M. 3D Cartesian MRI with compressed sensing and variable view sharing using complementary poisson-disc sampling. Magnetic Resonance in Medicine 2016;77(5):1774-85.
f
Mann LW, Higgins DM, Peters CN, Cassidy S, Hodson KK, Coombs A, et al. Accelerating MR
oo
[41]
quantitative evaluation. 2015;278(1):247-56.
Buckley DL, Shurrab AE, Cheung CM, Jones AP, Mamtora H, Kalra PA. Measurement of single
e-
[42]
pr
imaging liver steatosis measurement using combined compressed sensing and parallel imaging: a
Pr
kidney function using dynamic contrast-enhanced MRI: Comparison of two models in human
Sanford T, Colby EJLas. Effect of xylazine and ketamine on blood pressure, heart rate and
rn
[43]
al
subjects. Journal of Magnetic Resonance Imaging 2006;24(5):1117-23.
[44]
Jo u
respiratory rate in rabbits. 1980;30(3):519-23. Hermoye L, Annet L, Lemmerling P, Peeters F, Jamar F, Gianello P, et al. Calculation of the renal perfusion and glomerular filtration rate from the renal impulse response obtained with MRI. 2004;51(5):1017-25. [45]
Port RE, Knopp MV, Brix GJMRiMAOJotISfMRiM. Dynamic contrast‐enhanced MRI using Gd‐ DTPA: interindividual variability of the arterial input function and consequences for the assessment of kinetics in tumors. 2001;45(6):1030-8.
[46]
Bane O, Wagner M, Zhang JL, Dyvorne HA, Orton M, Rusinek H, et al. Assessment of renal
Journal Pre-proof
25
function using intravoxel incoherent motion diffusion‐weighted imaging and dynamic contrast‐ enhanced MRI. Journal of Magnetic Resonance Imaging 2016;44(2):317-26. [47]
Padhani AR, Hayes C, Landau S, Leach MOJNiBAIJDttD, Vivo AoMRI. Reproducibility of quantitative dynamic MRI of normal human tissues. 2002;15(2):143-53.
[48]
Cutajar M, Thomas DL, Hales PW, Banks T, Clark CA, Gordon IJEr. Comparison of ASL and DCE
oo
f
MRI for the non-invasive measurement of renal blood flow: quantification and reproducibility.
[49]
pr
2014;24(6):1300-8.
Chonko A, Osgood R, Nickel A, Ferris T, Stein JJTJoci. The measurement of nephron filtration
Barraclough MA, Guignard JP, Jones NFJJoap. Renal concentrating ability during hypertonic
Pr
[50]
e-
rate and absolute reabsorption in the proximal tubule of the rabbit kidney. 1975;56(1):232-5.
Rieg T, Richter K, Osswald H, Vallon V. Kidney function in mice: thiobutabarbital versus
rn
[51]
al
saline and mannitol diuresis in the rabbit. 1970;28(2):209-12.
[52]
Jo u
alpha-chloralose anesthesia. N-S Arch Pharmacol 2004;370(4):320-3. Fusellier M, Desfontis JC, Madec S, Gautier F, Debailleul M, Gogny M. Influence of three anesthetic protocols on glomerular filtration rate in dogs. Am J Vet Res 2007;68(8):807-11. [53]
Fleck C. Determination of the glomerular filtration rate (GFR): Methodological problems, age-dependence, consequences of various surgical interventions, and the influence of different drugs and toxic substances. Physiol Res 1999;48(4):267-79.
[54]
Artz NS, Wentland AL, Sadowski EA, Djamali A, Grist TM, Seo S, et al. Comparing Kidney Perfusion Using Noncontrast Arterial Spin Labeling MRI and Microsphere Methods in an
Journal Pre-proof
26
Interventional Swine Model. Investigative Radiology 2011;46(2):124-31. [55]
Sadick M, Attenberger U, Kraenzlin B, Kayed H, Schoenberg SO, Gretz N, et al. Two non-invasive GFR-estimation methods in rat models of polycystic kidney disease: 3.0 Tesla dynamic contrast-enhanced
MRI
and
optical
imaging.
Nephrology
Dialysis
Transplantation
2011;26(10):3101-8.
feedback-controlled
renal
constant
infusion
2012;82(3):314-20.
clearances
in
rats.
Kidney
International
e-
Zollner FG, Schock-Kusch D, Backer S, Neudecker S, Gretz N, Schad LR. Simultaneous
Pr
Measurement of Kidney Function by Dynamic Contrast Enhanced MRI and FITC-Sinistrin
rn
al
Clearance in Rats at 3 Tesla: Initial Results. Plos One 2013;8(11).
Jo u
[57]
oo
f
Schock-Kusch D, Shulhevich Y, Xie Q, Hesser J, Stsepankou D, Neudecker S, et al. Online
pr
[56]
Journal Pre-proof
27
Figure Legends
Figure 1. Diagram of VCS sampling and reconstruction strategy. a.) The k-space is divided into a
center region A and a periphery region B. b.) The region B randomly separates into three sub-regions (red, green and blue dot) and each sub-region is combined with one region A to build
f
the view-shared segments(Ki). One full k-space consists of three sub B regions (red, green and
oo
blue dot) and one center A region (yellow dot) in Ki. c.) On the basis of partial sampling in Ki,
pr
these segments are processed with a CS matrix to generate the view-shared compressed sensing measurement (𝑀𝑉𝐶𝑆 ) matrix, then, 𝑀𝑉𝐶𝑆1 , 𝑀𝑉𝐶𝑆2 , and 𝑀𝑉𝐶𝑆3 are used for consecutive image
e-
acquisitions. One CS matrix equals to three sub-B regions (red, green and blue dot) and one
Pr
center A region (yellow dot) in VCSi. d.) A diagram of view-shared CS reconstruction: each measurement acquires different k-space trajectory, before image reconstruction, the missing
al
portions in each measurement are copied from the adjacent k-space data in periphery regions.
rn
Three sequential pseudo-random partial sampled data from 𝑀𝑉𝐶𝑆1 , 𝑀𝑉𝐶𝑆2 , and 𝑀𝑉𝐶𝑆3 are shared
Jo u
to recombine a new k-space data set for image reconstruction by using the CS algorithm. For each combination, three sub-B regions (red, green and blue dot) and one center A region (dot that correspond to the color of the central arrow) in 𝑀𝑉𝐶𝑆𝑖 are added together as the new k-space data set. Figure 2. Representative VCS DCE-MR images at different points in time and signal time course. (a) Baseline image before contrast agent injection, (b) cortical peak contrast image 20s after bolus injection and (c) taken 2 min later showing increased medullar contrast. (d) The position of aortic ROIs is depicted (red circle). The concentration curves (blue line for cortex and green line
Journal Pre-proof
28
for medulla) and aortic input function (red line) of (e) left and (f) right kidney. The cortical ROI (blue circle) and medullary ROI (green circle) position are shown in (b). Figure 3. Typical GFR maps of normal kidneys and between-day reproducibility of GFR. (a), GFR maps of derived from VCS DCE-MRI in a healthy rabbit and the GFR values range from 0 (scaled to 0) up to 0.013 (scaled to 0.1). (b), Bland-Altman plot of between-day reproducibility of whole kidney GFR in healthy rabbits. Solid and dashed lines correspond to means and limits
oo
f
of agreement, respectively.
Figure 4. Typical dynamic contrast-enhanced VCS images of a rabbit with AERD after injection
pr
of contrast agent (10s, 20s, 30s). The green arrows indicate the order in which the images were
e-
obtained. The cortical lesion with reduced intensity could be observed in 10 VCS images (yellow
Pr
arrows) and the medullary lesion could be observed in 3 VCS images (red arrows). Figure 5. Typical GFR map, signal time course, renal specimen and histology result of an
al
embolized kidney. Lower values (yellow arrow) are observed in the cortex in (a) GFR map and
rn
the GFR values range from 0 (scaled to 0) up to 0.017 (scaled to 0.1). Signal time course (red
Jo u
line for cortical input function, blue line for cortex concentration curve and green line for medulla concentration curve) of the GFR reduction region (red square in (a)) is shown in (b). Scar tissues (red square) are observed in corresponding (c) renal specimen and confirmed by (d) histological finding. Figure 6. Pixel-wise GFR comparison between normal and embolized kidney. The pixel-wise GFR of lesion region is 0.0038 ± 0.005 ml/min and GRF of normal tissue is 0.0075 ± 0.0008 ml/min.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Hanjing Kong, Bin Chen, Xiaodong Zhang, Chengyan Wang, Min Yang, Li Yang, Xiaoying Wang, Jue Zhang PII:
S0730-725X(19)30078-5
DOI:
https://doi.org/10.1016/j.mri.2019.10.007
Reference:
MRI 9329
To appear in:
Magnetic Resonance Imaging
Received date:
3 February 2019
Revised date:
9 October 2019
Accepted date:
14 October 2019
Please cite this article as: H. Kong, B. Chen, X. Zhang, et al., Quantitative renal function assessment of atheroembolic renal disease using view-shared compressed sensing based dynamic-contrast enhanced MR imaging: An in vivo study, Magnetic Resonance Imaging(2018), https://doi.org/10.1016/j.mri.2019.10.007
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2018 Published by Elsevier.
Journal Pre-proof
1
Quantitative Renal Function Assessment of Atheroembolic Renal Disease using View-shared Compressed Sensing based Dynamic-contrast Enhanced MR Imaging: An in Vivo Study
Hanjing Kong1, Ph.D., Bin Chen1,2, Ph.D., Xiaodong Zhang3, Ph.D., Chengyan Wang1, Ph.D., Min Yang4, M.D. Li Yang5, M.D. Xiaoying Wang1,3, M.D. Jue Zhang1,6, Ph.D.
Academy for Advanced Interdisciplinary Studies, Peking University, 100871, Beijing, China
2.
Department of technical research and development, Instrumentation Technology and
oo
f
1.
pr
Economy Institute, 100088, Beijing, China
Department of Radiology, Peking University First Hospital, 100034, Beijing, China
4.
Department of interventional radiology and vascular surgery, Peking University First Hospital,
e-
3.
Pr
100034, Beijing, China
Renal Division, Peking University First Hospital,100034, Beijing, China
6.
College of Engineering, Peking University, 100871, Beijing, China
rn
al
5.
* Corresponding Author: Jue Zhang
Jo u
Address: 5 Yiheyuan Rd, Peking University, College of Engineering, Beijing, 100871, P.R.CHINA Tel: 8610-62755036, Fax: 8610-62753562 Email: [email protected]
Journal Pre-proof
2
Quantitative Renal Function Assessment of Atheroembolic Renal Disease using View-shared Compressed Sensing based Dynamic-contrast Enhanced MR Imaging: An in Vivo Study
Abbreviations: view-shared compressed sensing (VCS)
oo
f
Atheroembolic Renal Disease (AERD) glomerular filtration rate (GFR)
pr
Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI)
e-
digital subtraction angiography (DSA)
Pr
VCS reconstructed data (VCSD)
standard deviation (SD)
Jo u
rn
coefficient of variation (CV)
al
region of interest (ROI)
Journal Pre-proof
3
Abstract Atheroembolic renal disease (AERD) is the major cause of renal insufficiency in the elderly, and particularly, the diagnose of AERD is often delayed and even missed due to its nonspecific presentation and the sudden occurrence of an embolic event. To investigate the feasibility of the view-shared compressed sensing (VCS) based dynamic
oo
f
contrast enhanced magnetic resonance imaging (DCE-MRI) in the assessment of AERD in animal models. The reproducibility of VCS DCE-MRI based glomerular filtration rate (GFR)
pr
estimation was first evaluated using the three healthy rabbits. Animal models of unilateral AERD
e-
were then conducted. All the rabbits underwent VCS DCE-MRI and the GFR maps were
Pr
estimated by a commonly used cortical-compartment model. The whole kidney and suspicious lesion region GFR values of embolized kidneys were then compared with the corresponding
al
values of normal kidneys. Finally, the suspicious lesion regions were confirmed by the
rn
corresponding renal specimens and histological findings. The reproducibility of GFR
Jo u
measurements was analyzed using the coefficient of variation and Bland-Altman analysis. The GFR values of normal and embolized kidneys were compared using the Student t-test. Contrast-enhanced images with sufficient diagnostic quality and reduced motion artifacts are obtained at a temporal resolution of 2.5s. The Bland-Altman plot indicated close agreement between the GFR values estimated from between-day scans in healthy rabbits. Besides, there existed significant differences between the pixel-wise GFR values of normal and AERD kidneys in region-based comparison(P<0.0001). The suspicious lesions are consistent well with the renal specimen and histological findings. The preliminary animal study verified the feasibility of VCS DCE-MRI for renal function
Journal Pre-proof
4
evaluation, and the strategy could potentially provide a valuable tool to identify AERD.
Keywords: Atheroembolic Renal Disease, Glomerular Filtration Rate, DCE-MRI, View-sharing,
Jo u
rn
al
Pr
e-
pr
oo
f
Compressed Sensing
Journal Pre-proof
5
Introduction Atheroembolic Renal Disease (AERD) is part of a multisystemic disease and has attracted enhanced attention in recent years for its increasing incidence in the elderly[1]. It is a significant cause of renal loss in patients who suffer from valvular cardiopathy, aortic atheromatosis, and hypercoagulable states[2-5]. AERD is caused by showers of cholesterol emboli from the atherosclerotic aorta to many organs, and because of its variable clinical
renal damage in a significant number of patients [6-8].
oo
f
presentation, the diagnosis may be late or missing and cause more severe or less irreversible
pr
In fact, the emboli may dislodge from the aorta or other major arteries spontaneously or during
e-
or after intravascular trauma with angiographic catheters or after the use of anticoagulants and
Pr
thrombolytic agents. And the kidney is usually involved because of the proximity of the renal arteries to abdominal aorta, wherein the erosion of atheromatous plaque is most likely to occur.
rn
clinic.
al
Therefore, the diagnosis and renal function assessment of AERD is of great importance in the
Jo u
The definitive diagnosis of AERD is made by renal biopsy[9]. However, many patients are too acutely ill to tolerate renal biopsy. In recent years, several imaging methods including radioisotope renogram, renal angiogram, and contrast-enhanced CT have been applied to aid in the diagnosis of renal diseases. However, radioisotope renogram and renal angiogram are invasive and not employed as a first line study[5]. Contrast-enhanced CT has relatively poor sensitivity and patients are exposed to iodinated contrast during their diagnostic workup, placing them at high risk for contrast-associated nephropathy. MR imaging skirts these handicaps by offering anatomic and functional information simultaneously without exposure to ionizing radiation and could provide unilateral renal
Journal Pre-proof
6
functional information. Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) involving temporally resolved imaging and contrast agent injection has been established as a reliable method for estimating of single kidney glomerular filtration rate (GFR) and assessment of renal disease[10]. High spatial resolution and high temporal resolution in DCE-MRI are desirable to obtain
oo
f
accurate measurements of renal functional parameters[11, 12]. To balance the tradeoff between spatial and temporal resolution, strategies employing k-space undersampling have been proposed,
pr
including view sharing methods[13-20], Reduced-encoding Imaging by Generalized-series
e-
Reconstruction[21], k-t Broad-use Linear Acquisition Speed-up technique[22], parallel imaging
Pr
methods[23, 24], and compressed sensing methods[23, 25, 26]. Recently, Levine et al. proposed a method termed view shared compressed sensing (VCS) that combined view-sharing and
al
compressed sensing technique together and showed promising results in breast imaging[11]. The
rn
combination of compressed sensing and view sharing, can be used to trade off between temporal
Jo u
footprint and k-space sampling density. Thus this technique is particularly useful for abdomen imaging for its high temporal resolution and robustness to motion. The high temporal and spatial resolution, less invasive, and function evaluation characteristics of VCS DCE-MRI implying that it may be ideally suited for AERD detection and evaluation. Thus, in this study, we aim to assess the diagnostic performance of unilateral AERD by using a VCS DCE-MRI in a rabbit model. A commonly used cortical compartment model was served to DCE-MRI examinations of 15 rabbits and the discrimination power of the GFR of the healthy kidneys with respect to AERD kidneys was analyzed.
Journal Pre-proof
7
Materials and Methods Animals The in vivo animal experiments were approved by the Hospital Ethics Committee for Animal Research. Experiments were performed on 15 male New Zealand White rabbits (weighing 2.8– 3.3 kg). All the rabbits were free fed with standard animal food and tap water at room temperature and stopped feeding 6 hours before experiments.
oo
f
MRI
In the experiment of DCE-MRI, a dose of 0.5 ml/kg of pentobarbital sodium was injected
pr
through the marginal ear vein in the rabbit with a 24G catheter for anesthetization before
e-
scanning. All the animals were placed in a fixed device in a supine position to limit abdominal
Pr
motion during scans. Heart rate was continuously monitored during the scan by using a pulse sensor of the MR scanner.
al
DCE-MRI data were acquired using a GE 3.0T MR scanner (Signa ExciteTM; General Electric
rn
Medical Systems, Milwaukee, WI, USA) with 8-channel TORSOPA coil. Before DCE-MRI
Jo u
acquisition, data for constructing T1 map were acquired with the identical parameters using conventional GRE imaging sequence with variable flip angles method[27]. Subsequently, the coronal VCS DCE-MRI scan was performed with the following parameters: TR = 3.2 msec / TE = 1.3 msec, flip angle = 12°, acquisition matrix = 256 × 256 × 12, NEX=1, slice thickness = 4 mm, field of view = 180mm, VCS acceleration factor = 4. In dynamic contrast-enhanced experiments, five frames of non-enhanced volumes of the kidney were acquired before the bolus administration. Then, 0.1 mmol/kg body weight of Gd-DTPA (Magnevist, Bayer Schering Pharma AG, Berlin, Germany) was injected, followed by a 5.0 ml saline flush. Images were acquired immediately and totally scan time is about 8 minutes for each rabbit.
Journal Pre-proof
8
To test the between-day repeatability of the VCS method, a rescan was implemented on the second day under the same conditions for 3 healthy rabbits. GFR was calculated based on VCS-DCE images by the cortical-compartment model. The scanning position and protocols stayed the same in all the repetitive measurements. In the experiment of embolized kidneys, 12 rabbits were included. Anesthesia was induced by an intravenous injection of pentobarbital sodium before surgery. Unilateral renal embolization was
oo
f
induced by injection of a 50000-microsphere dose (acryl beads, 40-120 um in diameter) into the right renal artery. A 4F catheter was first injected into the right femoral artery and moved to the
pr
ostium of the right renal artery under the guidance of digital subtraction angiography (DSA).
e-
Then microspheres suspended in 2.0 ml of physiological saline were injected slowly. The
Pr
animals’ body temperature was maintained about 38° by using a hotplate during surgery and
experiments were implemented.
rn
VCS undersampling scheme
al
exposed to an infrared light in the box before MR scans. One week later, the DCE-MRI
Jo u
In this study, the view-shared compressed sensing sampling approach, which integrated the CS undersampling technique into the view-sharing strategy, was adopted. As demonstrated in Fig.1a, the ky-kz plane in typical k-space is divided into two independent regions: the center region 𝐴, which performs fully sampling, and the outer peripheral region B, which performs pseudo-random partial k-space sampling. First, the region 𝐵 is randomly separated into several nonoverlapping sub-regions, 𝐵𝑖 (𝑖 = 1,2 … 𝑁). Then, each sub-region 𝐵𝑖 combines a region 𝐴 to build an initial undersampled k-space, 𝐾𝑖 = 𝐴 + 𝐵𝑖 (𝑖 = 1,2 … 𝑁) for each frame, seen in Fig.1b. All the consecutive 𝐾𝑖 datasets could add up to the full k-space by sharing the data points in the 𝐵𝑖 regions. Subsequently, a two-fold CS undersampling matrix Φ𝐶𝑆 is generated according to
Journal Pre-proof
9
CS theory[28]. The CS matrix is selected uniformly at random and kept largely incoherent with some sparse transformation, used for the peripheral k-space undersampling on the basis of 𝐾𝑖 . Then, in our 3D k-space trajectory, the 𝐾𝑖 matrix is processed with the CS matrix Φ𝐶𝑆 to produce the view-shared CS measurement (MVCS), that is, 𝑀𝑉𝐶𝑆𝑖 = 𝐵𝑖 ∙ Φ𝐶𝑆 (𝑖 = 1,2 … 𝑁) and is used for view-shared CS undersampling scheme in each continuous frame, seen in Fig.1c.Here, ∙ stands for element-wise product of two matrix. The CS matrix has variable density sampling with
oo
f
denser sampling near the center region A, which matches the energy distribution in k-space. Thus, the order of acquisition is also progressively undersampled away from the center k-space
pr
according to the MVCS matrix. Here, N is set to be 3 in this study, thus, 𝑀𝑉𝐶𝑆1 , 𝑀𝑉𝐶𝑆2 , and
e-
𝑀𝑉𝐶𝑆3 are successively used for each undersampling measurement in DCE-MRI, achieving
Pr
much shorter acquisition time of each frame, and the corresponding raw data from each accelerated measurement is presented in terms of 𝐷𝑖 (𝑖 = 1,2 … 𝑁).
al
Subsequently, for image reconstruction, the view-shared compressed sensing algorithm was
rn
performed. The k-space points in MVCS are recombined by view sharing the adjacent 𝐷𝑖 (𝑖 = view-shared CS reconstructed data (𝐷𝑉𝐶𝑆 ), which equals the
Jo u
1,2 … 𝑁) data sets to generate the
CS trajectory according to the proposed acquisition scheme, seen in Fig.1d, that is, the missing portions in region 𝐵 of each frame are shared from the adjacent undersampled k-space data in sub-regions 𝐵𝑖 . Here, three subsequent partial k-space data sets in 𝐵𝑖 are recombined into the current k-space to produce the corresponding 𝐷𝑉𝐶𝑆 for further CS reconstruction,
𝐷𝑉𝐶𝑆𝑖 =
𝐷𝑖−1 ∙ 𝐵𝑖−1 + 𝐷𝑖 + 𝐷𝑖+1 ∙ 𝐵𝑖+1 The k-space samples in the new recombined 𝐷𝑉𝐶𝑆 datasets are corresponding to the points of the identified CS undersampling matrix, and CS reconstruction is then carried out according to the equation as follows: minimize λ1 ‖𝛹𝑆𝑖 ‖1 + λ2 𝑇𝑉(𝑆𝑖 ),
(1)
Journal Pre-proof
10
subject to ‖𝐹(𝑆𝑖 ) − 𝐷𝑉𝐶𝑆𝑖 ‖2 ≤ 𝜀,
(2)
where 𝑆𝑖 is the 𝑖-th image to be recovered, and Ψ denotes the linear operator that transforms from pixel representation into a sparse representation, λ1 and λ2 trade Ψ sparsifying transform with finite-differences sparsity. The typical initial values of λ1 and λ2 were both set to be 0.02 and optimized according to a previous study [29]. Ψ was set to identity transform and TV represents total variation of 𝑆. F is the Fourier transform corresponding to the given 𝑀𝑉𝐶𝑆𝑖 undersampling
oo
f
scheme, and 𝐷𝑉𝐶𝑆𝑖 is the current undersampled k-space of each frame, 𝑖 indicates the frame number.,. Additionally, ε controls the fidelity of the reconstruction which is usually set below the
pr
expected noise level[28].
e-
In this study, the data points percentage in A region were set to be 25%, and data points
Pr
percentage in 𝐵 region were set to be 25%. On the other hand, the CS matrix samples 50% of the total data points. Thus, (25%+25%)*50%=25% data points were finally sampled from the
al
full k-space of each frame, achieving a 4-fold acceleration for VCS measurement. During the
rn
image reconstruction, view-shared compressed sensing reconstruction was successfully
GFR Analysis
Jo u
performed as described in detail above.
GFR was quantitatively evaluated by using the cortical compartment model[30]. A 3×3 voxel region of interest (ROI) within the abdominal aorta distal to the branch of the renal artery was drawn to generate the arterial input function (AIF). To further reduce the noise effect related to respiratory motion, the tail of AIF was fit to a biexponential decay according to previous studies [31, 32]. The tissue signal intensity curves were generated from the DCE-MR images. T1 mapping was obtained by the acquired multi-flip angle data with a proposed method [33]. The contrast agent concentration curve in aorta 𝐶𝑎 (𝑡) and the contrast agent concentration curve in
Journal Pre-proof
11
renal cortex 𝐶𝑅𝑂𝐼 (𝑡) were calculated from the signal intensity-time curves according to the following equation: △ 𝑅1 = 𝑅1𝑝𝑜𝑠𝑡 − 𝑅1𝑝𝑟𝑒 = 𝑟1 𝐶,
(3)
Where △ 𝑅1 is the longitudinal relaxation rate change, 𝑅1𝑝𝑜𝑠𝑡 is the postcontrast longitudinal relaxation rate, 𝑅1𝑝𝑟𝑒 is the precontrast longitudinal relaxation rate, 𝑟1 is the longitudinal
f
relaxivity, and C is the concentration of the contrast agent.
oo
The cortical compartment model was adopted, the contrast agent concentration curve in the ′
pr
glomeruli 𝐶𝑎 (𝑡) could be estimated from 𝐶𝑎 (𝑡) according to the following equation: ′
1
1
e-
𝐶𝑎 (𝑡) = 𝐶𝑎 (𝑡 − 𝜏) ⊗ (𝑑) 𝑒 −𝑑
(4)
Pr
τ is a delay, d is a dispersion constant, and ⊗ is the convolution operation.
rn
al
The contrast agent concentration in the renal cortex is described as following:
′
𝑡
′
Jo u
𝐶𝑅𝑂𝐼 (𝑡) = 𝑓𝑎 𝐶𝑎 (𝑡) + 𝑘21 ∫0 𝐶𝑎 (𝜏)𝑒 −𝑘12 (𝑡−𝜏) 𝑑𝜏.
(5)
where 𝑓𝑎 is the fractional plasma volume，𝑘12 is the output of renal tubules, 𝑘21 is the GFR, The parameters 𝑘21 , 𝑘21 , 𝑓𝑎 , and 𝜏 of Eqs. (4) were fitted with the Levenberg-Marquardt nonlinear leastsquares algorithm. Considering the robustness of Levenberg-Marquardt algorithm (LMA) in the pharmacokinetic modeling of DCE-MRI, it was used to perform the non-linear least squares regression for pixel-wised analysis[34]
Journal Pre-proof
12
Histology Under sterile conditions, kidneys were taken off for pathological examinations one month after the embolization surgery. Kidneys tissues were fixed in 10% formalin and embedded in paraffin for light microscopic observation. Kidneys were sectioned into 2-mm slides and stained with hematoxylin-eosin-saffron. A pathologist specializing in kidney diseases reviewed histological findings. The pathologist was blinded to the imaging findings and the results of pathology and
oo
f
renal specimen were considered as a reference standard for assessing the kidney injury. Data Analysis
pr
DCE-MR images with the best visualization were selected by two radiologists for analysis. For
e-
the reproducibility study, the kidney was segmented by the radiologists and the whole kidney
Pr
GFR was calculated. For the AERD rabbit’s analysis, the lesion ROIs were carefully drawn by a radiologist with thirteen years’ experience and corresponding normal regions were also selected
al
on contra-lateral normal kidney images. Then, the lesion size and the mean value of pixel-wise
rn
GFR in each ROI were calculated for further comparison. All calculations were performed by
Jo u
MATLAB (MathWorks, Natick, MA). Statistical Analysis
Results were represented in mean ± standard deviation (SD). Between-day reproducibility of the GFR measurement was evaluated by coefficient of variation (CV), which was calculated as the standard deviation divided by the mean of the GFR from two scans. Also, Bland–Altman analysis of the scan and rescan (between-day) whole kidney GFR was implemented to validate the reproducibility of VCS method. Student t-test was performed to compare the value of pixel-wise GFR between normal tissue and contra-lateral embolized lesion. P value < 0.05 was considered statistically significant.
Journal Pre-proof
13
Results A representative VCS DCE-MR image and contrast agent uptake curve in normal kidneys are shown in Fig 2. The detailed spatial resolution enables display of the renal parenchyma and renal artery without obvious artifacts. Cortical ROIs were manually drawn (blue and green circle in Fig 2b) and slice that with a branch of the renal artery (red circle in Fig 2d) was selected for aortic ROI drawing. Representative concentration curves at a temporal resolution of 2.5s derived
oo
f
from corresponding ROIs are shown in Fig 2 e,f.
Typical GFR maps of a healthy rabbit derived from VCS DCE-MRI are showed in Fig 3a. The
pr
maps are generally smooth and differentiate between central and subcortical regions, but isolated
e-
pixels or smaller clusters with deviating values can be observed in all maps. Reproducibility
Pr
results of VCS DCE-MRI based GFR are summarized in Fig 3b. The between-day CV of whole kidney GFR measurements of VCS methods was 8.96%. Bland-Altman plot of VCS DCE-MRI
al
derived whole kidney GFR in healthy rabbits are shown in Fig 3b, and all the whole kidney GFR
rn
values lay within the limit of agreement (d-1.96s and d+1.96s).
Jo u
Fig 4 shows the representative twelve frames VCS DCE-MRI of an embolized kidney. As pointed by yellow arrows, the cortical lesion was seen in 10 VCS images and medullary lesion could be observed in 3 VCS images (Red arrows). Dark areas in DCE-MR images showed close spatial correspondence to the embolized regions in the renal specimen (Fig 5c) and is confirmed by histology result (Fig 5d). Representative GFR map and corresponding concentration curves derived from cortical and medullary ROIs are shown in Fig 5 a,b. GFR reduction regions on the GFR map (Fig 5a) matched closely to the heavily embolized regions in the renal specimen (Fig 5c), but the lesion region in GFR map showed a certain degree of underestimation compared with specimen. In
Journal Pre-proof
14
histological findings (Fig 5d), the glomeruli showed ischemic and wrinkled features with thickening change of Bowman’s capsule, and necrosis of the renal tubular epithelial cells was observed, the basement membrane was exposed. Meanwhile, the brush border of some tubular epithelial cells fell off and the tubular epithelial cells became flat and tubular lumen expanded. A total of 17 lesions were found in all rabbits by the renal specimen and confirmed by histological findings. 14 lesions were found by VCS DCE-MRI and the lesion size is 0.14 ± 0.07
oo
f
cm2. For the GFR comparison, the pixel-wise GFR of the embolized lesion was significantly lower than GFR of normal tissue (0.0038 ± 0.005 ml/min vs 0.0075 ± 0.0008 ml/min, P<0.0001)
Jo u
rn
al
Pr
e-
pr
in Fig 6.
Journal Pre-proof
15
Discussion In this in vivo study, we explored the ability of VCS DCE-MRI to detect the impaired renal function in AERD kidney of rabbits. The results confirmed that VCS DCE-MRI could obtain not only diagnostic image quality with reduced motion artifacts and high temporal resolution but reliable GFR estimation with high reproducibility. In addition, compared with normal kidneys, the significant lower GFR values in AERD kidneys corroborated the hypothesis that VCS
oo
f
DCE-MRI was capable to distinguish between AERD and normal kidneys, and thus provide a valuable tool for detection patients with AERD.
pr
AERD involves the acute occlusion of the renal artery or its branches because of an embolus
e-
from a distant source[5]. In this study, we use microsphere that injected directly into the renal
Pr
artery under the guidance of DSA to mimic the embolism process[35]. Also, the catheter was first injected into the right femoral artery and moved to the ostium of the right renal artery, which
rn
and reliable.
al
greatly reduces the surgery trauma, and make the AERD model in rabbits more effective, safe
Jo u
Once the emboli enter the blood circulation, it will stay in a small artery about 150-200 μm in diameter[36] and further causing small artery occlusion and inflammation. The focal feature of AERD, placing greater demand on the spatial resolution of imaging modality. In this study, the spatial resolution of VCS DCE-MRI is 0.7mm*0.7mm. The high resolution and reduced partial volume effect enabled precise visualization of lesion and detection of a majority of lesions (14/17). For GFR quantification, besides spatial resolution, the temporal resolution also plays an important role. Theoretically, a higher temporal resolution yields a higher quality of AIF measurement, and therefore improve GFR estimation. In recent years, view-sharing and
Journal Pre-proof
16
compressed sensing strategies have already exhibited potential contribution in DCE-MRI to improve the temporal resolution[37-39]. By further integrating CS into view-shared peripheral k-space in each 3D volume, the temporal resolution of this VCS acquisition can be significantly improved according to the k-space sampling percentage using the VCS matrix in k-space. We recommend using VCS strategy instead of TWIST or CS based method alone for two reasons.
oo
f
First, all view sharing schemes would introduce temporal blurring of the high spatial frequencies,
pr
making them sensitive to both motion and signal changes due to contrast enhancement[40].
e-
Although it could achieve high resolution, it may also be unstable for clinical abdomen DCE MR imaging. Second, although CS sampling schemes showed promising results in many studies, they
Pr
suffer from image quality degradation under low k-space sampling density[41]. The combination
rn
k-space sampling density.
al
of compressed sensing and view sharing, can be used to trade off between temporal footprint and
Jo u
Theoretically, the VCS strategy could achieve higher resolution compared with CS or view sharing strategy alone. In this study, considering the image quality, we are conservative in choosing the acceleration factor. In this study, a temporal resolution of 2.5s was achieved by VCS strategy and was adequate for renal DCE-MRI. By utilizing a well-established cortical compartment model[30, 42], the GFR values estimated of normal kidneys with high reproducibility was close to literature reports[27, 30], which suggested that VCS was able to provide reliable GFR measurements. The reproducibility is evaluated by CV and Bland-Altman plot. Due to the complexity of DCE
Journal Pre-proof
17
experiments, there are many factors that may affect the consistency of experimental results, including health and metabolism condition, effects of anesthesia on blood pressure and renal perfusion during MRI[43, 44], ROI selection for arterial input function[45]. Many impactful Study reported the CV value of GFR reproducibility in the range of 19.4 to 28 (CV=22.4%[46], CV=19.4%-26.4%[47], CV=28%[48]) and the width of the limits of agreement is within clinical
oo
f
acceptability[49, 50]. Based on these results, we believe that VCS is a relatively stable method for GFR estimation.
pr
In our study, GFR of AERD lesion was significantly lower than that of normal tissue. The
e-
changes in embolized kidneys are also confirmed by renal specimen and histology, which shows
Pr
typical hallmarks of AERD. Altogether, these support our conclusion that VCS MRI is feasible to distinguish AERD from normal kidney, and the derived GFR can be served as a biomarker to
al
differentiate kidneys with AERD from healthy controls.
rn
There are still some limitations. First, only animal experiments were conducted in this study.
Jo u
Additional evaluation of the VCS DCE-MRI is desirable in patients with confirmed pathologies to confirm the clinical relevance of the sequence ultimately. Second, we chose to examine only VCS DCE-MRI in renal imaging for its superior performance in abnormal imaging, and other extended or alternative strategies would also be deserved to further explore. For example, parallel imaging is widely used in clinical scanners and can be further combined with VCS to provide a more practical imaging method for clinic using. Third, we only followed the animals at one week after surgery, it would be interesting to assess at other times to investigate the disease development process. Fourth, the thickness in this study is 4mm, which may cause partial volume effect, so high isotropic‐resolution imaging should be used for further study.
Journal Pre-proof
18
No accepted gold standard for GFR measurements such as creatine or inulin clearance was employed. Though these indicators are probably more sensitive, they are highly invasive and might bias the results by factors like additional anaesthesia[51-53] and repeated measurements[54, 55]. Moreover, imaging the animals by such gold standard only allows for estimation of global GFR[54, 55], and resulting in a potential decrease of sensitivity in the detection and accuracy of the unilateral injury. Besides, recent studies showed that MRI based
oo
f
GFR measurement correlated well to GFR measured by FITC-sinistrin clearance which could be considered as gold standard for measuring the GFR[56, 57]. We believed that the above
pr
limitation might be minor.
e-
In conclusion, this preliminary animal study demonstrated the feasibility of using the VCS
Pr
DCE-MRI in quantitative renal function evaluation, with satisfactory reproducibility. The efficacy of the VCS strategy was carefully verified by histology findings. To the best of our
al
knowledge, this is the first study to perform the function evaluation of AERD using DCE-MRI.
rn
This VCS DCE-MRI could potentially provide a valuable tool to identify AERD and monitor the
Jo u
success of therapeutic methods in clinical practice.
Journal Pre-proof
19
References [1]
Scolari F, Ravani P. Atheroembolic renal disease. The Lancet 2010;375(9726):1650-60.
[2]
Fort J. Renal artery embolism. Renal Failure 1997;19(6):R7-R8.
[3]
Nicholas GG, Demuth WE. Treatment of Renal-Artery Embolism. Arch Surg-Chicago
[4]
oo
f
1984;119(3):278-81. Korzets Z, Plotkin E, Bernheim J, Zissin R. The clinical spectrum of acute renal infarction. Isr Med
Gen Intern Med 2008;23(5):644-7.
Lessman RK, Johnson SF, Coburn JW, Kaufman JJ. Renal-Artery Embolism - Clinical Features
al
[6]
e-
Kansal S, Feldman M, Cooksey S, Patel S. Renal artery embolism - A case report and review. J
Pr
[5]
pr
Assoc J 2002;4(10):781-4.
Mugge A, Gulba DC, Frei U, Wagenbreth I, Grote R, Daniel WG, et al. Renal-Artery Embolism -
Jo u
[7]
rn
and Long-Term Follow-up of 17 Cases. Ann Intern Med 1978;89(4):477-82.
Thrombolysis
with
Recombinant
Tissue-Type
Plasminogen-Activator.
J
Intern
Med
1990;228(3):279-86. [8]
Modi KS, Rao VK. Atheroembolic renal disease. J Am Soc Nephrol 2001;12(8):1781-7.
[9]
Wilson DM, Salazer TL, Farkouh ME. Eosinophiluria in atheroembolic renal disease. The American journal of medicine 1991;91(2):186-9.
[10]
Hahn OM, Yang C, Medved M, Karczmar G, Kistner E, Karrison T, et al. Dynamic Contrast-Enhanced Magnetic Resonance Imaging Pharmacodynamic Biomarker Study of Sorafenib in Metastatic Renal Carcinoma. Journal of Clinical Oncology 2008;26(28):4572-8.
Journal Pre-proof
[11]
20
Levine E, Daniel B, Vasanawala S, Hargreaves B, Saranathan M. 3D Cartesian MRI with compressed sensing and variable view sharing using complementary poisson‐disc sampling. Magnetic resonance in medicine 2017;77(5):1774-85.
[12]
Heisen M, Fan XB, Buurman J, van Riel NAW, Karczmar GS, Romeny BMT. The Influence of Temporal Resolution in Determining Pharmacokinetic Parameters From DCE-MRI Data. Magnetic
oo
Herrmann KH, Baltzer PA, Dietzel M, Krumbein I, Geppert C, Kaiser WA, et al. Resolving arterial
pr
[13]
f
Resonance in Medicine 2010;63(3):811-6.
phase and temporal enhancement characteristics in DCE MRM at high spatial resolution with
Le Y, Kroeker R, Kipfer HD, Lin C. Development and evaluation of TWIST Dixon for dynamic
Pr
[14]
e-
TWIST acquisition. Journal of Magnetic Resonance Imaging 2011;34(4):973-82.
al
contrast‐enhanced (DCE) MRI with improved acquisition efficiency and fat suppression. Journal of
Song T, Laine AF, Chen Q, Rusinek H, Bokacheva L, Lim RP, et al. Optimal k‐space sampling for
Jo u
[15]
rn
Magnetic Resonance Imaging 2012;36(2):483-91.
dynamic contrast‐enhanced MRI with an application to MR renography. Magnetic resonance in medicine 2009;61(5):1242-8. [16]
Lim R, Shapiro M, Wang E, Law M, Babb J, Rueff L, et al. 3D time-resolved MR angiography (MRA) of the carotid arteries with time-resolved imaging with stochastic trajectories: comparison with 3D contrast-enhanced Bolus-Chase MRA and 3D time-of-flight MRA. American Journal of Neuroradiology 2008;29(10):1847-54.
[17]
Riederer SJ, Haider CR, Borisch EA, Weavers PT, Young PM. Recent advances in 3D
Journal Pre-proof
21
time-resolved contrast-enhanced MR angiography. Journal of Magnetic Resonance Imaging 2015;42(1):3-22. [18]
Haider CR, Hu HH, Campeau NG, John Huston I, Riederer SJ. 3D High Temporal and Spatial Resolution Contrast-Enhanced MR Angiography of the Whole Brain. Magnetic Resonance in Medicine 2008;60(3):749.
f
Saranathan M, Dan WR, Hargreaves BA, Clarke SE, Vasanawala SS. DIfferential Subsampling
oo
[19]
pr
with Cartesian Ordering (DISCO): a High Spatio-temporal Resolution Dixon Imaging Sequence for Multiphasic Contrast Enhanced Abdominal Imaging. Journal of Magnetic Resonance Imaging
Zimmer F, Zöllner FG, Hoeger S, Klotz S, Tsagogiorgas C, Krämer BK, et al. Quantitative renal
Pr
[20]
e-
2012;35(6):1484–92.
al
perfusion measurements in a rat model of acute kidney injury at 3T: testing inter-and
Webb AG, Liang ZP, Magin RL, Lauterbur PC. Applications of reduced‐encoding MR imaging with
Jo u
[21]
rn
intramethodical significance of ASL and DCE-MRI. 2013;8(1):e53849.
generalized ‐ series
reconstruction
(RIGR).
Journal
of
Magnetic
Resonance
Imaging
1993;3(6):925-8. [22]
Tsao J, Boesiger P, Pruessmann KP. k‐t BLAST and k‐t SENSE: Dynamic MRI with high frame rate exploiting spatiotemporal correlations. Magnetic resonance in medicine 2003;50(5):1031-42.
[23]
Rosenkrantz AB, Geppert C, Grimm R, Block TK, Glielmi C, Feng L, et al. Dynamic contrast‐ enhanced MRI of the prostate with high spatiotemporal resolution using compressed sensing, parallel imaging, and continuous golden‐angle radial sampling: Preliminary experience. Journal of
Journal Pre-proof
22
Magnetic Resonance Imaging 2015;41(5):1365-73. [24]
Feng L, Grimm R, Block KT, Chandarana H, Kim S, Xu J, et al. Golden‐angle radial sparse parallel MRI: Combination of compressed sensing, parallel imaging, and golden‐angle radial sampling for fast and flexible dynamic volumetric MRI. Magnetic resonance in medicine 2014;72(3):707-17.
[25]
Zhang T, Chowdhury S, Lustig M, Barth RA, Alley MT, Grafendorfer T, et al. Clinical performance
oo
f
of contrast enhanced abdominal pediatric MRI with fast combined parallel imaging compressed
[26]
pr
sensing reconstruction. Journal of Magnetic Resonance Imaging 2014;40(1):13-25. Zhang T, Cheng JY, Potnick AG, Barth RA, Alley MT, Uecker M, et al. Fast pediatric 3D free‐
e-
breathing abdominal dynamic contrast enhanced MRI with high spatiotemporal resolution. Journal
Chen B, Zhang Y, Song X, Wang X, Zhang J, Fang J. Quantitative estimation of renal function with
al
[27]
Pr
of Magnetic Resonance Imaging 2015;41(2):460-73.
[28]
Jo u
2014;9(8):e105087.
rn
dynamic contrast-enhanced MRI using a modified two-compartment model. PloS one
Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magnetic resonance in medicine 2007;58(6):1182-95.
[29]
Li B, Dong L, Chen B, Ji S, Cai W, Wang Y, et al. Turbo fast three‐dimensional carotid artery black‐blood MRI by combining three‐dimensional MERGE sequence with compressed sensing. Magnetic resonance in medicine 2013;70(5):1347-52.
[30]
Annet L, Hermoye L, Peeters F, Jamar F, Dehoux J-P, Van Beers BE. Glomerular filtration rate: Assessment with dynamic contrast-enhanced MRI and a cortical-compartment model in the rabbit
Journal Pre-proof
23
kidney. Journal of Magnetic Resonance Imaging 2004;20(5):843-9. [31]
Tofts PS, Kermode AG. Measurement of the blood‐brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magnetic resonance in medicine 1991;17(2):357-67.
[32]
Misselwitz B, Schmitt-Willich H, Ebert W, Frenzel T, Weinmann H-J. Pharmacokinetics of
[33]
pr
in Physics, Biology and Medicine 2001;12(2):128-34.
oo
f
Gadomer-17, a new dendritic magnetic resonance contrast agent. Magnetic Resonance Materials
Cheng HLM, Wright GA. Rapid high‐resolution T1 mapping by variable flip angles: Accurate and
e-
precise measurements in the presence of radiofrequency field inhomogeneity. Magnetic
Marquardt DWJJotsfI, Mathematics A. An algorithm for least-squares estimation of nonlinear
al
[34]
Pr
resonance in medicine 2006;55(3):566-74.
rn
parameters. 1963;11(2):431-41.
Kassirer JP. Atheroembolic renal disease. New England Journal of Medicine 1969;280(15):812-8.
[36]
Eliot RS, Edwards JE, Kanjuh VI. Atheromatous Embolism. Circulation 1964;30(4):611-&.
[37]
Rapacchi S, Han F, Natsuaki Y, Kroeker R, Plotnik A, Lehrman E, et al. High spatial and temporal
Jo u
[35]
resolution dynamic contrast ‐ enhanced magnetic resonance angiography using compressed sensing with magnitude image subtraction. Magnetic resonance in medicine 2014;71(5):1771-83. [38]
Wang D, Shao J, Rapacchi S, Middione MJ, Ennis DB, Hu P. Phase contrast MRI with flow compensation view sharing. Magnetic resonance in medicine 2015;73(2):505-13.
[39]
Chandarana H, Feng L, Block TK, Rosenkrantz AB, Lim RP, Babb JS, et al. Free-breathing
Journal Pre-proof
24
contrast-enhanced multiphase MRI of the liver using a combination of compressed sensing, parallel imaging, and golden-angle radial sampling. Investigative radiology 2013;48(1). [40]
Levine E, Daniel B, Vasanawala S, Hargreaves B, Saranathan M. 3D Cartesian MRI with compressed sensing and variable view sharing using complementary poisson-disc sampling. Magnetic Resonance in Medicine 2016;77(5):1774-85.
f
Mann LW, Higgins DM, Peters CN, Cassidy S, Hodson KK, Coombs A, et al. Accelerating MR
oo
[41]
quantitative evaluation. 2015;278(1):247-56.
Buckley DL, Shurrab AE, Cheung CM, Jones AP, Mamtora H, Kalra PA. Measurement of single
e-
[42]
pr
imaging liver steatosis measurement using combined compressed sensing and parallel imaging: a
Pr
kidney function using dynamic contrast-enhanced MRI: Comparison of two models in human
Sanford T, Colby EJLas. Effect of xylazine and ketamine on blood pressure, heart rate and
rn
[43]
al
subjects. Journal of Magnetic Resonance Imaging 2006;24(5):1117-23.
[44]
Jo u
respiratory rate in rabbits. 1980;30(3):519-23. Hermoye L, Annet L, Lemmerling P, Peeters F, Jamar F, Gianello P, et al. Calculation of the renal perfusion and glomerular filtration rate from the renal impulse response obtained with MRI. 2004;51(5):1017-25. [45]
Port RE, Knopp MV, Brix GJMRiMAOJotISfMRiM. Dynamic contrast‐enhanced MRI using Gd‐ DTPA: interindividual variability of the arterial input function and consequences for the assessment of kinetics in tumors. 2001;45(6):1030-8.
[46]
Bane O, Wagner M, Zhang JL, Dyvorne HA, Orton M, Rusinek H, et al. Assessment of renal
Journal Pre-proof
25
function using intravoxel incoherent motion diffusion‐weighted imaging and dynamic contrast‐ enhanced MRI. Journal of Magnetic Resonance Imaging 2016;44(2):317-26. [47]
Padhani AR, Hayes C, Landau S, Leach MOJNiBAIJDttD, Vivo AoMRI. Reproducibility of quantitative dynamic MRI of normal human tissues. 2002;15(2):143-53.
[48]
Cutajar M, Thomas DL, Hales PW, Banks T, Clark CA, Gordon IJEr. Comparison of ASL and DCE
oo
f
MRI for the non-invasive measurement of renal blood flow: quantification and reproducibility.
[49]
pr
2014;24(6):1300-8.
Chonko A, Osgood R, Nickel A, Ferris T, Stein JJTJoci. The measurement of nephron filtration
Barraclough MA, Guignard JP, Jones NFJJoap. Renal concentrating ability during hypertonic
Pr
[50]
e-
rate and absolute reabsorption in the proximal tubule of the rabbit kidney. 1975;56(1):232-5.
Rieg T, Richter K, Osswald H, Vallon V. Kidney function in mice: thiobutabarbital versus
rn
[51]
al
saline and mannitol diuresis in the rabbit. 1970;28(2):209-12.
[52]
Jo u
alpha-chloralose anesthesia. N-S Arch Pharmacol 2004;370(4):320-3. Fusellier M, Desfontis JC, Madec S, Gautier F, Debailleul M, Gogny M. Influence of three anesthetic protocols on glomerular filtration rate in dogs. Am J Vet Res 2007;68(8):807-11. [53]
Fleck C. Determination of the glomerular filtration rate (GFR): Methodological problems, age-dependence, consequences of various surgical interventions, and the influence of different drugs and toxic substances. Physiol Res 1999;48(4):267-79.
[54]
Artz NS, Wentland AL, Sadowski EA, Djamali A, Grist TM, Seo S, et al. Comparing Kidney Perfusion Using Noncontrast Arterial Spin Labeling MRI and Microsphere Methods in an
Journal Pre-proof
26
Interventional Swine Model. Investigative Radiology 2011;46(2):124-31. [55]
Sadick M, Attenberger U, Kraenzlin B, Kayed H, Schoenberg SO, Gretz N, et al. Two non-invasive GFR-estimation methods in rat models of polycystic kidney disease: 3.0 Tesla dynamic contrast-enhanced
MRI
and
optical
imaging.
Nephrology
Dialysis
Transplantation
2011;26(10):3101-8.
feedback-controlled
renal
constant
infusion
2012;82(3):314-20.
clearances
in
rats.
Kidney
International
e-
Zollner FG, Schock-Kusch D, Backer S, Neudecker S, Gretz N, Schad LR. Simultaneous
Pr
Measurement of Kidney Function by Dynamic Contrast Enhanced MRI and FITC-Sinistrin
rn
al
Clearance in Rats at 3 Tesla: Initial Results. Plos One 2013;8(11).
Jo u
[57]
oo
f
Schock-Kusch D, Shulhevich Y, Xie Q, Hesser J, Stsepankou D, Neudecker S, et al. Online
pr
[56]
Journal Pre-proof
27
Figure Legends
Figure 1. Diagram of VCS sampling and reconstruction strategy. a.) The k-space is divided into a
center region A and a periphery region B. b.) The region B randomly separates into three sub-regions (red, green and blue dot) and each sub-region is combined with one region A to build
f
the view-shared segments(Ki). One full k-space consists of three sub B regions (red, green and
oo
blue dot) and one center A region (yellow dot) in Ki. c.) On the basis of partial sampling in Ki,
pr
these segments are processed with a CS matrix to generate the view-shared compressed sensing measurement (𝑀𝑉𝐶𝑆 ) matrix, then, 𝑀𝑉𝐶𝑆1 , 𝑀𝑉𝐶𝑆2 , and 𝑀𝑉𝐶𝑆3 are used for consecutive image
e-
acquisitions. One CS matrix equals to three sub-B regions (red, green and blue dot) and one
Pr
center A region (yellow dot) in VCSi. d.) A diagram of view-shared CS reconstruction: each measurement acquires different k-space trajectory, before image reconstruction, the missing
al
portions in each measurement are copied from the adjacent k-space data in periphery regions.
rn
Three sequential pseudo-random partial sampled data from 𝑀𝑉𝐶𝑆1 , 𝑀𝑉𝐶𝑆2 , and 𝑀𝑉𝐶𝑆3 are shared
Jo u
to recombine a new k-space data set for image reconstruction by using the CS algorithm. For each combination, three sub-B regions (red, green and blue dot) and one center A region (dot that correspond to the color of the central arrow) in 𝑀𝑉𝐶𝑆𝑖 are added together as the new k-space data set. Figure 2. Representative VCS DCE-MR images at different points in time and signal time course. (a) Baseline image before contrast agent injection, (b) cortical peak contrast image 20s after bolus injection and (c) taken 2 min later showing increased medullar contrast. (d) The position of aortic ROIs is depicted (red circle). The concentration curves (blue line for cortex and green line
Journal Pre-proof
28
for medulla) and aortic input function (red line) of (e) left and (f) right kidney. The cortical ROI (blue circle) and medullary ROI (green circle) position are shown in (b). Figure 3. Typical GFR maps of normal kidneys and between-day reproducibility of GFR. (a), GFR maps of derived from VCS DCE-MRI in a healthy rabbit and the GFR values range from 0 (scaled to 0) up to 0.013 (scaled to 0.1). (b), Bland-Altman plot of between-day reproducibility of whole kidney GFR in healthy rabbits. Solid and dashed lines correspond to means and limits
oo
f
of agreement, respectively.
Figure 4. Typical dynamic contrast-enhanced VCS images of a rabbit with AERD after injection
pr
of contrast agent (10s, 20s, 30s). The green arrows indicate the order in which the images were
e-
obtained. The cortical lesion with reduced intensity could be observed in 10 VCS images (yellow
Pr
arrows) and the medullary lesion could be observed in 3 VCS images (red arrows). Figure 5. Typical GFR map, signal time course, renal specimen and histology result of an
al
embolized kidney. Lower values (yellow arrow) are observed in the cortex in (a) GFR map and
rn
the GFR values range from 0 (scaled to 0) up to 0.017 (scaled to 0.1). Signal time course (red
Jo u
line for cortical input function, blue line for cortex concentration curve and green line for medulla concentration curve) of the GFR reduction region (red square in (a)) is shown in (b). Scar tissues (red square) are observed in corresponding (c) renal specimen and confirmed by (d) histological finding. Figure 6. Pixel-wise GFR comparison between normal and embolized kidney. The pixel-wise GFR of lesion region is 0.0038 ± 0.005 ml/min and GRF of normal tissue is 0.0075 ± 0.0008 ml/min.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6