In vivo, label-free, three-dimensional quantitative imaging of kidney microcirculation using Doppler optical coherence tomography (DOCT)

Abstract

Doppler optical coherence tomography (DOCT) is a functional extension of optical coherence tomography (OCT) and is currently being employed in several clinical arenas to quantify blood flow in vivo. In this study, the objective was to investigate the feasibility of DOCT to image kidney microcirculation, specifically, glomerular blood flow. DOCT is able to capture 3D data sets consisting of a series of cross-sectional images in real time, which enables label-free and non-destructive quantification of glomerular blood flow. The kidneys of adult, male Munich-Wistar rats were exposed through laparotomy procedure after being anesthetized. Following exposure of the kidney beneath the DOCT microscope, glomerular blood flow was observed. The effects of acute mannitol and angiotensin II infusion were also observed. Glomerular blood flow was quantified for the induced physiological states and compared with baseline measurements. Glomerular volume, cumulative Doppler volume, and Doppler flow range parameters were computed from 3D OCT/DOCT data sets. Glomerular size was determined from OCT, and DOCT readily revealed glomerular blood flow. After infusion of mannitol, a significant increase in blood flow was observed and quantified, and following infusion of angiontensin II, a significant decrease in blood flow was observed and quantified. Also, blood flow histograms were produced to illustrate differences in blood flow rate and blood volume among the induced physiological states. We demonstrated 3D DOCT imaging of rat kidney microcirculation in the glomerulus in vivo. Dynamic changes in blood flow were detected under altered physiological conditions demonstrating the real-time imaging capability of DOCT. This method holds promise to allow non-invasive imaging of kidney blood flow for transplant graft evaluation or monitoring of altered renal hemodynamics related to disease progression.

Intrarenal hemodynamic abnormalities are thought to be a primary factor associated with the onset and progression of acute injury [1], and other various nephropathies like diabetic nephropathy [2, 3], and focal segmental glomerulosclerosis (FSGS) [4]. Real-time assessment of renal morphological and hemodynamic changes could help to evaluate the kidney condition and offer valuable information to predict the prognosis of injury or disease lending to the development of patient-specific management strategies. Currently, though, there is no sensitive and objective tool for clinical monitoring of renal microcirculatory changes. The ability to monitor alterations in renal microcirculation due to vascular or glomerular disease may improve diagnostic and therapeutic interventions for renal health care.

Renal blood flow (RBF) has been monitored using a number of different imaging modalities including positron emission tomography (PET) [5–10], and magnetic resonance angiography (MRA) [11–14], Doppler ultrasound (US) [15–18] and contrast-enhanced (CE) US [19, 20]. While these techniques allow for non-invasive, wide field-of-view (FOV) imaging, they do not have sufficient resolution to detect changes in renal microcirculation (glomerular). Regarding kidney transplantation, studies have suggested that blood perfusion within glomerular capillaries may be correlated with intermediate and long-term graft function [21, 22]. Therefore, immediate detection of microcirculatory changes could provide decisive criteria for therapeutic interventions promoting graft salvage and long-term function [22–24].

Optical imaging techniques, though, that have higher resolutions and greater sensitivities could be a more feasible method for monitoring and evaluating microcirculatory changes, especially in an intra-operative setting. Several optical imaging techniques have been employed to study RBF. Studies using confocal [25–27] and multi-photon microscopy [28–30] have demonstrated the ability to image kidney microstructure and function (blood flow and filtration rate) on animal models, but the penetration depth has been limited to several hundreds of microns [31], and they require the need to administer exogenous contrast agents to determine flow. Therefore, using these imaging modalities for human studies poses certain challenges due to the capsule surrounding the human kidney, which can be several hundred microns thick, and the need to inject contrast agents into the patient.

Optical coherence tomography (OCT) [32] and its functional extension Doppler OCT (DOCT) [33] are emerging imaging technologies that have the capacity to provide real-time images (i.e., immediate imaging) of tissue in a non-invasive fashion (i.e., non-destructively) with high-resolution near that of conventional histopathological images. Previous studies by us have demonstrated the ability of OCT to resolve the renal corpuscle and uriniferous tubules [34] and evaluate the real-time morphological changes in these structures associated with ischemia-reperfusion injury in vivo in an animal model [35]. Also, we have previously demonstrated that OCT can penetrate the capsule surrounding human kidneys thereby enabling the characterization of renal tubules, glomeruli, and cortical blood vessels in human kidneys [36, 37]. OCT represents a laser echo microscopy and has the potential to become a powerful tool for functional optical renal biopsy in the near future [38].

High-speed, Fourier-domain DOCT can be used to visualize blood flow non-invasively by measuring Doppler frequency shifts in the OCT interference signal caused by moving scatterers that are label-free, such as red blood cells [39]. DOCT has been used to image and quantify blood flow in vivo for multiple clinical applications including retina [40–43], skin [44–46], and gastrointestinal tract [47], among others. DOCT imaging of kidney blood flow has also demonstrated translational potential.

In the present study, our objective was to demonstrate the feasibility of using Doppler OCT (DOCT) to observe microcirculation within the living kidney in real time and quantitatively compare physiologically induced changes.

Materials and Methods

Optical Coherence Tomography (OCT) and Doppler OCT

A high-speed, high-resolution OCT system was used in this study. The details of the OCT system have been previously described [36]. Briefly, a Fourier-domain OCT system consisting of a swept-source laser with 100 nm bandwidth at 1310 nm center wavelength yielding an axial resolution of ~10 μm in tissue was used. A 10x objective (Olympus Plan N, NA=0.25) was used in the sample arm to achieve a lateral resolution of ~2–3 μm determined by USAF resolution target. The laser source operated at a sweep rate of 16 kHz allowing a series of 2D cross-sectional images to be captured in real time to form a 3D data set (Voxel size: 1024 [X] by 256 [Y] by 512 [Z]; Dimension: 0.280 mm [X] by 0.325 mm [Y] by 1.90 mm [Z]; acquisition time <20 sec). OCT fringe data was acquired to obtain both amplitude and phase information after Fourier transform, and Doppler OCT signals were computed using standard phased-resolved algorithm [39].

Animal Model and Preparation

The animal protocol has been approved by the committees on animal care and use in both the University of Maryland and Georgetown University. Male Munich-Wistar rats (n=3, ~400 g) were used in this study. This strain of rat has numerous superficial glomeruli in the outer cortex that are accessible for observation [48]. Rats were regularly fed normal rat chow with free access to water at all times. During the study, rats were anesthetized by intraperitoneal injection of pentobarbital sodium (100 mg/kg body weight) and monitored routinely by tactile stimulation with supplemental anesthetic added as needed to maintain initial depth of anesthesia. After being secured on a portable surgical apparatus, the left kidney was exposed through laparotomy of the left flank region and the kidney was securely placed in a lucite holder. The right femoral vein was then cannulated with polyethylene tubing for administration of pharmacologic agents. Rats were then placed beneath the DOCT microscope for in vivo imaging. To induce alterations in renal blood flow, 0.1–0.2 ml mannitol (250 mg/ml), which increases renal blood flow [49–51], and 200μg/kg body weight angiotensin II [52], which decreases renal blood flow [53–56], were administered by a bolus injection into the femoral vein. Three injections of mannitol at 45 min intervals were performed followed by DOCT imaging after each injection and then a final injection of angiotensin II was given 10 min after the third mannitol injection followed by DOCT imaging. DOCT imaging was also performed approximately 15–20 min after each infusion of mannitol when the physiologic effects of the drug were diminishing. Data taken during this period is labeled “recover” to distinguish it from the data taken immediately after each mannitol injection. At each stage, multiple glomeruli (n ≥ 3) from different locations were imaged. Thus, the same glomeruli were not re-imaged at each stage.

OCT/DOCT Image Processing and Quantification

Glomeruli were located by scanning the surface of the cortex, and 3D data sets, including the fringe data for Doppler analysis, were acquired for each glomerulus (total n=44). OCT intensity images as well as DOCT images were computed from the fringe data for each frame in the 3D data sets [57]. Quantification of single glomerular microcirculation parameters was performed offline using a custom image analysis program written in Matlab (Mathworks Inc., Natick, MA, USA).

The glomerular volume (GV) was computed from the OCT intensity image according to the equation

$$\text{GV}=\frac{2}{3}{\text{S}}_{\text{A}}\text{h}$$ (1)

where S_A is the surface area through the mid-section of the glomerulus and h is the height of the glomerulus. The surface area and height of each glomerulus was manually selected by encircling the perimeter of the glomerulus in cross-sectional and en face images passing through the center of the glomerulus. The 2/3 scaling factor converts the cylindrical geometry of the calculation to the more realistic elliptical or spherical geometry of glomeruli.

All 3D DOCT data were subsequently analyzed by computing cumulative Doppler volume and Doppler flow range parameters. The cumulative Doppler volume (CDV) was calculated by summing together the volume of segmented voxels within the glomerular volume containing DOCT signals that were above the background threshold as defined below in Equation 2, where V(x,y,z) is individual voxel volume.

$$\text{CDV}=\sum _{\text{xyz}}\text{V}(\text{x},\text{y},\text{z})$$ (2)

The threshold value (|v_z| = 0.05 mm/s) for computing CDV was set to slightly above the average DOCT signal from regions of static tissue (|v_z| = 0.01 mm/s). The CDV is not equivalent to actual “Blood Volume” as DOCT is only sensitive to scatterers (red blood cells). Therefore, the CDV we measured is expected to be smaller than actual “Blood Volume” being coupled with the level of hematocrit.

Doppler flow range (DFR) was defined as the flow rate at 99% area under the curve (AUC) of the 3D flow histogram (see Figure 4). 3D flow histograms were calculated by integrating the DOCT signal over the lateral cross-section (en face) area of the segmented DOCT signal at each depth position. Integration over the en face plane eliminates the angle-dependent uncertainty of the Doppler velocity therefore providing an accurate quantitation of flow as defined in Equation 3 below [58]. Doppler flow calculations were performed in en face images according to Equation 3, where v_z(x,y,z) is the segmented DOCT velocity according to a threshold of 0.05 mm/s integrated over the xy segmented DOCT area.

Figure 4.

Blood flow histogram comparison. (A–B) Baseline blood flow histogram. (C–D) Blood flow histogram following injection of mannitol. (E–F) Blood flow histogram following injection of angiotensin II. Each histogram represents a compilation of the segmented DOCT signal at every en face plane through the entire depth of the glomerulus. Plots on the lower row are closeup versions at the base of the histograms (below the dashed line) in the upper row. Each group (A–B), (C–D), and (E–F) represents different glomeruli imaged in the same rat.

$$\text{Flow}=\underset{xy}{\iint }{\text{v}}_{\text{z}}(x,y,z)\mathit{dxdy}$$ (3)

Statistical Analysis

Data is given as mean ± standard deviation. One-way ANOVA test was performed to evaluate sampling variance followed by Tukey’s multiple comparison test with respect to the “baseline” values to determine statistical significance (p < 0.05 were considered significant). All histograms were computed using consistent bin size of 5×10⁻⁶ μl/s and minimum threshold equal to 0.05 mm/s.

Results

After the left kidney was exposed beneath the microscope, several glomeruli were located by scanning the beam across the surface of the kidney. 3D data sets of glomeruli were acquired and stored for subsequent analysis. With PIP (picture-in-picture) mode enabling simultaneous OCT/DOCT on-screen viewing, we were able to see blood flow in each glomerulus in real time while scanning the surface of the kidney. OCT/DOCT fused images were constructed offline to verify the spatial correlation of the captured DOCT signal. OCT/DOCT imaging of kidney glomerulus revealed intra-glomerular blood flow from the internal capillary network. Figure 1 shows images of three en face planes at different depths in a representative glomerulus: the top row is the upper region (A–C), the middle row is the middle region (B–F), and the bottom row is the lower region (G–I). Figure 1A,D,G show OCT intensity images revealing the kidney microstructure. Uriniferous tubules are readily identified surrounding the glomerulus, which is the circular structure in the middle of the image which is surrounded by the crescent shaped capsular space of Bowman. Figure 1B,E,H show corresponding DOCT images from the same plane depicting red blood cell velocity in numerous glomerular capillaries. As indicated by the colormap scale, red to yellow represents increasing velocity of blood flow in one direction while blue to cyan represents increasing velocity of blood flow in the opposite direction. Therefore, the mixture of these colors seen in the DOCT images demonstrates the varying velocities as well as the convoluted nature of blood flow through the glomerular capillaries. Figure 1C,F,I are the fused OCT/DOCT images demonstrating the spatial correlation of the DOCT signal within the glomerulus seen in the OCT images. Side-by-side OCT/DOCT images sectioning through the entire 3D glomerular volume are shown in Supplementary Video 1. DOCT clearly visualizes the vascular pole showing the afferent or efferent arterioles (Figure 1H, arrow). Note that this imaging information is not readily apparent in the corresponding OCT structure image. Figure 2A shows a top-down view of the complete 3D DOCT signal captured from the glomerulus in Figure 1 where the vascular pole with an arteriole is readily identified (arrow). Figure 2B offers a 3D perspective of the OCT/DOCT signal from the same glomerulus. The grayscale colormap represents the segmented OCT structural image showing the renal tubules and Bowman’s space, and the red/blue colormap represents the DOCT signal. Beneath the broad, gray, umbrella-like structure (Bowman’s space) is the glomerulus interior where numerous DOCT signals are nested together representing the blood flow in capillary tufts (see Supplementary Video 2 with 3D camera rotation to enable viewing from different perspectives).

Figure 1.

OCT and DOCT imaging of rat glomerulus. (A,D,G) OCT en face view of single glomerulus. (B,E,H) DOCT en face view of the same glomerulus. (C,F,I) Fused OCT/DOCT image showing spatial agreement between the OCT image and the corresponding DOCT image. The three images in each row correspond to the same depth imaging plane. Depth (A–C) = 440 μm; (B–F) = 470 μm; (G–I) = 545 μm; Image size: 325 × 278 μm. See Supplementary Video 1 for entire 3D OCT/DOCT en face sectioning of the glomerulus.

Figure 2.

(A) Top-down (en face) view showing numerous clustered DOCT signals within the glomerulus representing red blood cell velocities. Image size: 325 × 278 μm. (B) Three-dimensional (3D) rendering of the OCT/DOCT data set of the glomerulus in Figure 1 reconstructed from a series of cross-sectional images (see Supplementary Video 2). Grayscale colormap represents segmented kidney microstructures from OCT and red-blue colormap represents bi-directional blood velocity from DOCT. Volume size: 325 × 278 × 580 μm.

In vivo DOCT imaging was performed during three separate physiological states: baseline, following injection of mannitol, and following injection of angiotensin II. Figure 3 shows representative glomeruli imaged under each condition. Baseline DOCT is depicted in Figure 3B with mannitol and angiotensin II in Figure 3D and Figure 3F, respectively. The increase of the DOCT signal is readily visible following administration of mannitol and the decrease following angiotensin II. Corresponding OCT images are presented in Figure 3A,C,E for identification of the glomerular region in the corresponding DOCT images.

Figure 3.

Representative OCT and DOCT images (XZ) from 3 different physiological states. AB) baseline, C–D) after mannitol, and E–F) after angiotensin II. Comparison among the images shows differences in the observed DOCT signal in different glomeruli under altered blood flow conditions. Image size: 294 × 278 μm.

Analyzing the 3D data sets enabled quantitative evaluation of the DOCT signal within the glomerular region. By segmenting the DOCT velocity signal at each depth position (en face plane) according to a minimum background threshold (|v_z| = 0.05 mm/s) and multiplying by the cross-sectional area of the segmented signal, blood flow histograms of the 3D DOCT data sets were obtained. Figure 4A shows a representative baseline blood flow histogram from the 3D DOCT imaging with Figure 4B showing close up view of histogram base (region below dashed line in Figure 4A). Similarly, Figure 4C–D shows a representative blood flow histogram after injection of mannitol solution and Figure 4E–F after injection of angiotensin II. The histogram peak following injection of mannitol increased to 3 times higher than the baseline histogram peak. This indicates that cumulative Doppler volume increased dramatically after administering mannitol (a mild vasodilator), but also, the maximum blood flow rate increased as noted by the wider range of flow rates (both positive and negative) as seen in Figure 4D (arrows) that are absent in the baseline histogram. Blood flow resulting after injection of angiotensin II (a vasoconstrictor) was also significantly affected. Figure 4E–F show flow rates only within one or two of the lowest histogram bins depicting a noticeable decrease in overall blood flow.

Three-dimensional (3D) analysis of OCT and DOCT data sets was performed to extract the total glomerular volume (GV), Doppler flow range (DFR), and cumulative Doppler volume (CDV). Figure 5 shows the computed GV, DFR, and CDV from one representative rat kidney. Figure 5A shows the glomerular volume from each of the 7 physiological states induced during the experiment. Even though a slight increase in glomerular volume was noticed after injecting mannitol the first time and a slight decrease after injecting angiotensin II, no significant change in glomerular volume was observed during all data sets collected. Figure 5B–C shows the DFR and CDV values from the 7 different physiological states in the same rat. For each state, an average of n≈4 glomeruli were analyzed from all different glomeruli. It is interesting to note that after both the first and second mannitol injection, the DFR and CDV values returned to near-baseline levels during the “recover” period indicating that the effects of the mannitol were wearing off (~15–20 minutes elapsed since injection). Approximately 10 minutes after administering the third mannitol injection to increase blood flow, angiotensin II was injected. This injection of angiotensin II reduced both DFR and CDV dramatically, well below the baseline level, even against the opposing effects of mannitol.

Figure 5.

Analysis of in vivo DOCT imaging of rat glomeruli. (A) Glomerular Volume. (B) Doppler Flow Range. (C) Cumulative Doppler Volume. Each of these parameters was measured in multiple glomeruli with different glomeruli each time under 7 separate conditions: baseline (n=3), following injection of mannitol (n=3), after the effects of mannitol diminished (n=4), following a second injection of mannitol (n=3), after the effects of mannitol diminished (n=3), following a third injection of mannitol (n=3), and finally following an injection of angiotensin II (n=6). Asterisk (*) indicates p < 0.05 compared to “baseline.” Labels: Mann=mannitol; Ang II=angiotensin II.

Discussion

Optical imaging using DOCT, a functional extension of OCT, is able to reveal single glomerular blood flow in vivo. It is important to note that all the observations were performed without introducing artifacts otherwise associated with penetrating the kidney (i.e., performed in a non-invasive fashion). Furthermore, DOCT requires no exogenous contrast agents like multi-photon fluorescence microscopy [28, 29, 59] and confocal laser scanning microscopy [60] and offers deeper tissue penetration than most optical microscopy techniques. Although DOCT can penetrate to a depth of about ~1–2 mm, monitoring renal microcirculation less than 500 μm below the kidney surface following renal transplantation has been reported to yield a promising predictive marker for post-transplant graft function [61]. An additional advantage of DOCT is that it can acquire numerous cross-sectional images in rapid succession yielding comprehensive 3D volumes of tissue. Such images can be used to determine the size, shape, and blood flow through glomeruli, thereby providing pathological information regarding glomerular disease. Fast imaging speed [62] and also the ability to be incorporated into needle [63, 64] and laparoscopic [65] devices for imaging blood flow in various orientations and even deeper in solid tissue or in not easily assessable locations allows DOCT to perform quick, repetitive 3D scans that can provide vital information for kidney evaluation. These unique aspects of DOCT might make this imaging technology useful for renal imaging in clinical settings.

In our study, we demonstrated the ability of DOCT to quantitatively image glomerular blood flow in vivo. Blood cell velocities (between ±1 mm/s) obtained with DOCT agree well with previous optical imaging studies [21, 60] as well as capillary blood flow (~1–4×10⁻⁴ μl/sec) as compared with values reported in the rat cerebral cortex [58]. However, DOCT quantifies velocity by detecting phase difference and therefore has an inherent phase-wrapping effect (since phase detection is limited by ±π). The maximum detectable velocity (without phase-wrapping) is determined by the axial scan speed [66] and the minimum is based on the inherent phase noise present from the laser. In our system, 16 kHz axial scan speed yields our upper detection limit of ±3.90 mm/s with minimum limit of ±0.01 mm/s, which encompasses reported values of intra-glomerular blood flow. However, when imaging blood flow in arterioles, the phase wrapping effect might be present (see Figure 1H). Future studies using higher-speed laser source [62, 66] would solve this limitation.

One limitation involved in the DOCT signal segmentation is the fact that the intra-glomerular capillaries are at or below the axial resolution of our system and therefore speckle noise might be a significant factor affecting quantification of the DOCT signal. Since the segmented DOCT signal is composed of numerous small DOCT clusters representing the small vessels within the glomerulus, a large number of the voxels are boundary voxels and thus may only represent partially filled voxels. Some of the partial voxel effects might be alleviated by de-speckling but all the values we report were calculated directly from the raw segmented DOCT signal prior to any image processing.

Furthermore, if the vascular pattern is of more interest, other techniques such as Doppler variance imaging [41], speckle variance imaging [67], optical coherence angiography [68], and optical microangiography (OMAG) [69] can be applied. Future studies are also needed to cross-validate the blood flow measured by DOCT with confocal or two-photon microscopy by using multi-modal optical systems [70–73].

While the Munich-Wistar rat model provides superficial glomeruli that can be readily observed, the capability of DOCT to detect glomerular blood flow in the human kidney has yet to be demonstrated. Previous studies show the ability of OCT to image glomeruli in the intact human kidney ex vivo [36, 37] and several of the studies mentioned above report using optical imaging methods at depths much shallower than DOCT to monitor renal circulation in animal models and transplant patients. Therefore, future studies using DOCT to quantify microcirculation changes compared with immediate and long-term transplant outcome and compared with renal disease progression would provide needed insights regarding possible future clinical applications of DOCT in renal care. However, the sensitivity of DOCT to detect subtle changes that can be associated with disease states (such as diabetes or renal artery stenosis) or with abnormal states (such as anemia or partial/full nephrectomy) has not yet been demonstrated [74]. Thus, further studies are needed to evaluate native blood flow responses to renal disease in animal models and human patients in order to better understand the capability of OCT/DOCT to provide quantitative measurements of the kidney.

In these studies, we have demonstrated three-dimensional (3D), label-free imaging of microcirculation in the rat glomerulus in vivo under differing physiological conditions using DOCT. A significant difference in DFR and CDV was observed following injection of both mannitol and angiotensin II. The ability of DOCT to visualize hemodynamic changes in vivo could provide a direct method for clinicians to evaluate kidney perfusion and response to therapy in real time. Furthermore, intraoperative visualization and analysis of alterations in renal perfusion associated with injury or disease may help improve renal health care in the future.