2.1.

Large Animal Model for Ischemia/Reperfusion Injury

2.2.

Visible Reflectance Imaging System

The VRIS comprised of several components (Fig. 1): light source, mirror, filter system, charge-coupled device (CCD), and personal computer. The light source is a 30 W quartz-halogen lamp (Oriel, Irvine, California) which is reflected onto the area of interest via adjustable fiber-optic rings. The diffusely reflected light from the surgical field is then reflected to a liquid crystal tunable filter (LCTF) (Cambridge Research Instruments, Woburn, Massachusetts) by a mirror angled at 45 deg incident to the surgical plane. The CCD ($768\times 512\text{pixels}$; Photometrics, Tuscon, Arizona) acquires images (100 ms per image) as the LCTF adjusts to each desired wavelength. The LCTF transmits light from 420 to 700 nm. To improve the signal-to-noise ratio of the spectra contained in the data set, the image is binned by 3 pixels. This ultimately results in a $256\times 170\text{pixel}$ image plane. Each image cube contains 126 image planes, with a single image plane collected at single spectral increments of 1 nm from 520 to 645 nm. For background measurements, a 99% diffuse reflectance standard was used (Labsphere, North Sutton, New Hampshire).

Fig. 1

Schematic of visible reflectance imaging system (VRIS). The light is generated by a broadband quartz tungsten halogen lamp (200 to 1100 nm) and delivered to the operating plane (indicated by the kidney) via fiber optics. A mirror directs reflected light to a tunable filter (420–700 nm), which then passes the light to a thermoelectrically cooled detector.

The image data sets are stored and analyzed using a personal computer (Gateway, Irvine, California). Data analyses are carried out using in-house and commercially available scripts in Matlab™ (Mathworks, Natick, Massachusetts). Briefly, the spectral response of the tissue examined is subjected to deconvolution for calculation of various tissue parameters (oxy- and deoxy-hemoglobin, water). Oxygenated hemoglobin exhibits distinct spectral bands at 537 and 567 nm, whereas deoxygenated hemoglobin exhibits a markedly different spectral band at 553 nm.

2.3.

Data Analysis

According to the Beer–Lambert law, the absorbance of a sample ($A$) depends on three parameters: pathlength ($l$), concentration ($c$), and the molar absorption coefficient ($\epsilon$). Absorbance, however, can also be defined as the logarithmic ratio between the intensity of light incident on a sample ($I_0$) and the intensity of light transmitted through a sample ($I_1$), where $A={\log }_{10}(I_0/I_1)$. Finally, one can extrapolate concentration of the sample from the ratio of light intensity, because ${\log }_{10}(I_0/I_1)=\epsilon cl$. Absorbance spectra were used in the prediction of oxygenated hemoglobin concentration because of the linear relationship between concentration and absorbance spectra. Absorbance image cubes were created by referencing each raw spectral cube ($I_0$) by the reflectance of a standard reference material ($I_1$). The relative concentration units for deoxygenated hemoglobin utilized throughout this study are percentage oxygenated hemoglobin ($\%{\mathrm{HbO}}_2$), which is defined as the percentage of the concentration of oxygenated hemoglobin divided by the total hemoglobin concentration.¹⁵ All data analysis was performed using algorithms and software written in-house using the programming languages Matlab and IDL/ENVI (ITT Visual Information Solutions, Boulder, Colorado).

The calculation of relative oxygenated hemoglobin concentration was a two-step process. The first step involves using the classic least squares method, sometimes referred to as the $K$-matrix method,¹⁶ which is explained in detail in elsewhere.^17,18 In brief, the measured absorbance spectra and concentration of pure components (oxygenated and deoxygenated hemoglobin) are linearly proportional to each other via a $K$-matrix, $A=CK$, where $A$ is the absorbance spectra matrix ($\text{number of measured spectra}\times \text{number of wavelengths}$), $K$ is the matrix of pure components (number of components in the model × number of wavelengths), and $C$ is a concentration matrix (number of measured spectra × number of pure components in the model) with relative concentration units of $\%{\mathrm{HbO}}_2$. In this method, the absorptivities usually defined in univariate Beer law fitting are normalized to constant path length, resulting in $K$ being a normalized absorption coefficient matrix.¹⁸ These pure component spectra were measured external to these in vivo measurements. Also, the number of measured spectra in $A$ and number of concentrations in the $C$ matrix vary with each data set, because the number of pixels that include the kidneys varies, but in every data set is $>1000$. One limitation of this $K$-matrix method occurs when one or more spectroscopically active components are not represented in the calibration set.^17–19 In the case of visible reflectance spectra of kidneys, both chemical and optical nonhemoglobin spectral contributions can limit the accuracy of the concentration prediction. The second step in the concentration determination is to overcome this limitation by applying externally obtained relative oxygenated hemoglobin concentrations to correct for these nonhemoglobin contributions using augmented classic least squares, described elsewhere.¹⁹

Statistical significance was determined with the Student $t$-test with the degrees of freedom modified by the Greenhouse–Geisser correction.²⁰ This correction was chosen to provide more conservative estimates of the statistical significance in this study by mitigating pixel-to-pixel spatial correlations and temporal correlations. These correlation analyses were carried out using the CAR package²¹ in the R statistical computing language.²²

3.1.

VRIS Correlates with Measured Hemoglobin Oxygenation

Figure 2 shows a comparison between the oxy-hemoglobin concentration predicted by visible reflectance spectroscopy and catheter-drawn kidney blood oxy-hemoglobin concentration determined externally (canine nephrectomy). These externally determined oxy-hemoglobin concentrations ($\%{\mathrm{HbO}}_2$) are from single blood samples taken as a function of induced ischemia time (29%, 32%, 40%, and 96% ${\mathrm{HbO}}_2$, respectively). The visible reflectance–determined oxy-hemoglobin concentrations are spatially averaged concentrations (calculated for approximately 6000 individual pixels). The error bars represent one standard deviation from the spatially averaged concentrations. The predicted and the externally measured oxygenated hemoglobin concentrations ($\%{\mathrm{HbO}}_2$) correlate quite well (Fig. 2). The dashed line demonstrates a linear regression with a slope of 1.0 for reference. The cross-validation standard error of calibration was estimated for the calibration data, and the influence of each calibration point is insignificant compared to the standard error of calibration estimated by including all calibration data within a 95% confidence interval. This analysis suggests that no point in the calibration set holds significant influence or leverage over the calibration, and therefore, the calibration is valid.¹⁸

Fig. 2

Calibration plot for deconvolution of percent oxygenated hemoglobin ($\%{\mathrm{HbO}}_2$). Mean oxyhemoglobin concentrations calculated from the visible reflectance images are plotted versus actual oxyhemoglobin concentrations (blood gas measurements).

Figure 3, a set of images of the canine kidney during ischemia injury, illustrate the ability of the visible reflectance images to measure renal tissue oxygenation intraoperatively. The largest image (left) is a visible-color image of the kidney. The sequential images A–E are concentration images of the kidney before and during the ischemia/reperfusion injury. The intensity of the grayscale in the images (A–E) corresponds to the oxy-hemoglobin concentration scale shown in the lower right of the figure. According to the image colormap, the highest concentration of ${\mathrm{HbO}}_2$ is white, and the lowest concentration of ${\mathrm{HbO}}_2$ is black. These intensity values were determined using algorithms developed in-house. The red shaded area in Fig. 3(a) indicates the ROI from which mean and standard deviation oxy-hemoglobin concentrations were determined. As indicated by the visible reflectance images of the kidney during warm ischemia, the ${\%\mathrm{HbO}}_2$ decreases with ischemia time, and the kidney is darkest after 30 min of ischemia, demonstrating significantly lowered ${\%\mathrm{HbO}}_2$.

Fig. 3

Left, isolated right kidney; right, visible reflectance images indicating $\%{\mathrm{HbO}}_2$ at baseline (a) and 5 (b), 10 (c), 20 (d), and 30 (e) min of warm ischemia. The selected region of interest for calculated values is indicated in red.

3.2.

Effects of Regional Perfusion on Ischemia

Figure 4 shows a set of sequential oxy-hemoglobin concentration images of a left porcine kidney during ischemia/reperfusion injury. The intensity values in these images were determined using algorithms developed in-house described in the Data Analysis section and correspond to the scale on the right of the figure. In Fig. 4(a), three regions of the kidney are defined by two diagonal lines corresponding to the superior, middle, and inferior poles of the kidney. These three regions are used to calculate the mean regional oxy-hemoglobin concentration as a function of ischemia/reperfusion clamping and unclamping times (shown in Fig. 5). It is clear in Fig. 4(b)–4(h) that the superior pole of the kidney remains more oxygenated than the middle or inferior poles during ischemia. After reperfusion [Fig. 4(i)–4(o)], however, kidney oxygenation appears to be relatively homogeneous. The average $\%{\mathrm{HbO}}_2$ concentrations have been plotted in Fig. 5.

Fig. 4

Sequential oxygenation images of a left kidney at baseline (a) and during cold ischemia (b–h) and reperfusion (i–o). Ischemia time points: 5, 10, 15, 20, 25, 30, and 35 min after vessel clamping. Reperfusion time points: 5, 10, 15, 20, 25, 30, and 35 min after vessel unclamping. SP, superior pole; MP, middle pole; IP, inferior pole.

Fig. 5

Profiles of kidney oxygenation before vessel clamping, during ischemia, and after reperfusion. (a), Mean ${\mathrm{HbO}}_2$ concentrations for the entire kidney. (b) Mean ${\mathrm{HbO}}_2$ concentrations for the superior pole, middle pole, and inferior pole.

Figure 5 is the scatter plot of the oxy-hemoglobin concentrations calculated for the entire kidney [Fig. 5(a)] and each kidney segment [Fig. 5(b)] and is representative of all kidneys examined in this study. Although the regional differences in tissue oxygenation are not readily apparent in the image of the kidney itself [Fig. 4(b)], examination of the calculated $\%{\mathrm{HbO}}_2$ values clearly demonstrate a difference in the mean oxy-hemoglobin concentrations of the superior pole and middle and inferior poles ($91.9\%\pm 4.5\%$, $88.5\%\pm 3.9\%$, and $88.0\%\pm 1.8\%$, respectively; $P<0.05$). This spatial difference in oxy-hemoglobin concentration persists for the duration of the ischemia, with an average $\%{\mathrm{HbO}}_2$ of $75.8\%\pm 19.9\%$ for the superior pole and $58.9\%\pm 12.0\%$ and $57.3\%\pm 10.2\%$ for the middle and inferior poles, respectively ($P<0.05$). After reperfusion, statistically significant differences in oxy-hemoglobin concentrations abate and the final mean oxy-hemoglobin concentrations of the superior, middle, and inferior poles are $73.4\%\pm 17.1\%$, $82.5\%\pm 26.4\%$, and $76.0\%\pm 15.4\%$, respectively ($P>0.05$). For Figs. 4 and 5, the means and standard deviations were calculated from ROIs (not shown for clarity) that encompass approximately 60% of the visible area in each kidney pole.

3.3.

Effects of Cold Ischemia on Tissue Oxygenation

Additionally, mean oxy-hemoglobin concentrations were compared for cold and warm ischemia cases. In Table 1 are means and standard deviations for baseline, ischemic, and reperfused kidney oxy-hemoglobin concentrations from all cold and warm ischemia/reperfusion injuries. The average baseline values were $89.0\%\pm 4.0\%$ and $85.0\%\pm 10.0\%$ ${\mathrm{HbO}}_2$ for hypothermic and normothermic kidneys, respectively; note that these values were acquired before the addition of the ice slush for the hypothermic kidneys. The oxy-hemoglobin concentrations of hypothermic kidneys after 30 min of ischemia are significantly higher than those of normothermic kidneys ($64.0\pm 10.0$ versus $24.1\pm 20.0$, $P<0.05$). The renal oxygenation of the hypothermic kidneys after 30 min of ischemia is approximately 70% of the baseline oxy-hemoglobin concentration, whereas the normothermic kidneys retain only 30% of the baseline oxy-hemoglobin concentration after 30 min of ischemia. After approximately 30 min of reperfusion, the mean oxy-hemoglobin concentrations of hypothermic and normothermic kidneys is comparable ($77.2\pm 8.0$ versus $63.5\pm 20.0$) and returns to within 80% of the average baseline $\%{\mathrm{HbO}}_2$ values.

Table 1

Comparison of mean baseline, ischemic, and reperfused kidney HbO2 concentrations for cold and warm ischemia.

For all kidneys, normothermic and hypothermic, visible reflectance imaging demonstrated a spatially distinct decrease in the oxy-hemoglobin concentration of the superior pole compared to the middle or inferior pole of the kidney. Mean oxy-hemoglobin concentrations decrease more significantly during ischemia for normothermic kidneys compared to hypothermic kidneys.