Synchrotron radiation (SR) diffraction enhanced imaging (DEI) of chronic glomerulonephritis (CGN) mode

2.1 Animal mode preparation

In accordance with Institutional Animal Care and Use Committee Guideline, ten two-month old healthy male New Zealand white rabbits, weighing between 2 to 2.2 kg (normal sized) were carried out for this experiment. They were randomly separated into two groups, and each group contained five rabbits. For the CGN model, one group was administered l7 mg/kg doxorubicin hydrochloride via ear vein injection, followed by a second dose after one week. The other group as control group were raised up in the same way for two months without any medication. Both groups were housed in wire cages and maintained in a temperature-controlled room, with free access to water and standard rabbit chow. In two months, the rabbits were humanely sacrificed by hyperanesthesia (intravenous sodium pentobarbital overdose injection). In a standard animal operation room, kidneys were removed from animals in both groups. The harvest was done under renal artery, renal vein and ureter ligation, which was good for maintainence of renal microstructure. A portion of each kidney was removed for histological examination, and the rest were fixed in 4% formalin solution for one week and labeled separately according to CGN group and control group respectively. Formalin solution could help vessels become concretionary, in case blood flow while sample excision for DEI experiment. The renal tissues were imaged with a 4W1A beam line of Beijing Synchrotron Radiation Facility (BSRF). Due to the limitation of DEI experimental visual field, each slice of kidney tissue containing both cortex and medulla had to be narrowed down to approximately to 2mm×7mm×7mm, well-distributed each surface [11].

2.2 Laboratory test and histology

All clinical tests (renal function blood test and urine protein test) had been done in laboratory of Shanghai No. 10 people’s hospital. Laboratory exams showed the blood urea nitrogen (BUN), creatinine (Cr) and the protein quantification. All histological examination had been done by pathology department of Shanghai NO.10 people’s hospital. The rich part in cortex and medulla was chosen for histological examination, and they were observed under light microscope after fixation, embedding, hematoxylin-eosin stain, etc.

2.3 DEI experiment

The DEI experiment of rabbit CGN mode was performed at the 4W1A beam line of Beijing Synchrotron Radiation Facility (BSRF). The schematic set-up system is showed as Fig. 1(a) and the photography of real device showed as Fig. 2. Figure 1(b) is the theoretical calculation of 4W1A light source spectrum curve. The storage ring energy was 2.5 GeV with a 250 mA Current, while the energy of X-ray beam received in experiment hunch was 14KeV. To acquire higher quality images, two silicon (1 1 1) crystals were equipped in this set-up. The X-ray beam from wiggler was monochromatized by a perfect silicon (1 1 1) crystal and refracted by an analyzer crystal after transmitting sample tissue, then it was detected by a high sensitive CCD with the pixel size approximately to 10.9μm×10.9μm (X-ray Fast Digital Imager 18mm camera system, Photonic-Science Ltd., UK) [20, 26]. When the analyzer was positioned on different angular settings, different sensitivities on tissues and their interfaces were provided. The size of experimental vision field (facula) was about 9mm×9mm. The kidney sample was fixed on a plastic needle by super contact glue and the needle was placed on rotary table between two analyzers, with sample adjusted into the visual field precisely within a tentative 360° spin. The rotary table was driven by stepping motors. First the rocking curve was scanned, then the right low slope position B (Fig. 3a) on the curve was selected as light cutting-in direction to make CT scan. According to our repeated trials (Fig. 3b), position B gained the highest resolution though previous study suggested peak position (position A) [21, 24]. The refraction-angle resolution describes ability of the DEI system in differentiating the x-rays refracted by the sample [22]. It evaluates the characteristics of the DEI system, which are very different from the conventional x-ray imaging system [23]. Both group samples were acquired by tomography phase contrast on routine scanning procedure. To acquire images from each angle, the sample needle axis was rotated 360°, each step 0.5°. At last 721 projections were acquired in total, including one extra background image scanned in advance, and the exposure time of each projection was 120 milliseconds. The phenomenon of position offset was discussed and validated in the computer simulation experiment based on the equivalent rectilinear propagation model [24]. All the parameters were selected by repeated trials, in order to obtain high quality images [25].

2.4 Image processing

Although the DEI can provide high-resolution images, image processing is required during the analysis of the images [9]. The original projections are usually confused with spatial superimposition of lesions; small lesions overlapped inside the tissue could not be detected. Therefore, the professional software package named Computed Tomography of Diffraction Enhanced Imaging V2.0 (invented by Dr. Zhang Kai, High Energy Physics, Beijing, China 2014 Chinese Physics C [26]), was used to convert those projections into 3-Dimensional images. It combines BSRF source characteristics and provides methods for tomography reconstruction and information separation reconstruction algorithm. The software derived from Matlab principle has a parallel computing algorithm, which extracts the sample’s refractive index and refractive direction derivative along the X-axis and the Y-axis out of the two-dimensional CT reconstruction. Firstly, the geometric correction of diffraction enhanced CT projection data, including homogenization correction, angle of rotation correction and calibration were executed. Secondly, during CT slop (B position) data reconstruction, it can get the refractive index from derivative distribution of the two-dimensional segmental information along their X and Y-axis direction. Thirdly, sample date such as absorption, refraction and scattering width were separated and extracted. Filtered background projection algorithm was adopted as the first step of information separation reconstruction algorithm. Figure 4 (a) shows images on process. Since we had completed a 360° turn of the sample, with each 0.5° acquired one slide, so there was unequal balance between each two opposite slides in one plane, due to the irregular form of sample. Then correction method of angle and rot-axis was applied to convert the incongruity between slides. Two projections from 0° and 180° were selected, then 5 characteristic symmetric spots from them was picked up randomly and matched up. Consequently, a theoretical value of angle and rot-axis was obtained automatically with this software. To correct data from all slides in a more accurate way, an optimized value was chosen after an initial scan based on the theoretical value. Only when all data correction completes, can the refractive index of images be reconstructed. The relevant parameters were selected according to experimental process, such as the rotate range, the steps, the intervals, and the region of interest. Then another soft ware called Avizo Fire was adopted after DEI reconstruction software. It converted 2D straightforward visualization and measurement into advanced 3D visualization image by processing, quantification and skeletonization, which were also called image processing and image segmentation, and volume-rendering technique. At the end, date quantitative analysis was set up on both 2D and 3D display.

2.5 Statistical analysis

Statistical analysis was crucial for this study, after we collected all datasets from DEI experiment, laboratory, and pathological department. Statistical analysis was performed by software IBM SPSS 19 and Matlab (R2013a). Blood urea nitrogen (BUN) and creatinine (Cr) was tested regarding to renal function and urine protein was regarding to damage of glomerular basement membrane (GBM). Albuminuria assumed damage in glomerular basement membrane (GBM). More over, the gray values were extracted from the respective segmentations of the DEI of each 20 kidney datasets (each group contained 5 rabbits, each rabbit had 2 kidneys) and the largest data size was around 400 million data points (2000×2000 points in one matrix). The cortex part was the most important area for nephritis lesion, thus cortex was selected as region of interest (ROI). Due to enormous elements in one ROI matrix, we manually selected one kidney with positive urine test and positive histological test from CGN group, the other one from control group randomly. Statistical analysis was performed using the randomized sampling function, randomly selecting 36,000 data points (450×80) from each kidney ROI compartment.

3.1 Kidney function laboratory test result

Blood urea nitrogen (BUN) and creatinine (Cr) were shown as following Fig. 5 (a) and (b), which were closed to the reference value of normal blood urea nitrogen (BUN) (6.47±0.33 mmol/l) and creatinine (Cr) (67.11±1.33 umol/l). The significance of both two-sample T tests result of BUN and CR were 0.1995 and 0.971 respectively (Fig. 5), indicating there were no significant differences between the two groups in renal function test and both groups had normal function. The urine protein quantification(the albuminuria test was carried out by sulfosalicylic acid solution) is listed in Table 1. According to normal reference value, protein should be negative (equal to 0) in urine test. CGN group was implied to contain nephrons damage inside, while the control group expressed normally in urine protein.

3.2 Histology result

Renal histological sections were evaluated under optical microscopes. The cortex part was chosen as ROI and it was evaluated with respect to integrity of glomeruli structure, mesangial cell proliferation, irregularity in the tubular cell size, renal interstitial inflammatory. The qualitative observation of the control cortex and the CGN cortex in all three different modalities (histological images, DEI images, DEI 3D image) were provided in Fig. 6. Images (a) (d) showed H&E stained 100 times magnified cortex section under optical microscope, from control group and CGN group respectively. Image (a) from control group indicated that the glomeruli structure, mesangial cell and interstitial space were clearly displayed in normal formation and in good condition. The glomerular capillary loops were well defined and thin. The endothelial cell and mesangial cell numbers were normal. The surrounding renal tubules were normal. While image (d) from CGN group was found that the swelling in the renal tubular epithelial cells, mild damage in structure of Bowman’s capsule, visible hyperemia in juxtaglomerular angiectasis, cell infiltration. Due to the damage of podocyte and the loss of normal charge barrier, albumin was selectively leaked out and the subsequent proteinuria happened in CGN group. The capillary loops were surrounded by endothelial cells, epithelial cells, mesangial cells and increased neutrophils. The cortex of the CGN kidney could be easily distinguished by its different morphological structure. Both image (a) and (d) were very similar to DEI image (b) and (e).

3.3 DEI image result

DEI images of the control and CGN cortex showed in Fig. 6(b) and (e). Figure 6(b) and (e) were magnified cortex part of Fig. 7(a) and (c). The result showed that no matter the control or the nephritis kidney, the major features including artery, vein, straight collecting ducts and papillary tubules were clearly visible with the spatial resolution reach to 10μm. but the difference between Fig. 6(b) and (e) was hard to tell except that Fig. 6(b) control cortex was remarkable in nephron, renal tubules and renal interstitial while Fig. 6(e) CGN cortex interstitial fuzzy exudative change, mesangial proliferation among its microstructures. Figure 6 (c) and (f) were 3-D DEI. (c) The control kidney microstructure was clear, normal vessels form. (f) CGN group structure was a bit disorder, twisted capillary hyperplasia. Figure 7 (a) (c) Coronal sections of kidney DEI images, the cortex (CO), outer stripe of the outer medulla (OSOM)[33, 34], inner stripe of the outer medulla (ISOM), and the inner medulla (IM) were clearly distinguishable by different gray values, also the interlobar vessels, arcuate vessels, even small interlobar vessels were clearly displayed. Figure 8 (a) 3-D reconstruction of transversal section of CGN group. (b) 3-D reconstruction of coronal section nephritis group (c) 3D reconstruction colored image of control kidney. Details of interlobar vessels, cortical radial vessels and glomerulus can be seen in cortex. Vasa recta were visible in both juxtamedullary region and cortical region. Afferent arterioles branched from cortical radial artery leading into glomerulus while efferent arterioles leading out glomerulus were distributed towards cortical peritubular capillaries. However, there was a lot of similarity between histologic images.

3.4 Statistic and quantitative data analysis result

Evaluation of glomeruli is one of the most important aspects of the diagnosis of CGN. The most common and non-invasive way is to compare their morphologic change by DEI images (as we discussed above), the other way we have done was selecting equal area of each kidney ROI, extracting and calculating the grey value of each matrix, and applying MATLAB and SPSS for analysis, shown as Fig. 7 (b)(d). First, the mean grey value of each cortex from both groups drew out by MATLAB, as Fig. 7(f) was shown. The mean grey value rank of each random cortex from nephritis group was 91 to 112, and from control group was 121 to 141. The significance of two sample paired T test (p <  0.05), which indicated grey value from two groups differed greatly from each other, and CGN group was much lower comparatively. Second, one image from CGN group positive in proteinuria Fig. 7(b) and the other random image from control group Fig. 7(d) were picked out for accurate analysis. An equal small area of cortex part (matrix in 450×80) was drawn out from both images for grey value analysis. The grey value of 36,000 dots was counted and shown as Fig. 7(e). With the same total number of voxel between control group (green curve) and nephritis group (black curve), the green curve towards right moved according to black curve, which implied the grey value of nephritis group (ranking from 55 to 160) was lower than it was from the control group (ranking from 75 to 175). Furthermore the saltatory frequency of black curve showed that cortex of nephritis group had remarkable grey value difference in regions, which suggested those diffused dots (low grey value) in CGN cortex arrayed within micro regions.

4.1 DEI process merits and defects

DEI images have almost same up-sizing features with histologic image. Based on this point, we ventured to explore more reliable data from this new experiment without contrast media, finally it was proved that modern radiologic method could provide same level picture which used to be able to obtain only under a microscope. The result shows that under DEI device, the renal architecture including vessels, tubules and glomeruli of about 10μm spatial resolution can be demonstrated clearly without any contrast agent, which makes it possible to diagnose different renal disease associated with vessel by measuring or comparing the feature size. The major obstacle in realizing the picture is the motion artifacts caused by vaporization of formalin, which would lead to serious blur during scanning or reconstruction.

The purpose of this study was to exploit the diagnostic value of renal disease by SR DEI phase-contrast imaging technology. Imaging morphological changes on macrovessels, microvessels, tubules and glomeruli were crucial for chronic glomerulo nephritis (CGN), so that proper treatment strategy could be performed to cure it or postpone the progression of disease. One way, DEI images morphologic analysis based on the doxorubicin hydrochloride injected CGN animal mode, were able to correspond with the images obtained through histological technique. The other way was Grey Value Analysis between CGN and normal kidney. Without any contrast agents, DEI revealed the cortex (CO), outer stripe of the outer medulla (OSOM), inner stripe of the outer medulla (ISOM), and the inner medulla (IM), and the architecture of vasculature and the relationship between vessels were clearly visible. For a more objective segmentation, the area of image processing and machine learning provides a wide variety of adaptive dedicated algorithms [39, 40]. The 3-dimensional model of kidney had been successfully established which provided an excellent view of renal microstructure and artery, vein, straight collecting ducts and papillary tubules the interlobar vessels, arcuate vessels, even small interlobar vessels were clearly displayed in DEI magnified images and reconstruction images. But the details of glomeruli were not as clear as histologic images. Thus the Grey Value analysis was carried out for accurate differentiation of normal glomeruli and chronic nephritis glomeruli. The CGN group had lower grey value than normal group in cortex.

Kidney sample and histologic and DEI image segment as ROI was essential for this experiment.We especially thank pathologist Professor Feng Li-Jin for guidance throughout experiment. Evaluation grey value and analysis were based on Matlab and SPSS. Dr. Xu Xu-Dong working for precision optical Engineering designed program into Matlab. First, grey value from color map was selected, then ROI was chosen and input to program.

Another potential source of error was the influence of formalin fixation on kidney tissue signal characteristics in DEI imaging. Thus the gray values displayed in this study were proportional to the respective density of the formalin fixed sample.

4.2 SR future improvement

Currently, most experiments had to be done in vitro due to limitations of resolution and scan vision field. Although we did not use of any kind of contrast agent, we failed in in-vivo performance because of those limitations. High-speed X-ray DEI may be another possible solution to image live animal by substantially decreasing imaging time. In fact, completing a CT scan in one respiration phase is already possible but the rotation speed of the sample stage and the time resolution of the detector still need to be improved [34]. DEI was still associated with high radiation doses and the present study has not been optimized in this regard. However, alternative acquisition and post-processing techniques as well as dedicated iterative reconstruction algorithms should be simplified. Translation to a clinical setting is still dependent on further technological developments, such as larger gratings to cover a larger field of view [41].Further research on fine-tune of SR resolution of in vivo intact animal imaging is eagerly requested.