Abstract
BACKGROUND:
The outstanding functional importance of the brain implies a strong need for brain imaging modalities. However, the current imaging approaches that target the brain in rodents remain suboptimal.
OBJECTIVE AND METHODS:
In this paper, X-ray propagation-based phase contrast imaging combined with equally sloped tomography (PPCI-EST) was employed to nondestructively investigate the mouse brain.
RESULTS:
The grey and white matters, which have extremely small differences in electron density, were clearly discriminated. The fine structures, including the corpus callosum (cc), the optic chiasma (ox) and the caudate putamen (CPu), were revealed. Compared to the filtered back projection reconstruction, the PPCI-EST significantly reduce projection number while maintaining sufficient image quality.
CONCLUSIONS:
It could be a potential tool for fast and low-dose phase-contrast imaging to biomedical specimens.
Introduction
Brain diseases, such as cerebral vascular disease and brain tumors, are the leading causes of mortality in the world. Rodent models have been widely accepted and are important in preclinical research. A non-invasive and three-dimensional (3D) imaging method is necessary to understand the pathology of brain diseases and to develop appropriate therapeutic strategies. The outstanding functional importance of the brain implies a strong need for brain imaging modalities. However, the current imaging approaches that target the brain in rodents remain suboptimal.
Magnetic resonance imaging (MRI) as a three-dimensional imaging technique, can yield superb contrast between white and gray matter, but only limited spatial resolution. High-resolution small animal MRI equipped with stronger gradient systems and operating at higher magnetic field has been used to visualize the mouse brain with voxel sizes of tens of micrometers. However, usually more than 10 hours is needed to collect a set of mouse brain data [1, 2].
Micro-CT is a non-destructive technique that could provide fully quantitative 3D data that ultimately attains a higher spatial resolution. However, its contrast for brain tissue is extremely weak because the image contrast is based on differences between X-ray absorption coefficients, which are intrinsically small between tissues. Several approaches have been developed to enhance the contrast. Contrast agents have been used in micro-CT to differentiate between grey and white matter [3, 4]. Another approach has been to utilize X-ray phase shift, which has been termed phase contrast tomography (PCT). Grating-based differential phase contrast tomography (GDPCT), as a phase sensitive PCT technique, was used to study the mouse brain recently. And lots of results were obtained [5 –7]. However, It needs the phase stepping procedure [6 –8] and this requirement leads to a high radiation dose and a long data acquisition time.
Here, we introduce X-ray propagation-based phase contrast imaging combined with equally sloped tomography (PPCI-EST) for 3D imaging of the mouse brain. This paper describes the methods and materials in Section 2, the results and discussion in Section 3, the resulting conclusions in Section 4.
Methods and materials
PPCI-EST method
PPCI-EST is the combination of propagation-based phase contrast imaging (PPCI) and EST algorithm which uses tomographic projections collected at equally sloped views.
It is well known that PPCI is the simplest way to visualize the phase contrast effect, because no optical elements and complex scan procedures is required. Only one image is needed for each projection. When a coherent X-ray wave-front goes through the sample and the distorted wave-front propagates sufficiently far, the small differences in the phase propagation cause interference, and variations of intensity are observed. This pattern of intensity is the so-called phase contrast image [9, 10]. It is mathematically described as
With phase retrieval, the phase shifts of the sample could be recovered, which could greatly facilitate observation of the structures of low contrast samples, such as soft tissues [11].
EST is a novel data acquisition and reconstruction method for tomography, which utilizes a set of equally sloped projections. The equally sloped projections are collected at the views satisfying the following equation
With the PPCI-EST method, it is possible to obtain 3D images for low contrast sample using less projections which means shorter acquisition time and lower dose.
Animal surgical procedures and experimental protocols were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) and the Bioethics Committee of School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China. A BALB/c mouse (Shanghai Laboratory Animal Research Center, Shanghai, China) was deeply anesthetized via an intraperitoneal injection of pentobarbital (50 mg/kg) and perfused with 0.01 M phosphate-buffered saline (PBS) followed by 4% paraformaldehyde. The brain was then dissected and post-fixed in 4% paraformaldehyde overnight. The brain was further dehydrated using graded ethanol at room temperature prior to the tomographyexperiments.
Experiments and processing
The X-ray propagation-based phase contrast tomography (PPCT) experiments were performed on the BL13W beamline at the Shanghai Synchrotron Radiation Facility (SSRF). BL13W is a wiggler beamline that produces monochromatic X-rays with a flux density of ∼3.4×1010 photons/s/mm2 at 20keV and a narrow energy band pass (ΔE/E<5×10–3, where E is the photon energy) by a Si(111) double-crystal monochromator. A typical PPCT experimental set-up at the SSRF is illustrated in Fig. 1. The mouse brain sample-to-detector distance (SDD) was set to 80 cm for good contrast enhancement according to the experimental experience. The X-ray energy was chosen to be 14 keV by considering both the penetrating ability and phase contrast. The projections were recorded using an X-ray charge-coupled device (CCD) detector (X-ray VHR camera, Photonic Science Ltd. East Sussex, UK) with 4008×2672 pixels and straight fiber-optic coupling with a YAG:Ce scintillator of 100 μm thickness. The CCD sensor had an active input area of 36 mm (H)×24 mm (V), and the pixel size was 9μm. For the acquisition of a full tomographic data set, the sample was rotated around the rotation axis and images were recorded for each projection angle.
For the exact reconstruction of parallel CT, the number of projections (P n ) required was determined by P n ≈πW /(2s) , where W is the effective width of the sample and s is the detector pixel size [18]. In this experiment, 1553 equal-angle projections were collected during a 180° rotation with an exposure time of 18 ms for each projection. The flat- and dark-field images were collected for X-ray fluctuation correction and background normalization. Flat-field images were obtained with X-ray beam illumination without the sample in the field of view (FOV) before and after sample rotation to correct for the X-ray intensity decay, whereas dark-field images were acquired for the detector background without the X-ray beam. For whole brain imaging, three adjacent scans were acquired in the vertical direction.
In order to evaluate the performances of PPCI-EST and PPCI-FBP (propagation-based phase contrast CT using the FBP reconstructed algorithm from equal-angle projections), the same datasets should be used. Here, 400 equally sloped projections (N = 400) were extracted from 1553 CT projections according to equation (4). The errors induced by the EST extraction of the projections should be negligible [15].
Raw projection images were corrected with background images (flat- and dark- field). Then the background-corrected projections were processed using a phase retrieval algorithm to reconstruct the phases, which is based on the assumption that the ratio between the refractive index decrement and the absorption index (δ/β) in the sample is constant [19]. And then slices with phase information were reconstructed by FBP algorithm. All data sets were merged to obtain a total reconstruction for the whole brain. The EST algorithm was used to reconstruct slices from the 400 extracted equally sloped projections.
Image quality evaluations
For the evaluation of the performances of PPCI-EST and PPCI-FBP, the parameters such as the signal-to-noise ratio (SNR), the contrast-to-noise ratio (CNR), the spatial resolution (δ SR) and image quality were calculated. For each subject, manually traced regions of interest (ROI), which consisted of grey matter and white matter, were used for calculations.
The SNR and the CNR represented the parameters of tissue contrast. First, the histograms of the ROIs were obtained from the images. Second, Gaussian curve fittings were performed for the corresponding peaks of grey and white matter in the histograms, while the center positions (x ci ) and the variance of the corresponding peaks were obtained. The SNR was an important characteristic of CT imaging and was calculated by the equation SNR = |x ci |/σ i [20]. As an extension of the SNR, the CNR evaluates the differences in various tissue types, which is, in part, determined by the contrast in the images between the tissue types of interest. The CNR between two anatomical structures can be defined as CNR = 2|x c1 - x c2|/|σ 1 + σ 2| [21].
In addition to the CNR and SNR, another criterion, the spatial resolution δ SR, was considered to assess image quality. δ SR was determined based on the method of Fourier analysis [22]. A line profile of the ROI was obtained, and its power spectral density (PSD) was calculated. For more robust and full automation, a mean PSD was obtained from many line profiles in the ROI. The mean PSD converged towards a value that could be defined as the noise baseline. When the value of the mean PSD at the noise baseline was doubled and matched to the respective spatial frequency, it yielded the resolution.
The image quality factor (q) represents an overall evaluation of an image, which considers both the criteria contrast (CNR) and the spatial resolution; it was calculated using q = CNR/δ SR [22].
The SNRs, CNRs, δ SR and image quality factor of the cerebrum and the cerebellum for PPCI-FBP and PPCI-EST were calculated, respectively.
Results and discussion
Reconstructed slices and 3D rendering
Slices with phase information reconstructed by PPCI-FBP (a, b) and PPCI-EST (c, d) algorithm are shown in Fig. 2. Figure 2(a, c) and (b, d) are the slices of cerebellum and cerebrum respectively. The internal anatomical features of the brain can be visualized. From the direct observation, the characteristic morphology of the mouse brain was visible, and the grey and white matter tissues could be clearly distinguished in slices by both algorithms. Compared with slices by PPCI-FBP, there is slightly increased noise in the PPCI-EST reconstructed slices. In addition, there is a streak artifact visible in Fig. 2(c). Quantitative evaluations will be calculated below.
Figure 3(a, b, c) shows the coronal, horizontal and sagittal sections. The slices of the phase tomogram (refractive index slices) coincide with the mouse brain sections with myelin staining [23]. In the coronal slice, the major nerve tracts, such as the corpus callosum (cc), optic chiasm (ox), the anterior commissure anterior part (aca), and the dorsal fornix (df), are displayed in bright color, while the grey matter, such as the frontal cortex (Fr), is shown in dark. The sub-cortical nuclei, such as the caudate putamen (CPu) and the septo fimbrial nucleus (SFi), are also clearly visualized. In the horizontal slice, the sub-fields of Ammon’s horn of the hippocampus can be distinguished. For 3D imaging, all slices were rendered with the grey value using Amira software [24]. The 3D rendering and the segmented structures were shown in Fig. 3(d, e) as a false-color rendering. The cc and the major fiber bundles in the subcortex become evident. In the frontal view, the cc connects the two cerebral hemispheres. The fiber bundles of the ci pass through the CPu along the ventral-medial to the dorsal-lateral direction and reach the cc. The aca extends to both anterior sides. These structures observed in 3D are consistent with the results obtained from myelin staining.
Comparison between PPCI-FBP and PPCI-EST reconstructed slices
The SNRs and CNRs between the grey and white matter in the frames in Fig. 2(a, b) and (c, d) were calculated, which were related to PPCI-FBP and PPCI-EST, respectively (Table 1). It was demonstrated that the SNRs and CNRs for PPCI-FBP were close to but slightly higher compared with PPCI-EST. From this aspect, the PPCI-FBP has a slightly better tissue contrast. The results also indicated that the SNRs and CNRs of the cerebellum were higher compared with the cerebrum, which indicates the soft tissues in the cerebellum had a better image signal versus noise compared with the cerebrum. This finding explains the reason why a better image quality could be achieved for the cerebellum.
The δ SR values of the cerebrum and cerebellum slices corresponded to 37.50 μm and 39.13 μm for PPCI-FBP, respectively, and to 35.29 μm and 38.30 μm for PPCI-EST, respectively, as shown in Table 1. The results were confirmed with an alternative method in which the resolution was derived from a line profile taken along an edge in the same areas as previously described. The results indicated that PPCI-EST has a slightly better spatial resolution compared with PPCI-FBP for both the cerebrum and the cerebellum.
Using the CNR and its corresponding spatial resolution, quality factors of 0.032 μm–1 and 0.048 μm–1 were obtained for PPCI-FBP in the cerebrum and the cerebellum, respectively, and 0.028 μm–1 and 0.045 μm–1 for PPCI-EST were obtained in the cerebrum and the cerebellum, respectively. The results indicated that the image quality of the cerebellum was better compared with the cerebrum, which is supported by the direct observations shown in Fig. 2(a, b) and (c, d). The results demonstrated that the ∼73% reduction of the projections using PPCI-EST results in a deterioration in image quality of 6% for the cerebellum and 13% for the cerebrum. Therefore, the PPCI-EST results in a significant reduction in projection number while retaining acceptable imaging quality, which could reduce the radiation doses and data acquisition time.
Conclusions
X-ray propagation-based phase contrast imaging combined with equally sloped tomography (PPCI-EST) has been used for the nondestructive 3D imaging of the mouse brain. The experimental results have demonstrated that the structures, such as the corpus callosum, the optic chiasma, and the caudate nucleus, could be clearly depicted. The 3D structures of the whole brain were obtained. The results shows that this method has the ability to image extremely low contrast soft tissue samples. Compared with PPCI-FBP, PPCI-EST reduced the projection number by ∼73% , with a deterioration in the image quality of 6% for the cerebellum and 13% for the cerebrum. It could reduce the radiation dose and speed up the data acquisition process while simultaneously retaining sufficient contrast sensitivity and image quality. This method has the potential to be utilized for fast and low-dose phase-contrast 3D imaging to biomedical specimens.
Footnotes
Acknowledgments
This work is supported by CAS-CSIRO cooperation research project grant GJHZ1303, the Joint Funds of the National Natural Science Foundation of China (Grant No. U1232205), the National Nature Science Foundation of China (Grant No. 11405261, 11405260), the knowledge innovation frontier project of CAS and Youth Innovation Promotion Association of CAS (No. 2014235).