Background
Non-destructive whole-volumetric phenotyping studies through three-dimensional (3D) image reconstruction have gained increasing popularity among biomedical researchers over recent decades, especially with recent advancements in imaging acquisition and processing techniques. Three-dimensional visualization techniques may be divided into two categories: serial sectional image reconstruction and whole-volume imaging. The former includes confocal microscopy, episcopic microscopy [
1], and the latter includes optical projection tomography (OPT) [
2], micro-magnetic resonance imaging (micro-MRI) [
3,
4], and micro-computed tomography (micro-CT) [
5].
Whole-volumetric imaging is superior to serial imaging in many aspects. Serial imaging reconstructions such as confocal microscopy or confocal laser scanning microscopy (CLSM) can provide detailed and fluorescence-labelled representations of the sample, but of only limited thickness, typically <2–300 μm due to the need for optical transparency [
6]. Although image reconstruction may be employed to obtain the entire 3D view of the sample, the process is laborious and often with operator-dependent results. Episcopic microscopy overcomes this shortfall by adopting automatic serial slice alignment, and reconstructs relatively well-preserved images; however, such a process destroys the tissue sample post-imaging and prohibits further sample use [
1,
7]. Whole-volumetric imaging, on the other hand, does not have the above restrictions. For example, optical projection tomography (OPT) can be used to localize and measure structures within a whole organ, but only for restricted sample thicknesses [
8]. Micro-MRI can offer great soft-tissue contrast with adequate resolution images, but its use is limited by its restricted availability, high cost and extended scanning time [
4,
9,
10]. Micro-CT is now a widely accessible imaging modality, offering a mean for accurate visualization of three-dimensional structures, quantitative volumetric measurements, and tissue characterizations such as bone stress and vascular density [
11‐
15]. Although many are still unaware of this versatile tool, micro-CT has gradually becoming a popular whole-volume scanning research method, in large part due to its versatility for volume exploration, ease of tissue preparation, and quantitative analysis potentials.
Although many recent papers have published various staining techniques for successful micro-CT scanning of biological samples, including chicken embryo,
Xenopus embryo, mouse embryo, tumour angiogenesis, and other animal internal organs [
5,
16,
17], few study have clarified the scanning protocols for micro-CT scans on postnatal model animals or made direct comparisons between the ex vivo micro-CT, in vivo micro-CT, and histology scans of rat neuroanatomy. Furthermore, even though prior study has suggested staining with iodine or phosphotungstic acid (PTA) were efficient and effective for embryos, no comments have been made on their staining power for postnatal animals [
5]. In addition, the effect of iodine and PTA staining on subsequent microscopy processing has not been illustrated in the past. Lastly, we are also not aware of studies trialling image processing protocols on the rat brain micro-CT scans, which may facilitate future quantitative analysis.
Our study aims to complement previous studies by illustrating the following:
1.
Simple, safe, and effective staining method for successful micro-CT scans of both small and large rat brains.
2.
The potential neuroanatomical details of rat brain and image magnification and resolution achievable by micro-CT scanning using one of the most high-powered ex vivo micro-CT scanners available currently.
3.
Validate neuroanatomical information of micro-CT scan by direct comparison with H&E light microscopy.
4.
The difference in image quality between in vivo and ex vivo micro-CT scans of rat brain.
5.
Effectively improve image clarity and potentiate organs of interest through non-local means (NLM) denoising algorithms.
Discussion
Through proper micro-CT setup, sample preparation, and image processing, informative neural micro-CT scans can be generated with postnatal animals. These scans enable internal visualization and potential quantitative analyses including volumetric and dimensional measurements by providing images with neural-tissue differentiation comparable, if not better, to micro-MRI brain scans, as illustrated Denic et al. [
24]. Furthermore, micro-CT scans has a maximal spatial resolution of 1 μm/voxel with 16-h scanning time, significantly higher than 20 μm/voxel of micro-MRI achieved by 24 h of high-magnetic-fields scanning [
4,
25]. Consistently, this high-magnification power of micro-CT provides anatomical and histological details comparable to those of 4× H&E light-microscopy, Fig.
4b. As a result of these characteristics, 3D rendering of the ex vivo micro-CT scans can show detailed and refined micro-neuroanatomy, Fig.
3. This function can therefore be applied to morphological study on genetic-disease model animals or cancer studies. Moreover, because tissue sample can be preserved following micro-CT scanning for H&E processing, Fig.
4c and
d, micro-CT scan can serve as a screening tool for area of interest in the brain for further histology analysis. Although high-resolution micro-CT has the drawback of large dataset size, >4.6 GB/dataset, the high demand in computational and storage power for image analysis are met by the recent advancement in computing hardware, rendering this a practical research modality.
Although the difference between ex vivo and in vivo micro-CT has been addressed in previous studies, few have made direct comparisons of the quality difference of postnatal neural scans using these two modalities [
26,
27]. In this study, we demonstrated the difference in magnification and resolution potentials of micro-CT scans using a custom-built ex vivo micro-CT scanner versus a commercial in vivo micro-CT scanner, as shown by Fig.
5. Consistent with expectations, our results have shown that ex vivo micro-CT scans offer finer detailed scans than in vivo micro-CT scans, due to the higher magnification achieved by having the X-ray source closer to the sample (i.e., a short source-sample distance,
SSD). In theory, reducing the
SSD increases the image noises and blurs edges, an effect clearly shown in Fig.
5Ba and
Bb. However, image degradation can be minimized by using a low source current and careful focusing to minimize source spot size to obtain a best resolution of <2–3 μm/voxel while extending the scanning time to over 15 h to achieve high signal-to-noise ratio images. In comparison, commercial in vivo micro-CT scanners have limited magnifications and thus lower resolution due to the preset values of SSD. Furthermore, higher image noise in in vivo micro-CT scans is due to the shorter scanning time required to minimize the radiation exposure in live-animal scanning. In general, the best image quality needs the longest possible scanning time in either ex or in vivo micro-CT setups.
Based on the prior success of embryo study, we compared two contrast agents, iodine and PTA, for postnatal neural tissue staining [
16]. The contrasts were chosen for their safety, ease of handling, and the likely effectiveness based on prior studies. Perfusion was first trialled in the hope of preserving rat skull integrity. Unfortunately, iodine staining was ineffective when introduced this way, likely due to poor penetration through the blood brain barrier and slow diffusion rate through capillary walls to end-organs [
28,
29]. Contrary to prior embryo studies, simple diffusion staining with intact skull was also trialled with little success, suggesting contrast penetration through calcified skull was not possible [
16]. Hence, diffusion-staining through a small craniotomy was subsequently adopted and found effective using iodine staining on all tested tissues. However, PTA was less effective as a stain using the same protocols on partially encapsulated tissues, especially on large adult rat brains; staining remained incomplete after 2.5 years, Fig.
2d and
e, rendering this method impractical.
Due to concerns of prolonged staining may cause potential structural damage and prohibit tissues from future histology uses, as shown in Additional file
2: Figure S1, we tested PTA and iodine staining on dissected rat brains of small sizes with hope, Additional file
3: Figure S2. The result showed both stains yield similar quality of rat brain micro-CT images, as demonstrated by Fig.
2k and
n. However, iodine was able to complete tissue staining faster than PTA due to its much higher penetration. In addition, iodine was better in staining poorly vascularized tissues such as the central nervous system, where PTA had little success in staining tissues with size greater than 1 cm
3. This is likely due to the size and polarity difference of the two contrast agents: iodine is a non-polar molecule that is about 20 times smaller than the polytungstate anion of PTA [
30‐
33]. Nevertheless, we found staining times to be significantly longer than previously described [
5,
16]. To compensate for the lower penetration of PTA, limiting tissue volume to 1 cm
3 and complete organ isolation through dissection may be required for successful staining. While these modifications to the method have proven effective, adopting such an approach defeats the purpose of using micro-CT, since dissection and excessive handling increase the risks of structural damage, even by experienced dissectors, as illustrated by the micro-tears shown by micro-CT scans, Fig.
2h through
n. Based on our findings and the known difference in the properties of iodine and PTA, we recommend diffusion-iodine staining with partial craniotomy as the method of choice for micro-CT scans of postnatal rat brains. Additionally, successful H&E processing on both iodine- and PTA-stained brain tissues confirmed that prolonged contrast staining for micro-CT scanning compromises little, if any, tissue integrity and neuronal cell bodies can be visualized in higher magnification light micrographs, Fig.
4c and
d. This suggests micro-CT scans may serve as a targeting tool for regions of interest and reduce unnecessary histopathology processing in pathology study, thus reduce labour and time.
Post-acquisition image processing through NLM filtering was performed to enhance image clarity and structural boundaries. NLM algorithm was chosen for its property of spatial detail preservation [
21]. Although denoising only modestly improved ex vivo micro-CT scans, Fig.
6Aa and
Ba, due to the scan’s high intrinsic signal-to-noise ratio, it has proven useful for in vivo micro-CT scans, which has high intrinsic image-noises, Fig.
6Ca. This image-noise was significantly reduced after processing, Fig.
6Da. These improvements are obvious in selected magnified views of caudate putamen, Fig.
6Cb and
Db. We appreciate that commercially available in vivo micro-CT scanners are more widely available than custom ex vivo micro-CT scanners. By demonstrating the quality of in vivo micro-CT scans can be effectively improved through NLM algorithms, we support in vivo micro-CT data can also offer anatomical visualizations and accurate quantitative analysis. Lastly, denoising reduces the image noise, as demonstrated by the reduction in the variability of intensity profiles, Fig.
6Bc and
Dc; this is an important step for the future developments of accurate automated segmentation for organs of interests.
Conclusions
Micro-CT scanning, using either ex vivo or in vivo micro-CT scanners, is a powerful and effective neuroanatomy visualization modality for lab animals. Furthermore, micro-CT digital data is easy to store and manipulate. The anatomical details and accuracy offered by micro-CT scanning have been validated with the traditional H&E stained light micrograph. Moreover, the safety, simplicity, and tissue-preserving characteristic of the iodine-diffusion staining through craniotomy render micro-CT scanning a very easy-to-use study modality. Lastly, we appreciate that in vivo micro-CT scanners are more widely accessible and have a faster image processing time than ex vivo micro-CT scanners, the low signal-to-noise ratio due to the physical data acquisition restraints of in vivo micro-CT scans has been addressed. The improvement made by simple and efficient post-acquisition NLM processing reduces image degradation and enhances anatomical clarity, particularly with in vivo micro-CT scans, thus promoting versatile structural segmentation for quantitative analysis of animal morphology, including volumetric measurements. Based on described imaging methods, future studies may include: 1. internal organ visualization and quantitative analysis on animal models of genetic disease, e.g. Hirschsprung’s disease; 2. cancer treatment response studies on animal models; 3. identifying areas of interest for histological processing in pathology studies.