Open Access
Add to BibSonomy
Abstract
We present a new autofluorescence-imaging method for bone analysis. This method, based on the autofluorescence of bone, provides color images in microscopic scale. The color images are created from three monochrome images acquired with optimal excitation- and emission-wavelengths combinations. The choice of these combinations were determined from the study of two-dimensional distributions of bone-features-bispectral autofluorescence in the visible- and ultraviolet-spectral range. We demonstrate that main-bone features visualized with MG-staining method can also be visualized in the autofluorescence-color image. Furthermore, the autofluorescence-color image presents features hardly distinguished in a histological-bone section.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Bone is a complex and dynamic tissue, which has various interdependent features in macro-, micro- and chemical level structures. All these features affect individually the tissue properties mechanically and metabolically [1]. Bone-affecting diseases such as osteoporosis, systemic bone syndromes, infections, malformations or tumors can often lead to a pathologic fracture. Bone fragility can also be increased by medical therapies i.e. corticosteroid treatment and radiation therapy (RT) [2,3]. In addition to the fracture tendency, the inadequate bone quality can also lead to an impaired osseointegration of an orthopedic prosthesis or a dental implant [4]. Thus, bone analysis is important for research purposes and clinically in order to detect and treat viable bone to prevent bone fracture, or to obtain a long-term success in osseointegration of dental implants.
Various analysis methods for bone diagnostics have been developed [5–9]. Most common are the radiographic methods such as dual energy x-ray absorptiometry (DXA), native X-ray and computed tomography (CT). These imaging methods allow the evaluation of the bone macrostructure, mineralized bone volume (the volume of mineralized bone per unit volume of the sample to evaluate the variation in strength and stiffness of bone) or bone mineral density (density of mineral such as calcium in bones to estimate the bone strength). These are dominant markers in daily clinical practice to describe bone quality [10]. However, these methods require exposure to ionizing radiation and explain only part of the tissue quality. From the biomedical point of view, the bone quality can be assessed through the abilities of the bone matrix (mainly collagen I), abilities of the minieral component their heterogeneity, relation and constant dynamics (bone turnover) [10]. Nowadays, more interests has been developed to study the organic component such as collagen content provides more important information on fracture tendency [11–14].
Because of its composite tissue structure, the bone has to be assessed in both macroscopic and microscopic scale. The study of microscopic structure (histology), numeric microstructure evaluation (histomorphometry) and chemical composition of bone tissues provides precise information about the bone health [15,16]. However, all existing histological methods require staining of the bone to highlight particular features of interest [17]. Labelling methods introduce irreversible effects and variations in color in most of histologic samples [18].These variations create difficulty in image analysis [19]. In addition, not all features can be visualized by only one labeling method in the same section [17]. The commonly used staining method for bone analysis is Masson-Goldner trichrome which highlights the mineralized cortex, non-mineralized collagen and bone remodeling cells [20, 21]. However, visualization of some other properties or bone function, such as evaluating the rate of mineral apposition are visualizing by using a different method: the fluorochrome labelling methods [22]. In this case, several bone sections is needed to analysis bone. Histomorphometry, for example, requires two different sections: one with MG-staining and one with fluorescent markers. However, preparation of different bone sections is time consuming and requires experimented laborious work. Immunohistochemistry enables by multiplexing of antigens, to visualize several features in the same tissues section. However, the PMMA embedding can degrade some of the proteins and prevent label binding. For that reason, reliable immunohistochemistry on non-decalcified bone is challenging [23]. The number of tissue sections and laborious process can be reduced by using label-free methods. Label-free methods provides the possibility to utilize the same tissue section by another method for deeper analysis. Different label-free imaging methods was developed such as Raman or Fourier-transform infrared (FTIR) spectroscopy [24–26]. These methods allow the visualization of different bone features by providing the chemical map of the bone sample. Other label-free methods are based on the autofluorescence of bone, which was discovered already in 1967 [27]. Only recently, this bone properties has raised interest to develop tools for bone analysis such as bispectral autofluorescence spectroscopy [28–32]. However, these methods require a point-scanning of the bone sample to obtain an image, which can be time consuming [33,34]. The time consuming is reduced in the methods acquiring few autofluorescence images of bone without scanning the sample [35,36]. The existing methods use filters and monochrome or RGB camera to obtain enhanced bone features image in fast way. However, all these methods utilize wide spectral-bands for the detection, which leads to the visualization of only few features. In addition, the combinations of excitation and emission spectral bands, which has been reported for imaging tissue, are not optimum for bone analysis. None of them is useful for native bone analysis in fast way.
In this paper, we propose the optimal combinations of excitation- and emission-narrow-spectral bands (wavelengths), applicable for various bone imaging perspectives (quantitative and qualitative imaging). First, these combinations can be utilized to acquire enhanced color image of native bone section in fast and simple way. The color image provides wide enhanced tissue overview of bone tissue. This modality is applicable especially for traditional histopathology or quantitative histomorphometry. In addition, the color image can be obtained without scanning the sample, which leads to a faster acquisition than spectroscopic methods. Furthermore, in contrast to staining or fluorochrome labeling methods, the proposed method based on the bone autofluorescence does not require chemical labelling, which reduces considerably the laboratory work and time in both clinical and research practice. Only polymethyl methacrylate (PMMA) embedding, xylene and DePex solution is applied to the bone for sectioning bone into a thin section and mounting it on the microscope slide. Secondly, this bone preparation provides the possibility to visualize several interesting features in the same bone section while enables the possibility to utilize the same section for deeper analysis by another method such as FTIR for chemical analysis for example. As mentioned previously, the most important features regarding the bone strength and histopathology, are the collagen structure and content, mineralized collagen fibrils and cellular components for remodeling. The collagen and cellular components are autofluorescent. Therefore, autofluorescence method could well be considered for bone quality imaging [28]. The method is faster than spectroscopic methods since the sample does not need to be scanned and only three monochrome images are acquired. The acquired enhanced color image is consisted of only three monochrome images acquired with different bone excitation and emission narrow-spectral-bands combinations optimum for bone analysis. The choice of these monochrome images were determined from the two-dimensional distribution of the bispectral autofluorescence of bone tissues (i.e. two-dimensional distribution of the spectral fluorescence emitted from the bone components under illumination of different wavelengths). The color image consisted of images obtained with optimal narrow excitation and emission spectral bands, which demonstrated the full histology of bone tissue, has not been previously applied to bone.
2. Methods and materials
2.1. Materials and samples preparation
This study consists of bone samples of 34 patients with different states of health (healthy, cancer patient treated with ablative surgery without RT and cancer patients treated with RT). All the patients had given their consent to participate to the study and the work has been approved by the ethical committee: Medisch ethische toetsingscommissie (METC) in VU University medical center of Amsterdam in Netherland (2011/220).
The bone samples were harvested from the dental implant beds at the time of implant surgery in VU University Medical Center of Amsterdam (Netherland). A biopsy of 10 × 35 mm was processed in the alveolar bone of the mandible and immerged in 70% of ethanol. The bone samples were prepared as native non-stained samples according to the standard resin embedding and sectioning protocol in order to be observed by using a microscope. The samples were dehydrated with increasing the concentration of ethanol and embedding in polymethyl methacrylate resin (PMMA). The embedding in PMMA, due to its hardness close to bone hardness, provides the possibility of bone sectioning into thin slices of 5 to 10 μm of thickness without decalcifying the bone. The embedded samples were sectioned in thin layers (5 μm of thickness) by using a microtome (Polycut S, Reichert-Jung, Nußloch, Germany).
The sectioned bone samples were fixed on a microscope slide by using a Haupt solution (consisting of gelatin, phenol mastic, glycerol and water). The sectioned samples, fixed on the slide, were dipped in a xylene and DePex solution and covered by a coverslip. The xylene and DePex solution is a refractive-index-matching material in visible and ultraviolet spectral range (i.e. the refractive indexes of the bone, coverslip and the solution are closed to each other in the spectral range of the visible and ultraviolet light). This solution is used to fix the coverslip on the bone sample while providing a good clarity and contrast, since the small difference between refractive of different media (bone, solution and coverslip) decreases the light reflection, and light bending at the transition between different media. The sample is covered by a coverslip (MENZEL-GLÄSER 24 × 32 mm #1, thickness: 0.13–0.17 mm) enabling the sample illumination by ultraviolet light (UV) of wavelengths lower than 360 nm.
2.2. Experimental set-up
2.2.1. Autofluorescence-imaging set-up
The autofluorescence of bone samples was imaged by using an Olympus’ microscope (Olympus BX63) in epi-illumination configuration. The autofluorescence imaging set-up is shown in Fig. 1.
A short-arc xenon lamp (PE300 BUV) is coupled with angle-tunable band-pass filters (VersaChrome Edge thin-film filters technology, Semrock) mounted in an angle-tunable filter wheel (Tunable Filter Changer: Lambda VF-5TM, Sutter instrument Company) controlled electronically. A set of five angle-tunable band-pass filters of narrow spectral bandwidth (20 nm) are used to illuminate the bone sample with light of narrow spectral band. The central wavelength of each filter can be set with a precision of 1 nm, which makes it possible to create a sequence of illuminations of different central wavelengths from 338 nm to 620 nm shifted of 1 nm from one to another. The light of the sequence, transmitted by an optical guide (VIS380), is reflected by a dichroic-beam splitter (Di01-R355-25×36, Semrock) to the sample. The dichroic-beam splitter reflects light of wavelength less than 385 nm from the light source to the sample and transmits light of wavelgnth more than 385 nm from the sample to the band-pass filter. The light is focused on the sample via a 20-x microscope objective.
The light emitted from the sample is transmitted by the dichroic-beam splitter to a band-pass filter of bandwidth comprised between 14 nm and 18 nm. The band-pass filter between the camera and the beam splitter, allows the selection of the emission wavelengths (wavelengths of the light emitted by the sample) to be recorded. The band-pass filter was changed manually for bone visualization at different emission wavelength. A total set of 21 band-pass filters were used to acquire the emission spectra of the bone. The monochrome camera, sensitive to the electromagnetic radiation in the spectral range from 345 to 1,000 nm, acquires the bispectral-autofluorescence images of the bones. The camera is water cooled at −100°C to obtain a low-dark-current noise. The sensor resolution is 1,024 × 1,024 pixels. The sampling resolution of the acquired images is 0.68 μm/pixel. The exposure time (10 s) was chosen so that the camera’s dynamic range was optimally used while avoiding pixel saturation. In addition, the exposure time has to be the smallest possible to reduce the destructive effect due to UV illumination. The moving stage in x, y directions, allows the selection of the sample part to be imaged. The image focusing was obtained by moving the sample along the z-direction.
2.2.2. Stained-sample-imaging set-up
For identification of different features observed in the fluorescence image, the imaged-bone samples are stained with Masson-Goldner (MG)-staining by using a standard protocol [37]. Afterwards, the stained-bone samples were imaged by using a LEICA-zoom macroscope (LEICA Z6 APO). The magnification was 9-x, which provides a sampling resolution of 0.38 μm/pixel. The macroscope was used in transillumination mode as it is used usually for observations of MG-stained bone sample by the histologists. The bone samples were transillumined with light from a white LEDs panel of homogenous illumination. The color images of the stained samples were acquired by a RGB CCD camera (DFC 450c) of sensor size: 2,560 × 1,920 pixels. The white balance was processed without the bone sample under the objective. Several images of the bone-sample part were acquired at different focus. These images were stacked together by using the z-stacking of the LEICA software (Leica Applications Suite) to obtain a bone-sample image in focus for any part of the bone image.
2.3. Image acquisition and processing
2.3.1. Bispectral-autofluorescence-image acquisition
The bone samples were illuminated with a sequence of nine narrow-spectral-band lights of different central wavelengths (excitation wavelengths) from 340 nm up to 380 nm in 5 nm steps. For each illumination (excitation) of the sequence, the light emitted from the bone sample was filtered by the different narrowband band-pass filters (of central wavelength from 390 to 720 nm) in sequence. Each filtered light emitted by the sample, reaches the camera sensor which provides a monochrome image for each excitation- and emission-wavelengths combination. These acquired images form the bispectral autofluorescence image of the bone sample affected by the spatial and spectral uniformity of the system. The spectral uniformity of the system (Corr(λemc,λexc)) is described in Eq. (1).
Here, λemc and λexc represent the central wavelengths of the band-pass filter and the excitation spectrum, respectively. λem is the emission wavelength. Pe(λexc) is the spectral optical power of the light source (Short-arc xenon lamp coupled to an angle-tunable filter). Tex(λexc) and Tem(λem) are the transmittance of the optical elements situated in the illumination and imaging part of the imaging system. TPB(λem,λemc) is the transmittance of the band-pass filter of central wavelength λemc and η(λem) is the spectral efficiency of the camera.
Additionally, the system presents a spatial non-uniformity due to filters, lenses and uneven sample illumination. To determine the spatial uniformity of the system, an image of uniform and smooth-surface sample were acquired at each filtered-light emitted from these reference samples and each excitation wavelengths. The reference samples were chosen to emit light in the spectral range from 380 to 760 nm when they are excited by UV light in the spectral range (from 340 to 380 nm). The emission spectra of the reference samples overlap to obtain an image of the spatial uniformity for each wavelength in the range of 390 to 720 nm.
The bispectral autofluorescnce of the bone sample (ImgFluo(λemc,λexc,x,y)), obtained by taking in account the spatial and spectral uniformity of the system, is described in Eq. (2).
Here, Img(λemc,λexc,x,y) is the raw bispectral image. Dark(λemc,x,y) is the average-dark image. Corr(λemc,λexc) and Ref(λemc,λexc,x,y) are the spectral- and spatial-non-uniformity of the system in a function of the central wavelength (λemc, λexc), of the band-pass filter transmittance and illumination spectra, respectively.
2.3.2. Monochrome-Images selection
A color image, represented in the RGB (Red, Green and Blue) model, consists of three channels. Each channel is a monochrome image. This set of the three monochrome images are uncorrelated when color is observed in the resulting color image. The most uncorrelated set provides the best color contrast. The uncorrelated set of bone images can be consist of images acquired at different emission- and excitation-wavelength combinations, since the features, which are the best enhanced in an image acquired at a specific wavelength combination, are different at another wavelength combination. The uncorrelated set should contain monochrome images of high quality since images of low quality leads to a granulate color image, which provides difficulty to detect the bone features from its background. The ability to detect features from its background (neighboring pixels) is determined by the signal-to-noise ratio (SNR) [38] calculated according to the Eq. (3).
The signal is the level difference between the average-pixels intensity Img(λexc,λemc,p) of the feature of interest and the average-dark image Dark(λemc,p) (average image of images acquired without illumination). The noise (N) is the standard deviation of the intensity fluctuation in the dark images and p is the pixel number. M is the quantiy of pixels from the bone feature in the image acquired at the excitation wavelength λexc and the emission wavelength λemc.
To determine the images of high quality from the bispectral image, the SNR is calculated for each images acquired at different wavelengths combination (emission and excitation wavelengths) for different features of interest. The SNR superior to 3 dB (criteria considered as the boundary between low and high SNRs by engineers) indicates the images of high quality for the feature of interest. Most of the main features, possessing a SNR superior to 3 dB, are observed in the images acquired with an exposure time of 10 s and a 20-x microscope objective, at an emission wavelength in the range from 390 to 620 nm and an excitation wavelength comprised in the range of 340 to 380 nm. The images acquired at a higher emission wavelength than 620 nm become noisy. Therefore, the set of images candidate for the color-image creation are acquired at the emission wavelengths from 390 nm to 620 nm and excitation wavelengths from 340 nm to 380 nm.
The most uncorrelated images, from this candidate images, are determined by the lowest correlation coefficient of intensities averages of the features and their surrounding pixels, between two images pair. The correlation coefficient is calculated for all images-pair combinations from the group of high quality images. The pair of images possessing the lowest correlation coefficient are the images selected for creating the color image. The third image of the color image is the image from the set, which provides the lowest correlation coefficient with the two previously selected images. An example of a color image and its monochrome images in its Red, Blue and Green channel are shown in Fig. 2.
Fig. 2 Monochrome images of native-healthy bone in the Blue, Green and Red channel of the autofluorescence-color image (d), shown in (a), (b) and (c), respectively. Different bone features (the cortex (Ct), osteoid (O), osteoblasts (Ob), osteocytes (Ot), erythrocytes (RBC), connective tissues (CT)), are indicated by arrows in the autofluorescence-color image (d) and in the true-color image acquired after bone staining with MG-staining (e).
The color image of Fig. 2 (d) is created to obtain the erythrocytes, commonly known as the red blood cells (RBC), the best color contrasted with the surrounding features such as the cortex (Ct), the osteoblasts (Ob) and the connective tissues (CT). The two most uncorrelated images are the images acquired at the emission wavelengths of 390 and 560 nm and the excitation wavelength of 340 nm. As it can be seen in Fig. 2, the erythrocytes (RBC) are not distinguished from their surrounding features in the image acquired at the emission wavelength of 390 nm (Fig. 2 (a)). However, they are strongly distinguished in the image acquired at the emission wavelength of 560 nm (Fig. 2 (c)). The third image of the color image is the image acquired at the emission wavelength of 500 nm and excitation wavelength of 340 nm (Fig. 2 (b)).
The combination of these three images provides the erythrocytes (RBC), represented in canary yellow in Fig. 2 (d), color contrasted with their surrounding features such as the connective tissues (CT) in black chocolate color, the osteoblasts (Ob) in camel color, the cortex (Ct) in steel blue. In this figure, other features can be distinguished such as the osteoid (O) in sand color, the osteocytes (Ot) in camel color and the artifacts due to tissue folder in white. The MG-stained images (Fig. 2 (e)) of the bone allows the identification of most of the features observed in the color-fluorescence image.
The selection of images acquired from other emission- and excitation-wavelength combination, for creating the color image, allows the enhancement of other features, which are barely visualized with MG-staining. However, they can be clearly visualized with other methods such as different structures of the cortex (Based Structural Units : BSUs) which are clearly visualized with a third-harmonic generation microscope [39] or the lacunae (Lc) in cortex which can be obviously visualized a brightfield configuration, observed in the Fig. in 3 (a) and Fig. 3 (c), respectively. To identify the lacunae in the autofluorescence color image, a monochrome image of the same bone section trans-illuminated with halogen lamp light were acquired.
Fig. 3 Color images of a native-heathy bone enhancing different cortex structures (BSU1 and BSU2) and the lacunae (Lc) in (a) and (c), respectively. Images of the MG-stained bone, area represented in the fluorescence images (a) and (c), are shown in (b) and (d)).
The best images combination for the visualization of the cortex microstructure is the images acquired at the emission wavelengths of 420 nm, 540 nm and 620 nm, while the best combination for the visualization of the lacunae with or without their osteocyte, in cortex, is another set of images (images acquired at the emission wavelengths of 420 nm, 434 nm and 620 nm and at the excitation wavelength of 340 nm) for creating the color image. The stained-bone images shown in Fig. 3 (b) and (d) are the images of the same part of the bone showing autofluorescence in Fig. 3 (a) and (c). The stained-bone images allow the identification of the different features observed in the fluorescence images, such as the cortex represented in ocean green in the stained-bone image, the lacunae (Lc) in white and the osteocytes in Royal purple.
2.3.3. Color selection
It is well know that the choice of color has to be complementary colors (opposite colors in the hue circle of the representation HSV) for features, which are nearby each other (e.g. green and red, blue and orange or violet and yellow). The green and red is not chosen because the colorblind persons will not distinguish them. Therefore, the blue and orange-yellow is preferred. To visualize a feature in blue and another in yellow color, the blue feature has to possess the highest intensity value in the Blue channel and the lowest in the Green and Red channels of the RGB image. In contrast, the yellow feature has to possess the lowest intensity value in the Blue channel and the highest in the Green and Red channels. In the case of enhancing the erythrocytes in the color image, the cortex (the blue feature) presents the higher pixel intensities than the erythrocytes (the yellow feature) for the image acquired at the emission wavelength of 390 nm. Therefore, this image is in the Blue-channel, and the images acquired at 500 nm and 560 nm are in the Red and Green channels of the RGB image. However, the low intensity values of the cortex in the Blue-channel image and of the erythrocytes in the Red- and Green-channel images, prevent the visualization of these features in blue and yellow. To increase these values, the image of each channel is transformed according to the Eq. (4).
ImgFluo(λemc,λexc,x,y) is a monochrome image of a color-channel which is in a function of the emission and excitation wavelengths (λemc and λexc). The coordinates x, y represent the pixel position in the image. The maximum (b(λemc,λexc)) and minimum (a(λemc,λexc)) intensity values are defined such as 99 % and 1% of pixel from the features of interest possess a value inferior to these intensity levels, respectively.
Thus, the intensity values of the blue feature in the Blue channel and of the yellow feature in the Red and Green channel are close to the maximum of the dynamic range (216 for 16-bits-grayscale image), while the intensity values of the yellow feature in the Blue channel and the blue feature in the Red and Green Channels are close to zero. Thus, the image for enhancing erythrocyte, transformed thus, presents the erythrocytes in yellow and the cortex in blue color (Fig. 2 (d)).
3. Results
3.1. Bispectral-autofluorescence image
The bispectral-autofluorescence image consist of monochrome images acquired at different excitation wavelengths and emission wavelengths. A representation of a bispectral-autofluorescence image of a heathy-bone part is shown in Fig. 4.
Fig. 4 Two-dimensional distribution of bispectral autofluorescence of a healthy-bone sample. The bone features are indicated by arrows for the cortex (Ct), osteoblasts (Ob), osteocytes (Ot), connective tissues (CT), endosteum (Es), erythrocytes (RBC), osteoclasts (Oc) and osteoid (O).
The pictures in row and in column represent the autofluorescence intensities of bone tissues in a function of the emission and of the excitation wavelength, respectively. Here, only images acquired at an emission and an excitation wavelength in the range of 390 to 720 nm with a step of 100 nm, and of 340 nm to 380 nm with a step of 20 nm are shown in Fig. 4. The images acquired at an emission wavelength superior to 700 nm are noisy (SNR < 3 dB).
It can be noticed from this 2D distribution, that different excitation- and emission-wavelength combinations provide images of different pattern for the same bone area. For example, in Fig. 4, the osteoid (O) is not distinguished from the mineralized cortex at the excitation-/emission-wavelengths combination of 340 nm / 390 nm. However, it is strongly distinguished at the emission wavelengths superior to 520 nm. The osteoblasts (Ob) and osteoclasts (Oc) beginning to be visible at the emission wavelength of 475 nm when the excitation wavelength is 340 nm. However, they can be distinguished earlier at 420 nm at a higher excitation wavelength (360 nm). The erythrocytes (RBC) are strongly distinguished from the connective tissues at the combination 360 nm / 520 – 600 nm. In Fig. 4, it can be seen that the erythrocytes are highly distinguished at the emission wavelength of 520 nm and at the excitation wavelength of 360 nm than at the excitation wavelength of 340 nm.
These different excitation-emission profiles are due to different biomolecular background of the bone components. They all possess a different substance-characteristic signature. The Donaldson matrices of some main-bone features are shown in Fig. 5. These Donaldson matrices are the results of the pixels intensities of a disk area of eight-pixel diameter, averaged for each image combination. Each areas corresponds to a bone feature: cortex, osteoid, erythrocyte.
Fig. 5 Donaldson matrices of healthy-native-bone-tissue components: mineralized cortex in (a), osteoid in (b) and erythrocytes in (c).
As it can be seen, all these features present an intensity peak at an excitation- and emission-wavelength combination of 340 nm / 390 nm. This peak corresponds well with the collagen content since most of the organic component of bone possess collagen type I [27–29, 35]. However, the intensity slopes are more abrupt for some features than others are. For example, the osteoblasts present an emission-wavelength slope softer than the cortex’s slope. Therefore, the intensity variations at the excitation and emission wavelengths (Donaldson Matrices) are different from one feature to another, which enables the identification of the features.
3.2. False color image
Most of the main-bone features can be distinguished by creating a false-color image from the bone-autofluorescence images acquired at the excitation wavelength of 340 nm and at the emission wavelengths of 390 nm, 500 nm and 560 nm. This false-color image (RGB image) and its true-color image of the native and of the Masson-Goldner-stained bone, respectively, are shown in Fig. 6. The true-color image of the stained bone was used to identify the observed features in the false-color image.
Fig. 6 Autofluorescence image of a native bone in false color in (a) and its RGB image, after staining, represented in true color in (b). Main-bone features (osteoblasts Ob, osteoid O, osteocytes Ot, cortex Ct, connective tissues CT, eroded surface E, osteoclasts Oc and erythrocytes RBC) are indication by arrows
As it can be seen in Fig. 6 (a), it is possible to distinguish and identify the principal tissue components of the bone: cortex (Ct), osteoid (O), osteoblasts (Ob), osteoclasts (Oc), lacunae (Lc) with their osteocyte (Ot), erythrocytes (RBC), connective tissues (CT) in the same sample. These features are also visible by using the traditional staining technique, here by the Masson-Goldner staining (Fig. 6 (b).
Additionally, several other bone components and structures were observed in other imaged bone samples, which are hardly observed in this figure. These observed components and their optimal-imaging-wavelengths combination are listed in the following Tables 1–2. The features observed in both images (the false-color image of native bone and the true-color image of MG-stained bone) are presented in Tables 1. The quality of distinction is evaluated in 4-steps scale between clearly visible (++) and non-visible (–).
Table 1. Observed features common to images obtained with both imaging methods
Table 2. Observed features in autofluorescence image, barely contrasted in MG-stained bone image
The second column of the Table 1 indicates the color of the feature observed in the false-color image created for the most-main-features visualization. The monochromes images, in the Red, Green and Blue channel of this color image, are acquired at the excitation wavelength of 340 nm and the emission wavelengths of 560 nm, 500 nm and 390 nm, respectively. The emission wavelengths of monochrome images, which provide a color image with better enhancement of the feature, are shown in the fifth column. The fourth column indicates the excitation and emission wavelengths needed to obtain the gray-scale image of the bone with the best contrast between the feature of interest and its neighboring pixels (its background). This image is determined from the calculation of the Michelson contrast C between the feature and its background. [40]. The Michelson contrast is defined in Eq. (5).
Here C is the contrast value between the feature of interest of averaged intensity If and its neighboring pixels of averaged intensity Ibck. The image of wavelengths combination, which possess the highest contrast value, is the image chosen to present the best contrast between the feature and its background.
The mineralized cortex is visualized by using both methods, in blue, in the fluorescence image, optimized to distinguish the main-bone features and in green, in the stained-bone image. In both images, the feature contrast is high in relation to other tissue components. The cortical lamellae and the osteon structures are visualized better in the autofluorescence image. In addition, the choice of images at other wavelengths to construct the color image, provides a color image with higher color contrast between the features of interest. For example, the best contrast between the cortical lamellae and the osteon structures is visualized in the color image formed by the monochrome images acquired at the emission wavelengths of 420 nm, 540 nm and 620 nm. The monochrome image which presents the best contrast between these features is obtained at the excitation at 340 nm and at the emission at 390 nm. The cortical lamellae and the osteon structures are hardly visible by the staining methods. The centers of active ossification are visualized by using both methods. The sand color of the border sites, in the autofluorescence image, corresponds to the sites where the osteoid is present. This border is contrasted strongly in the monochrome image acquired at the excitation of 360 nm and the emission of 720 nm. The sites of inactive ossification or bone resorption (eroded surfaces) are indicated equally as an absence of cortical border fluorescence or absence of osteoid. These surfaces, dependent structures, are the sites where the cortex meets directly the connective tissue, after the resorbing osteoclast activity. The site registration is important in terms of bone-resorption-rate assessment. The loose connective tissues (bone marrow connective tissues) and the dense connective tissues (endosteum, periosteum) are visible in both images (autofluorescence- and stained-bone images). However, an experienced researcher is required to separate the medium purple color of the connective tissue from the adjacent slightly different purple of the osteoid in the stained-bone images. This can be a challenge in case of stain excess. The adipose tissues in bone marrow is seen as large circular hollow spaces separated by cell walls (cytoplasmic membrane) and connective tissue walls in both stained- and autofluorescence-bone images. The cellular components (the erythrocytes, osteoblasts, osteoclasts, osteocyte and bone marrow cells) are visible by using both methods. These features can be distinguished in the fluorescence image created for the main-features distinction. A very strong fluorescence is emitted from the erythrocytes at the emission wavelength superior to 475 nm especially at 600 nm compared to the other compounds. These cells are not as evidently observed in the stained bone images.
The autofluorescence images can enhance the bone features, which are hard to distinguish in the MG-stained-bone images acquired by the traditional method. The Table 2 lists the features, which are observed only in the fluorescence image. However, these features are hardly distinguished in the MG-stained bone image. One of these features is the bone basic structural units (BSU’s), which can be enhanced in the color image formed by the monochrome images acquired at the excitation wavelength of 340 nm and at the emission wavelengths of 420 nm, 540 nm and 620 nm in the Blue, Green and Red channel of the color image. The cement lines between the BSU’s are visible in the monochrome image acquired at an excitation at 340 nm and at an emission at 620 nm (Fig. 7 (a)). In addition, it can be observed in the color image created from the monochrome images acquired at the excitation wavelength of 340 nm and at the emission wavelengths of 390 nm, 500 nm and 560 nm. The BSU’s and the cement lines are barely seen in the stained bone image. Moreover, the fluorescence image presents borders around the lacunae, whereas the staining does not emphasize the lacunar borders (Fig. 7 (b) and (c)).
Fig. 7 Autofluorescence image acquired at the emission and excitation wavelength of 620 nm and of 340 nm with 20-x microscope objective in (a) showing cement line (Cm). Autofluorescence-color image acquired with a 40-x microscope objective, at the emission wavelengths of 390 nm, 500 nm and 560 nm with an excitation at 340 nm, showing the cortex (Ct), the lacuna (Lc) with its osteocyte (Ot) and without its osteocyte in (b) and (c), respectively.
Most of the features observed in the stained-bone images are also visualized in the false-color images of the native bones (Table 1). Some features are clearly visualized in both images and some only obvious in the autofluorescence images (Table 2). Actually, the staining does not yield additional information and rather involves a potential risk of staining artifacts. The autofluorescence images provide to reduce the laborious work, time, and difficulties in bone analysis due to staining color variation. However, the autofluorescence images are more affected by dust and air bubbles than the stained-bone images. The sample preparation needs to be performed with extra caution for the fluorescence imaging.
4. Discussion
Color image acquired with optimum narrowband-light excitation and emission has not previously been reported, neither applied for bone imaging. In the narrowband-autofluorescence imaging method (NABFI), the optimum monochrome images consisting the color image, are acquired with narrow-spectral-band excitation and emission combinations optimum to enhance the color contrast between bone features. The optimum monochrome images were determined from the study of the bispectral-autofluorescence image of bone. The bispectral-autofluorescence image was only a preliminary study to track the tissues-full-spectral profiles. After the optimal-light-wavelengths determination for the color-image creation, the bispectral-autofluorescence-image acquisition is not necessary. Only three optimum monochrome images neet to be acquired to obtain the color image, which reduces the acquisition time (10 s per image for 20-x microscope objective).
The color images obtained with the NABFI method enable the identification and localization of several mineralized bone features in bone samples. The BSU’s, observed in the color image created from monochrome images acquired at the emission wavelengths of 420 nm, 540 nm and 620 nm and excitation-light wavelength of 340 nm, are formed through the bone remodeling over the age. It has been previously suggested, that the bone fluorescence changes with the bone age [27]. The oldest bone emitted more light at 450 nm when they are excited at 360 nm than the youngest bone. Theses difference is also observed in our study at this combination of wavelengths. However, the contrast between these two different BSUs is the highest for the image acquired at the excitation and emission wavelengths of 340 nm and 390 nm, respectively. The BSUs are circled by thin cement lines. These lines possess different composition (relatively high mineral content and low collagen content, non-collagenous proteins (NCP’s), such as the osteopontin [41] than the BSUs [42,43]. This different composition enables the distinction of cement lines from the BSUs in the autofluorescence images. In the image acquired at the wavelengths combination 340 nm and 390 nm (Fig. 6 (c)), different BSUs emit light at different intensity levels. The cement lines, in this image, cannot be distinguished from the BSUs since the cement line emits light at the same intensity level than the brighter BSU. The cement lines are highly contrasted from the BSUs in the image acquired at the excitation and emission wavelengths combination of 340 nm and 620 nm. In the image, shown in Fig. 7 (a), it can be observed that the cement lines emit more fluorescent light at 620 nm than the BSUs, which confirms that they possess different composition.
The mineralized cortex is a microscopic-composite structure of the mineralized and differently mineralized collagen fibrils. The orientation angles of the fibrillary structures appear as cortical lamellae [1]. In a routine practice, these lamellae have been observed from native sections under polarized light microscope, which requires preparation of a native section adjacent to the stained section. In the autofluorescence-color images, creating for the main-features distinction, the lamellae were clearly visible from the same sample section without any need for additional native section as with staining.
The cellular components (the osteoblasts, osteoclasts, osteocytes, erythrocytes, fibroblasts and adipocytes) can be also observed since they possess some fluorescent compounds such as NADH/NADPH-coenzyme derived in different quantity [44]. The diagnostically relevant cellular observations are the osteoid-related osteoblasts indicating the active ossification, osteoclasts resorbing the bone, and osteocytes for the lacunar vitality count [16]. The lacunar vitality has been previously assessed according to the presence or non-presence of an osteocyte in the lacunae. Since the lacunae, osteocyte and mineral cortex possess different fluorescent compounds, they can be distinguished in the fluorescence image (Fig. 7). In the fluorescent images, an additional thin border lining the lacuna was observed. It has been proposed that the lining could be the structures of the osteocyte sheath [35]. In the necrotic specimen, the lacunar de-vitality increases [45]. It seems that this lining did not occur in the necrotic specimen. However, the magnification and resolution used in this study was not good enough to assess reliably the structure of the lining in the necrotic samples.
In addition, the borderline at the locations of the osteoid possess a clear fluorescent line in the monochrome image acquired at the emission wavelength of 620 nm and in the color image created from the images acquired at the excitation wavelength of 340 nm and emission wavelengths of 390 nm, 500 nm and 560 nm. In this color image, the osteoid forms clear lines of different fluorescence properties that appear as different color zones in the fluorescence-false-color image. Contrarily, in the stained-bone images, the osteoid is a relatively homogeneous area, which is sometimes difficult to separate from the connective tissue. The molecular mechanisms of the osteoid maturation into mineralized bone are relatively undiscovered. However, the structure in the osteoid zone goes through several conformation changes influenced by various osteogenic inductors [43], which can lead to different fluorescence properties. The autofluorescence imaging could be a useful tool in the study of the ossification process.
The connective tissue observations diagnostically is mainly relevant if it is compared to the relative area of the mineralized cortex. In certain pathologic states, i.e. preclinical osteoradionecrosis, the relative amount of fibrotic connective tissue is increased. The connective tissues could be devised into two forms: dense connective tissues (endosteum, periosteum) and loose connective tissues. Both of them are observed in the autofluorescence-color image and stained-bone images.
5. Conclusion
The choice of optimal excitation-emission-wavelength combinations to acquire a color image, provides enhanced bone-features images useful for bone research and diagnostics. The acquisition of the color image from three optimized monochrome images is fast and simple.
In addition, the acquisition of monochrome images of the autofluorescence of native bone provides the possibility to use the same bone section for deeper analysis by other method, since this method does not require any bone labeling. This reduces the quantity of bone section and time needed for analysis, as well as artifacts due to the staining process.
These obtained color images provide an overview of the bone microarchitecture, which covers textural, structural and chemical information. All the features observed in MG-stained bone images are observed in the bone-autofluorescence image (centers of remodeling, mineralized cortex, connective tissue and cellular components). Furthermore, the cortical lamellae, lacunar vitality, BSU’s and cement lines are better visualized in the autofluorescence-color image than in the traditional MG-stained-bone image.
Hence, the narrow-band-autofluorescence imaging with optimum spectral bands is a potential method for bone diagnostics. Further studies are required for deeper analysis of variation in autofluorescence signal at different tissue sites, especially regarding the effect of different forms of collagen folding and mineralization.
Acknowledgments
The authors acknowledge Olympus Corporation for lending the materials needed for the research and Ritva Savolainen for bones preparation.
Disclosures
The authors declare that there are no conflicts of interest related to this article.
References
1. S. Weiner and H. D. Wagner, “The material bone: structure-mechanical function relations,” Annual Review of Materials Science 28, 271–298 (1998). [CrossRef]
2. L. L. Furstman, “Effect of radiation on bone,” Journal of dental research 51, 596–604 (1972). [CrossRef] [PubMed]
3. A. Unnanuntana, B. J. Rebolledo, M. M. Khair, E. F. DiCarlo, and J. M. Lane, “Diseases affecting bone quality: beyond osteoporosis,” Clinical Orthopaedics and Related Research® 469, 2194–2206 (2011). [CrossRef]
4. S. Ihde, S. Kopp, K. Gundlach, and V. Konstantinović, “Effects of radiation therapy on craniofacial and dental implants: a review of the literature,” Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, and Endodontology 107, 56–65 (2009). [CrossRef]
5. D. Chappard, M.-F. Baslé, E. Legrand, and M. Audran, “New laboratory tools in the assessment of bone quality,” Osteoporosis International 22, 2225–2240 (2011). [CrossRef] [PubMed]
6. A. Dhainaut, M. Hoff, U. Syversen, and G. Haugeberg, “Technologies for assessment of bone reflecting bone strength and bone mineral density in elderly women: an update,” Women’s Health 12, 209–216 (2016).
7. T. Link, S. Majumdar, S. Grampp, G. Guglielmi, C. Van Kuijk, H. Imhof, C. Glueer, and J. Adams, “Imaging of trabecular bone structure in osteoporosis,” European radiology 9, 1781–1788 (1999). [CrossRef] [PubMed]
8. W. T. Trouerbach, J. L. Grashuis, A. W. Zwamborn, E. C. Clermonts, and H. J. Schouten, “Microdensitometric analysis of bone structures in x-ray images,” Skeletal radiology 16, 190–195 (1987). [CrossRef] [PubMed]
9. P. Guggenbuhl, F. Bodic, L. Hamel, M. Baslé, and D. Chappard, “Texture analysis of x-ray radiographs of iliac bone is correlated with bone micro-ct,” Osteoporosis international 17, 447–454 (2006). [CrossRef] [PubMed]
10. M. L. Bouxsein, “Bone quality: where do we go from here?” Osteoporosis international 14, 118–127 (2003). [CrossRef]
11. E. P. Paschalis, E. Shane, G. Lyritis, G. Skarantavos, R. Mendelsohn, and A. L. Boskey, “Bone fragility and collagen cross-links,” Journal of Bone and Mineral Research 19, 2000–2004 (2004). [CrossRef] [PubMed]
12. J. A. Inzana, J. R. Maher, M. Takahata, E. M. Schwarz, A. J. Berger, and H. A. Awad, “Bone fragility beyond strength and mineral density: Raman spectroscopy predicts femoral fracture toughness in a murine model of rheumatoid arthritis,” Journal of biomechanics 46, 723–730 (2013). [CrossRef]
13. A. L. Boskey, E. Donnelly, E. Boskey, L. Spevak, Y. Ma, W. Zhang, J. Lappe, and R. R. Recker, “Examining the relationships between bone tissue composition, compositional heterogeneity, and fragility fracture: A matched case-controlled FTIRI study,” Journal of Bone and Mineral Research 31, 1070–1081 (2016). [CrossRef]
14. S. Gamsjaeger, R. Mendelsohn, A. Boskey, S. Gourion-Arsiquaud, K. Klaushofer, and E. Paschalis, “Vibrational spectroscopic imaging for the evaluation of matrix and mineral chemistry,” Current osteoporosis reports 12, 454–464 (2014). [CrossRef] [PubMed]
15. P. Augat and S. Schorlemmer, “The role of cortical bone and its microstructure in bone strength,” Age and ageing 35, ii27–ii31 (2006). [CrossRef] [PubMed]
16. D. W. Dempster, J. E. Compston, M. K. Drezner, F. H. Glorieux, J. A. Kanis, H. Malluche, P. J. Meunier, S. M. Ott, R. R. Recker, and A. M. Parfitt, “Standardized nomenclature, symbols, and units for bone histomorphometry: a 2012 update of the report of the asbmr histomorphometry nomenclature committee,” Journal of Bone and Mineral Research 28, 2–17 (2013). [CrossRef]
17. C.-P. Adler, Bone diseases: macroscopic, histological, and radiological diagnosis of structural changes in the skeleton (Springer Science & Business Media, 2013).
18. J. D. Bancroft and M. Gamble, Theory and practice of histological techniques (Elsevier health sciences, 2008).
19. M. Veta, J. P. Pluim, P. J. Van Diest, and M. A. Viergever, “Breast cancer histopathology image analysis: A review,” IEEE Transactions on Biomedical Engineering 61, 1400–1411 (2014). [CrossRef] [PubMed]
20. A. R. Villanueva and K. D. Lundin, “A versatile new mineralized bone stain for simultaneous assessment of tetracycline and osteoid seams,” Stain technology 64, 129–138 (1989). [CrossRef] [PubMed]
21. C. Rentsch, W. Schneiders, S. Manthey, B. Rentsch, and S. Rammelt, “Comprehensive histological evaluation of bone implants,” Biomatter 4, e27993 (2014). [CrossRef] [PubMed]
22. S. M. van Gaalen, M. C. Kruyt, R. E. Geuze, J. D. de Bruijn, J. Alblas, and W. J. Dhert, “Use of fluorochrome labels in in vivo bone tissue engineering research,” Tissue engineering Part B: Reviews 16, 209–217 (2010). [CrossRef]
23. R. Yang, C. M. Davies, C. W. Archer, and R. G. Richards, “Immunohistochemistry of matrix markers in technovit 9100 new-embedded undecalcified bone sections,” Eur Cell Mater 6, 57–71 (2003). [CrossRef]
24. A. Boskey and N. P. Camacho, “FT-IR imaging of native and tissue-engineered bone and cartilage,” Biomaterials 28, 2465–2478 (2007). [CrossRef]
25. E. P. Paschalis, R. Mendelsohn, and A. L. Boskey, “Infrared assessment of bone quality: a review,” Clinical Orthopaedics and Related Research® 469, 2170–2178 (2011). [CrossRef]
26. M. D. Morris and G. S. Mandair, “Raman assessment of bone quality,” Clinical Orthopaedics and Related Research® 469, 2160–2169 (2011). [CrossRef]
27. A. Prentice, “Autofluorescence of bone tissues,” Journal of clinical pathology 20, 717–719 (1967). [CrossRef] [PubMed]
28. Z. K. Drzazga, A. Kluczewska-Gałka, A. Michnik, M. Kaszuba, and H. Trzeciak, “Fluorescence spectroscopy as tool for bone development monitoring in newborn rats,” Journal of fluorescence 21, 851–857 (2011). [CrossRef]
29. P. Kumar, S. Rao, S. D. Kamath, V. Prabhu, K. Satyamoorthy, and K. Mahato, “Autofluorescence of osteoporotic mouse femur bones: a pilot study,” Photomedicine and laser surgery 29, 227–232 (2011). [CrossRef]
30. F. Jamme, S. Kascakova, S. Villette, F. Allouche, S. Pallu, V. Rouam, and M. Refregiers, “Deep uv autofluorescence microscopy for cell biology and tissue histology,” Biology of the Cell 105, 277–288 (2013). [CrossRef] [PubMed]
31. G. P. Pilolli, A. Lucchese, E. Maiorano, and G. Favia, “New approach for static bone histomorphometry: confocal laser scanning microscopy of maxillo-facial normal bone,” Ultrastructural pathology 32, 189–192 (2008). [CrossRef] [PubMed]
32. C. Maggiano, T. Dupras, M. Schultz, and J. Biggerstaff, “Confocal laser scanning microscopy: a flexible tool for simultaneous polarization and three-dimensional fluorescence imaging of archaeological compact bone,” Journal of Archaeological Science 36, 2392–2401 (2009). [CrossRef]
33. G. Zhang, D. J. Moore, C. R. Flach, and R. Mendelsohn, “Vibrational microscopy and imaging of skin: from single cells to intact tissue,” Analytical and bioanalytical chemistry 387, 1591–1599 (2007). [CrossRef]
34. A. Papour, J. H. Kwak, Z. Taylor, B. Wu, O. Stafsudd, and W. Grundfest, “Wide-field raman imaging for bone detection in tissue,” Biomedical optics express 6, 3892–3897 (2015). [CrossRef] [PubMed]
35. L. Capasso, R. D’Anastasio, S. Guarnieri, J. Viciano, and M. Mariggiò, “Bone natural autofluorescence and confocal laser scanning microscopy: Preliminary results of a novel useful tool to distinguish between forensic and ancient human skeletal remains,” Forensic Science International 272, 87–96 (2017). [CrossRef] [PubMed]
36. O. Ristow and C. Pautke, “Auto-fluorescence of the bone and its use for delineation of bone necrosis,” International journal of oral and maxillofacial surgery 43, 1391–1393 (2014). [CrossRef] [PubMed]
37. H. E. Gruber, “Adaptations of goldner’s masson trichrome stain for the study of undecalcified plastic embedded bone,” Biotechnic & histochemistry 67, 30–34 (1992). [CrossRef]
38. F. Timischl, “The contrast-to-noise ratio for image quality evaluation in scanning electron microscopy,” Scanning 37, 54–62 (2015). [CrossRef]
39. R. Genthial, E. Beaurepaire, M.-C. Schanne-Klein, F. Peyrin, D. Farlay, C. Olivier, Y. Bala, G. Boivin, J.-C. Vial, D. Débarre, et al., “Label-free imaging of bone multiscale porosity and interfaces using third-harmonic generation microscopy,” Scientific Reports 7, 3419 (2017). [CrossRef]
40. E. Peli, “Contrast in complex images,” JOSA A 7, 2032–2040 (1990). [CrossRef]
41. J. Sodek, B. Ganss, and M. McKee, “Osteopontin,” Critical Reviews in Oral Biology & Medicine 11, 279–303 (2000). [CrossRef]
42. J. G. Skedros, J. L. Holmes, E. G. Vajda, and R. D. Bloebaum, “Cement lines of secondary osteons in human bone are not mineral-deficient: New data in a historical perspective,” The Anatomical Record Part A: Discoveries in Molecular, Cellular, and Evolutionary Biology 286, 781–803 (2005). [CrossRef]
43. R. T. Ingram, B. L. Clarke, L. W. Fisher, and L. A. Fitzpatrick, “Distribution of noncollagenous proteins in the matrix of adult human bone: evidence of anatomic and functional heterogeneity,” Journal of Bone and Mineral Research 8, 1019–1029 (1993). [CrossRef] [PubMed]
44. A. Croce and G. Bottiroli, “Autofluorescence spectroscopy and imaging: a tool for biomedical research and diagnosis,” European journal of histochemistry: EJH 58, 2461 (2014). [CrossRef]
45. M. Fenner, J. Park, N. Schulz, K. Amann, G. G. Grabenbauer, A. Fahrig, J. Karg, J. Wiltfang, F. W. Neukam, and E. Nkenke, “Validation of histologic changes induced by external irradiation in mandibular bone. an experimental animal model,” Journal of Cranio-Maxillofacial Surgery 38, 47–53 (2010). [CrossRef]
