John Czerski and Susanta K. Sarkar
Single Molecule Biophysics Laboratory, Department of Physics, Colorado School of Mines, Golden, USA
Medical practice can be categorized into three broad steps. First, the doctor performs a physical examination and studies the medical history of a patient. Second, the doctor orders diagnostic tests such as blood and/or urine tests, ultrasound, magnetic resonance imaging (MRI), and computed tomographic (CT) scans for conclusive diagnosis. Finally, the doctor prescribes treatment with drug therapy and/or surgery. Imaging materials and methods that connect these three steps with acceptable cost, biotoxicity, biodispersibility, biodistribution, and bioclearance are promising, and likely to obtain regulatory and insurance approvals. We often miss this broad bird's‐eye view of the process and work with toxic materials for biomedical imaging without the long‐term possibility of adoption by medical professionals. In the case of cancer, the top killer along with the heart diseases that causes financial and emotional ruin, multimodal uses of materials are particularly helpful. Many patients would benefit from a multimodal imaging material that detects the biomarkers for cancer in blood or urine with high specificity, enhances the contrast of common imaging modalities, helps surgeons detect cancer cells intraoperatively, and allows targeted drug delivery. In this context, carbon nanomaterials (CNMs) have generated significant interest as versatile imaging agents with low toxicity, in comparison to organic dyes and quantum dots (QDs), for biomedical applications.
The impressive range of optoelectronic properties, the availability of precursors for synthesis and functionalization, and the potential for biocompatibility make CNMs suitable for bioimaging. CNMs can be functionalized through well‐established organic chemistry techniques providing a versatile platform for multimodal imaging modalities. There are reviews on every aspect of nanoimaging including CNMs for bioimaging [1], clinically relevant in vivo diagnostic [2], clinical translation of nanotechnology [3], and fluorescent nanodiamonds (FNDs) for molecular and cellular bioimaging [4, 5]. In this chapter, we review some of the most promising CNMs for fluorescent imaging and their advantages, specific properties, and potential applications. In particular, we discuss FNDs for visible wavelengths and carbon nanotubes for near‐infrared (NIR) wavelengths in the context of biomedical imaging. The properties of FNDs and carbon nanotubes provide multiple pathways toward high‐contrast fluorescent imaging, contrast for other imaging modalities, and functional imaging.
While many CNMs are useful for bioimaging, we limit our discussion to FNDs, single‐walled carbon nanotubes (SWCNTs), graphene, and carbon nanodots (CNDs). FNDs are indefinitely photostable and can be used to make highly sensitive measurements of magnetic field, and temperature. SWCNTs have unique structural characteristics along with the intrinsic fluorescence and polarization anisotropy. Graphene has unique structural characteristics that offer a range of potential uses such as membranes and trackable drug delivery agent. CNDs are easily produced through a variety of reaction pathways and have tunable fluorescent properties. The number and variety of uses for these materials will only increase as these properties are better understood and more researchers use them for their specific imaging application.
Regardless of the CNM being used, they typically require functionalization prior to preclinical and clinical imaging. One general approach is to coat the nanomaterial with silica similar to sugar‐coated medicines. Figure 4.1 shows an example of silica‐coating (Figure 4.1a) and subsequent functionalization of FNDs (Figure 4.1b) [6]. This approach provides a thin enough coating to allow functionalization without increasing the size of the FNDs by any detectable amount, even by a transmission electron microscope (TEM). In addition, it purifies the FNDs and selectively separates them based on their size and the size of the lipid vesicles.
FNDs have gained significant attention due to the unique optical properties of the nitrogen vacancy (NV) defect centers, where one carbon is replaced by a nitrogen atom next to a vacant site (Figure 4.2a). FNDs are highly photostable, exhibit low biotoxicity, can be functionalized, and are sensitive to electromagnetic fields and temperature. NV defects can be negatively charged or neutral, resulting in zero phonon lines (ZPLs) at 637 or 575 nm, respectively. NV– centers are of particular interest because they provide a controllable isolated spin system that can be used to measure electric fields, magnetic fields, and temperature with unprecedented sensitivity and resolution [7–11]. FNDs are used in a variety of imaging applications from cellular biomarkers and background free imaging to microscope drift estimation [12–16]. Aside from their obvious utility as fluorescent markers, the unique spin properties of the NV¯ centers can be exploited to measure local temperature, magnetic field, and electric field in a system using optically detected magnetic resonance (ODMR) [17].
Carbon nanotubes are cylindrical rolls of one or more layers of graphene. They have been studied in depth because of their unique mechanical and electrical properties. SWCNTs are effectively one‐dimensional and can be fluorescent or conductive, depending on their structure. SWCNTs' fluorescence peaks depend on their chirality and diameter. They also exhibit significant polarization anisotropy, and their fluorescence can be tuned via ultraviolet illumination [18–20]. While initial measurements of the quantum efficiency of SWCNTs indicated that they may not be useful as fluorescent probes because of the low quantum yield ~3% [21], their emission in the NIR and strong Raman peaks with narrow linewidths have proven valuable for bioimaging. Carbon nanotubes have also been investigated as photoacoustic contrast agents as they absorb in the NIR and produce a strong photoacoustic signal [22]. They can also encapsulate other materials, which could be useful for multimodal imaging such as X‐ray fluorescence (XRF) microscopy and MRI [23, 24].
Graphene is a 2D material consisting of a hexagonal lattice of carbon atoms composed of sp2 hybridized bonds [25]. Graphene and its various derivatives such as graphene oxide (GO), graphene quantum dots (GQDs) have many properties that make them ideal candidates for biomedical applications [26]. Pure graphene does not exhibit fluorescent properties, but has many potential uses due to its high surface area to volume ratio and unique mechanical properties. GO and reduced graphene oxide (RGO) exhibit useful fluorescent properties such as excitation dependent emission spectra and quenchable fluorescence [27–29] The unique mechanical and chemical properties of these materials also make them of great interest as observable delivery media for drugs.
The term CND refers to a range of quasi‐spherical carbon nanoparticles typically smaller than 10 nm that exhibit a range of composition‐dependent fluorescent properties [30]. CNDs can be separated into two main groups based on their composition, crystalline carbon quantum dots (CQDs) and amorphous varieties. GQDs can be classified as CQDs, but exhibit photoluminescence (PL) properties based on size, edge effects, and electron‐hole recombination [31]. CNDs, particularly the amorphous variety, have gained significant attention due to their ease of production and tunable optical properties [32–36]. As with many other CNMs, CNDs are useful for a range of imaging applications due to their ease of functionalization and fluorescent properties. There has also been some research into their utility in other imaging modalities such as MRI and photoacoustic imaging (PAI) [37, 38].
Useful nanomaterials in the context of fluorescence imaging must be fluorescent or easily labeled with fluorescent particles. In the first case, this requires energy level gaps in the range of 0.9–3 eV and primarily radiative transition pathways. In the latter case, the material must be easily modified and conjugated with molecules such as fluorescein, cyanine, indocyanine green (ICG), or any other highly fluorescent particle using methods such as the silica coating shown in Figure 4.1. In the case of materials with intrinsic fluorescence, it is important to characterize their optical properties such as PL emission spectra, absorption or photoluminescence excitation (PLE) spectra, quantum yield, and photostability. Each of the CNMs included in this chapter outperform existing commercial organic fluorophores in at least one of these categories.
FNDs have indefinite photostability, broad excitation, and emission spectra in the visible and NIR range [39, 40], and magnetic field‐dependent fluorescence emission [41, 42]. Figure 4.2a shows an NV center in the diamond lattice. Figure 4.2b shows the energy level diagram of a negatively charged NV¯ center, which has a triplet ground state ( m S = 0 , m S = ± 1), a triplet excited state ( m S = 0, m S = ± 1), and a pair of metastable singlet states [42–45]. The spin quantum number m S is quantized along the N–V symmetry axis, and the degeneracy of m S = 0 and m S = ± 1 levels are lifted due to spin‐spin and spin‐orbit coupling where the lattice strain plays an important role. Spin conservation rules dictates that the optical transitions can only happen from m S = 0 (ground state) to m S = 0 (excited state) or from m S = ± 1 (ground state) to m S = ± 1 (excited state). However, an NV¯ center in the m S = ± 1 excited state is more likely to go to the metastable state and from there it goes to m S = 0 ground state instead of m S = ± 1 ground state. This pathway is nonradiative and has two effects: (i) it pumps the NV¯ centers to m S = 0 state within ~μs; and (ii) it increases the fluorescence because transitions involving m S = 0 states are radiative. In the presence of a magnetic field, m S = 0 and m S = ± 1 can be mixed. As a result, the nonradiative transition involving m S = ± 1 states improves and therefore, fluorescence drops in the presence of a magnetic field.
FNDs are extremely photostable in sizes as small as 5 nm [46]. One method for characterizing the photostability of a fluorophore is to fix it on a slide at pM concentrations and image a field of view such as the one shown in Figure 4.2c. This image was taken using a prism‐type total internal reflection fluorescence microscope (TIRFM) with a 532 nm excitation source. As shown in Figure 4.2d, the intensity of one of the FNDs (blue line) in Figure 4.2c did not show any reduction in fluorescence over the course of five minutes. In contrast, an AlexaFluor555 dye (red) showed distinctive single‐step photobleaching after only 200 seconds with five times lower excitation intensity. Continuing to expose the FNDs to high‐intensity excitation light for an hour did not change their fluorescence intensity. Along with this impressive photostability, FNDs are very bright with quantum yields ranging from 10 to 90% [47]. Another method for characterizing the brightness of a fluorophore is to measure its signal‐to‐noise ratio (SNR). The average SNR for the FNDs in Figure 4.2c was ~1210, while AlexaFluor555 dyes had an SNR of ~120 under identical imaging conditions. We define the SNR here as the mean fluorescence intensity divided by the standard deviation. An example of a typical intensity distribution is shown in Figure 4.2e.
The NV0 and NV¯ centers have ZPLs at 575 and 637 nm, respectively [48, 49], and PL spectra extending into the 750 nm range. Figure 4.2f shows the PLE (blue) and emission spectra (red) of an ensemble of 100 nm FNDs. It should be noted that PLE spectrum mimics absorption spectrum in the absence of nonradiative relaxation, and therefore, can be used to determine absorption of a nanomaterials. Despite their importance in imaging, PLE and emission spectra of FNDs and nanomaterials are difficult to measure due to scattering. The PLE spectra shown in Figure 4.2f were collected by imaging the FNDs using a TIRFM with a supercontinuum laser and a volume Bragg grating‐based monochromator. The analysis of PLE spectra suggested that two types of FNDs were present, illustrating the importance of proper characterization. Since NV centers exist in charged and neutral forms, the data pointed to the fact that on the single molecule level, different ratios of NV¯/NV0 will result in differing PLE spectra that can be used to distinguish FNDs with a majority of NV¯ or NV0 centers. The ensemble PL emission measurement shown in Figure 4.2f was taken using a home‐built fluorimeter. The sample was excited by focusing a 1.5 W laser at 532 nm through a cuvette very close to its edge. If the sample is excited in the middle of the cuvette, the scattering significantly reduces the PL signal. The fluorescence was focused with a 10X objective and filtered with a 550 nm long pass filter before detection using an Ocean Optics USB4000 spectrometer.
NV centers have a number of other properties that can be used to measure biological systems. In particular, NV¯ centers are sensitive to magnetic fields, electric fields, and local temperatures [7, 9, 10, 50–54]. Maze et al. were able to detect a 3 nT magnetic field at kHz frequencies following 100 seconds of averaging by measuring the increased ground state splitting in the presence of the field [9]. This corresponds to a sensitivity of 0.5 μT Hz−1/2 with 30 nm diameter FNDs at room temperature. Similar methods have been used to detect the electric field of a single electron 25 nm away [50]. This method could theoretically be used to detect fields with a sensitivity of 202 V cm−1 Hz−1 [51]. At this sensitivity, it is possible to map electric fields in neurons and other biological systems with unprecedented resolution. Along with their ability to measure electric and magnetic fields, the same ground state splitting measurements can be used to measure local temperatures [53]. Like electric and magnetic fields, higher temperatures split the ground state energy levels of NV¯ centers. This energy level splitting can be detected through ODMR or with an all optical technique by exploiting the Debye‐Waller factor [10]. Temperature detection via ODMR resulted in sensitivities of 10 mK Hz−1/2 , while measurements collected using the Debye‐Waller factor have a noise floor of 100 mK Hz−1/2 . In either case, these sensitivities surpass those of other biocompatible nanothermometry techniques and could be useful in a number of experiments including the photothermal treatments being tested with SWCNTs [55].
While FNDs provide emission light in the first NIR window, SWCNTs fluoresce in the second NIR window [56–61]. This spectral range is ideal for deep‐tissue imaging and situations where auto‐fluorescence from biomolecules creates excessive background noise. While a number of CNMs have multiple emission peaks, the chirality/diameter dependent PL spectra of individual SWCNTs provides numerous separated peaks that can be used to identify these important properties [20, 62]. SWCNTs are also extremely photostable showing no appreciable decrease in fluorescence after long periods of laser excitation. Unlike FNDs and CNDs, the quantum yield of SWCNTs is relatively low, around 3% [21]. Functionalization of the nanotube increases the quantum yield by up to 15 times, and remains a topic of interest [58, 63]. Improvements in the quantum yield along with their emission in the NIR and potential to be used as a highly multimodal imaging agent give SWCNTs significant potential as a fluorescent probe for biomedical imaging. Aside from their aforementioned fluorescence properties, SWCNTs have a number of interesting properties that could be exploited to enhance their utility. SWCNTs demonstrate significant polarization anisotropy [64], and their PL emission can be further red shifted by exposing them to ultraviolet (UV) light [18] or doping them at proximal sites [63]. Most importantly, they have been demonstrated to be useful deep‐tissue probes in a number of studies and are capable of acting as labels for extended periods [57, 58, 65, 66].
Long‐term success of fluorescent probes for preclinical or clinical studies depend on the following general considerations:
Regulatory approval and eventual adoption by medical professionals of many promising fluorescent probes have been hindered by lacking in one or more of these areas. However, in vitro diagnostics are less restrictive and have benefited from fluorescent probes with desirable fluorescent properties but are not likely to be approved for human use.
An ideal fluorescent probe should be indefinitely photostable, bright, and inexpensive. In general, six properties of a fluorescent probe can be used to achieve high‐contrast biological imaging:
In vivo fluorescence imaging mainly focuses on two wavelength windows, ~650–950 and ~950–1400 nm, because light in these wavelength ranges penetrates deeper due to relatively low tissue absorption and scattering. In vivo imaging typically uses the shorter wavelength window, due to the availability of dyes in the entire visible range and the overlapping applicability for in vitro fluorescence imaging. However, significant progress has been made in the longer wavelength window as well.
Before reviewing fluorescence imaging with CNMs, we briefly discuss three widely used preclinical and clinical imaging modalities, PAI, MRI, and X‐ray CT imaging. These are followed by a section on image registration and alignment, which is extremely important for medical imaging. Exploring carbon‐based fluorescent probes that could be useful for one or more of these is attractive because it enables connection of preoperative PAI/MRI/CT imaging with intraoperative guided fluorescent imaging with high specificity.
PAI combines the benefits of both optics and ultrasound by exciting the sample with laser and then detecting the ultrasound generated due to laser absorption and subsequent thermoelastic expansion [67]. PAI maintains high optical contrast while achieving high spatial resolution by detecting ultrasound, which scatters much less than light and is widely used in clinics. PAI can achieve sub 0.2 mm resolution at 19 mm depth with a penetration depth of up to 38 mm [68]. In combination with a contrast agent, which could be various CNMs, PAI is becoming very useful for bioimaging and possible photothermal therapies. SWCNTs have been used as PAI contrast agent to image expression of integrin in tumor [69]. Sentinel lymph node (SLN) imaging, where a contrast agent is injected near the primary tumor and taken up by adjacent lymphatic systems before moving to SLN, is another application where SWCNTs have been used to enhance PAI [70]. Graphene‐based PAI agents have shown ~5–10‐fold PAI signal enhancement in comparison to blood at 755 nm [71] and seems to have better dispersibility in biological systems [72]. Radiation‐damaged nanodiamonds with high near‐infrared absorption enhanced the PAI signal by 567% at 820 nm imaging 3 mm below the skin surface in rodents [73].
Both 2D X‐ray and 3D CT detect density‐dependent absorption and scattering of X‐ray in tissues. Typical spatial resolution of CT scans is ∼1 mm3 , but can reach down to 1 μm for ex vivo imaging of bones using micro‐CT. Potential carcinogenic effect of X‐ray is considered acceptable compared to the diagnostic benefit, but dosing and technical requirement limit the spatial resolution to 130 μm for in vivo bone imaging. SWCNTs could be filled with various materials to increase their usefulness for microscopic imaging. In the case of XRF, this includes nonbiological and even toxic materials that act as contrast agents but cannot otherwise be used for cellular imaging. Serpell et al. [23] filled and sealed SWCNTs with krypton, barium, and lead. These three elements provide signal peaks, 4.5, 10.5, and 12.6 keV, outside emission regions 2.0–3.7 and 5.9–8.6 keV, which contain numerous peaks from biological samples. Once the SWCNTs are sealed, they are then conjugated with specific biological agents using covalent bonding to the SWCNT surface, allowing them to act as contrast agents for various regions of a cell. This provides a means of differentiating various regions of the cell and can be used in conjunction with scanning XRF microscopy, high‐angle annular dark‐field scanning transmission electron microscopy, as well as Raman microscopy. In theory, this method for encapsulation could be extended to a variety of other materials and imaging techniques such as MRI and traditional fluorescence microscopy. Highly iodinated fullerene has been shown to be a promising contrast agent for X‐rays [74].
MRI typically uses the magnetic spin of hydrogen nucleus since it is present in water and fat – two major components of the human body. In MRI scanners, a magnet with field strength in the range 0.5–1.5 T is applied to align all the spins of hydrogen nuclei in a body part along the direction of the magnetic field. The resulting magnetic vector is then rotated by turning on a small radiofrequency magnetic field. When the radiofrequency source is turned off, the magnetic vector returns to the original orientation along the strong magnetic field with two characteristic timescales T 1 and T 2 that depend on the source of the hydrogen, the microenvironment, and the presence of a contrast agent such as gadolinium III (Gd III). Compared to X‐rays, MRI has better soft tissue contrast and spatial resolution in the range of 1 mm3 . FNDs are known to enhance the effect of clinically used MRI contrast agent, Gd (III) [75, 76].
In medical imaging, it is most often necessary to register and align images from different image modalities, images taken at different time points, and images taken with accompanying tissue movement. In addition, imaging system itself can drift during image acquisition. There has been significant progress in image registration and alignment research, but it is still an active area of research [77]. Two general approaches to image registration and alignment are feature‐based and fiducial marker‐based tracking. Colomb et al. [14] showed that FNDs can be used as efficient fiducial markers for correcting drift from Gaussian or non‐Gaussian noise. Using a generalized maximum likelihood method (MLE), the authors estimated microscope drift with less than 6 nm precision and accuracy. FNDs are suitable fiducial markers for multicolor and multimodal imaging due to their ~200 nm wide emission spectrum and beneficial properties for PAI/MRI/CT imaging. FNDs have been used to align images acquired using fluorescence microscopy and electron microscopy [15, 78–80]. Yi et al. [81] used FNDs as fiducial markers for multiplexed direct stochastic optical reconstruction microscopy (madSTORM), a super‐resolution technique for large‐scale multiplexing at the single molecule level. Techniques such as stochastic optical reconstruction microscopy (STORM) implement switchable fluorophores which allow large cell structures to be imaged with super‐resolution by imaging the sample repeatedly with these switchable dyes and localizing individual fluorophores at relatively high densities. Stage motion reduces the density at which these fluorophores can be localized as the fluorophores appear to overlap. To mitigate this problem, the authors used FNDs as fiducial markers and used a method known as average fiducial correction (AFC). This method allows the stage motion to be corrected and also provides a means for estimating the localization precision (2.6 nm) and distinguishing individual antibodies that had an approximate size of 12 nm2 .
While there are a variety of fluorescent CNMs, the most promising for biomedical imaging are FNDs and SWCNTs. Both are inexpensive, commercially available, not prohibitively biotoxic and biodispersible, and have bioclearance with suitable retention time for intraoperative imaging. FNDs have been used in many types of optical imaging primarily in the short visible wavelength window including one‐photon imaging, two‐photon imaging, background‐free imaging, lifetime imaging, and multicolor imaging. SWCNTs are used in the long‐wavelength window imaging including one‐photon imaging, multiphoton imaging, multicolor Raman imaging, and transient absorption imaging. Comparatively, imaging with FNDs is less expensive and combining with existing diagnostic imaging has more potential than SWCNTs.
Carbon dots and FNDs are most used CNMs for imaging in the short‐wavelength window. The first imaging of carbon dots in live mice was reported by Yang et al. [82]. They observed bright emission within the injection region at 525 and 620 nm with 470 and 545 nm excitations, respectively. One of the first application of FNDs involved imaging of 100 nm FNDs in rats that showed long‐term stability and biocompatibility over five months [83]. Recently, FNDs have been used to track human mesenchymal stem cells in miniature pigs [84]. FND‐based therapeutic delivery agents have proved to enhance treatment of chemoresistant tumor [85]. Steinert et al. reported room temperature imaging magnetic spins of paramagnetic oxygen, MnCl2, and Gd ions using FNDs [86]. Hall et al. showed wide filed imaging of neuronal activity using FNDs. Even though these studies were in vitro, they showed that FNDs can be used for functional imaging as well.
Both aberration and diffraction blur the image and, thus, determine the spatial resolution of the imaging system. Even if aberration is reduced by appropriate optical design and material quality, diffraction remains due to the fundamental wave nature of light and leads to the diffraction limit of resolution, d = λ/2NA , where λ is the detected wavelength and NA is the numerical aperture of the imaging system. Advances in fluorescence microscopy over the past decades have enabled resolution below the diffraction limit and several super‐resolution techniques have emerged with resolutions less than 50 nm [87–95]. Techniques such as PALM and STORM rely on stochastic switching of few fluorophores with low intensity excitation that does not lead to overlapping point spread functions (PSFs) due to emission from fluorophores. As such, individual PSFs can be fit to a distribution function and fluorophores can be localized with precision well below the diffraction limit. Therefore, the concentration of fluorophores for labeling biological samples for PALM and STORM needs to be optimally chosen. In contrast, techniques such as stimulated emission depletion (STED) rely on nonlinear depletion of excited fluorophores in the periphery using a doughnut‐shaped red‐shifted STED laser so that only the emission from fluorophores at the center is detected, which effectively narrows the PSF beyond the diffraction limit. These super‐resolution techniques have also been applied to nanodiamond‐based imaging using NA engineering such as structured illumination microscopy (SIM) [96], near‐field microscopy [97], and far‐field microscopy techniques such as STED microscopy [98] and madSTORM [81]. Recently, STED has been used to image individual NV¯ centers in FNDs with a measured PSF of 5.8 nm [99]. Each super‐resolution technique has advantages and disadvantages, and therefore, should be judiciously chosen for a particular biological application [100].
Fluorescence from FNDs can be modulated by external microwave or magnetic field to distinguish them from background fluorescence. Igarashi et al. [101] used the electron spin resonance to regulate the ground state spin configuration of NV centers in FNDs and imaged them inside Caenorhabditis elegans and mice. Sarkar et al. developed [13] a method for background‐free images of FNDs in mouse lymph nodes by modulating FND emission with an electromagnet. Figure 4.3a shows the clear intensity modulation of an FND with ~100 G magnetic field. Such selective modulation can be used for background‐free imaging. As an in vitro test, ~40 nm FNDs containing ~15 NV¯ defect centers were imaged with a TIRFM in the absence (Figure 4.3b) and the presence (Figure 4.3c) of ~1 μM Alexa647 dye solution. The mean (ImageOFF − ImageON) for 1000 pairs of images taken with and without the magnetic field of view is shown in Figure 4.3c. As shown in Figure 4.3d, the background noise was removed effectively. Next, the technique was applied in vivo to image FNDs in mice. Images were processed two different ways: (i) simple subtraction of images with and without the magnetic field and (ii) phase‐sensitive lock‐in detection of signal amplitudes. Figure 4.3e shows the image of a mouse with the simple subtraction method with the background (top left), the mean (ImageOFF − ImageON) of a mouse for 475 pairs of images (bottom left), and the combined overlaid image (right). FNDs were injected in the footpad of a mouse. To track the uptake by lymph nodes, the chest cavity of the mouse was surgically exposed and imaged. Figure 4.3f shows the combined image of the open chest cavity with both types of image processing: the simple subtraction of images with and without the magnetic field using (ImageOFF − ImageON) (left) and the phase sensitive pixel‐by‐pixel lock‐in processing (right). As a control experiment, the pixel values as a function time were tracked corresponding to the selected points in (e) and (f) to confirm the signal modulation as shown in Figure 4.3g. Pixel‐by‐pixel lock‐in processing reduced the background noise by nearly 100‐fold. This method is also versatile as it could be used in TIRF microscopes, confocal microscopes, or any other fluorescence microscope.
Another method being exploited for background reduction in FND fluorescence imaging is time‐gated imaging. FNDs have a long fluorescence lifetime, ~17 ns [102], compared to the fluorescence lifetime of ~5 ns [103] from autofluorescence in tissue and biological materials. An order of magnitude difference in the fluorescence lifetimes enables the use of fluorescence lifetime imaging (FLIM) and time‐gated imaging of FNDs to reduce the autofluorescence background. Wu et al. [104] used time‐gated imaging of FND‐labeled stem cells to track engraftment and regenerative capabilities of transplanted lung stem cells. Hui et al. [12] showed time‐gated imaging using a pulsed source and nanosecond intensified charge‐coupled device (ICCD). The authors achieved 599 nm excitation by frequency doubling a 1064 nm source and subsequently red‐shifted the resulting 532 nm light via a Ba (NO3)2 crystal. In the 599 nm region there is significantly less tissue absorption, allowing the excitation source to penetrate much deeper into the tissue. There is also significantly less fluorescence from the hemoglobin that has a broadband emission profile in the same range as FNDs, 550–750 nm. This technique enabled the authors to achieve a lateral resolution less than 0.5 μm using their wide‐field time‐gated fluorescence imaging, as well as a resolution of ~5 μm for samples covered by 0.1‐mm‐thick chicken breast.
For biomedical imaging, the longer wavelength window is desirable because of relatively low autofluorescence of tissues [105], significantly lower scattering, and lower absorption by the blood and tissues [106]. A mouse can be effectively made translucent by imaging at longer wavelengths, and it is possible to image in vivo vasculature [107] or through the skull without thinning [108]. One of the main obstacles in this imaging window is the scarcity of suitable long‐wavelength fluorescent probes. QDs such as InAs [109], PbSe, PbS, and CdHgTe fluoresce in the long‐wavelength window [110], but their biotoxicity prevents use for in vivo applications. While pulmonary toxicity of SWCNTs in mice can occur [111], NTs have been reported to show no acute toxicity [112].
In one‐photon imaging, SWCNTs have been used for deep‐tissue anatomical imaging in mice [57] and for imaging of tumor vessels under thick skin in mice [58]. CNMs have large two‐photon cross‐sections [113], making them suitable for two‐photon imaging of biological samples [114]. Two‐photon imaging was developed by Denk et al. [115], which excites samples only at the focal point resulting in lower scattering and deeper penetration compared to one‐photon imaging.
Raman imaging, based on inelastic scattering of light due to electron‐phonon coupling, has been performed in vivo using SWCNTs conjugated with RGD peptide in live mice [116, 117]. Raman imaging has the advantage that it can distinguish molecules spectrally due to narrow Raman linewidths and does not require fluorescence. This allows multicolor imaging with convenient multiplexing. For example, Liu et al. [118] synthesized SWCNTs with five different C13/C12 compositions with distinct Raman peaks, conjugated them to five different target ligands for specificity, and imaged cancer cells and tumor tissues. The primary drawback to this and other hyperspectral methods is the slow acquisition time that hinders dynamic imaging. The longer wavelengths of light also limit the effective resolution since the PSF of the imaging system is proportional to the wavelength of light. In this case the resulting images had ~1 μm xyz resolution with a SNR ratio of ~100. Roxbury et al. [62] reported spatially and spectrally resolving 17 distinct SWCNT species in live mammalian cells, murine tissues ex vivo, and zebrafish endothelium in vivo. If the SWCNTs could be conjugated according to their chirality, this would provide a platform for hyperspectral imaging with a large number of selectively conjugated dyes. SWCNTs have intense Raman peak due to the strong electron‐phonon coupling [119].
Transient absorption microscopy, where two lasers beams are used to measure the differential absorption, is a very sensitive phase‐based imaging technique. For SWCNTs, a pump beam at 707 nm and a probe beam at 885 nm can be used to access the E 11 transition of metallic SWCNTs (600–800 nm) and E 22 transition of semiconducting SWCNTs (850–1100 nm) [120]. The circulation of SWCNTs injected via tail vein in mice has been observed with high temporal resolution [121]. Transient absorption microscopy has negligible interference from autofluorescence [122] and can be used to image nonfluorescent CNMs as well.
In this chapter, we have highlighted the fundamental concepts and some in vivo examples of biomedical imaging with CNMs. In particular, we focused on those techniques using FNDs (shorter wavelengths) and SWCNTs (longer wavelengths). FNDs allow background‐free imaging because their emission can be selectively modulated using a magnetic field and their fluorescence lifetime is longer compared to tissue autofluorescence. SWCNTs allow background‐free imaging because they emit at wavelengths where tissue absorption and scattering are negligible and they have intense Raman peaks with narrow linewidths. Both FNDs and SWCNTs have large two‐photon cross‐sections and therefore, are suitable for multi‐photon imaging. Defects in diamonds have been studied much longer than SWCNTs and therefore, the properties of FNDs have been studied in greater depth. FNDs are probably the most suitable CNM for biomedical imaging if purity, inertness to biological environment, cost, biotoxicity, biodispersion, retention, and clearance are considered together with their diverse applications outside biomedical imaging. In addition, FNDs can be used to enhance MRI contrast, as a photoacoustic contrast agent, and allows functional imaging. FNDs have a good chance of being successfully used in clinics for in vitro diagnostics, for enhancement of contrast agent for preoperative noninvasive in vivo diagnostic, for intraoperative diagnostic of diseases enabling better surgical outcomes, and for targeted drug delivery.