Introduction

Bio-engineering, bio-technology and associated technologies have seen a continuous growth over the past decades by the development of concepts and tools that advance our understanding of life and disease. From system biology approaches to advanced diagnostic and therapeutic strategies, there is an increasing sophistication in approaches geared toward finding effective treatments and curing major diseases. It comes then often as a surprise when one realizes that for a large part of medical visualization, for example in surgery, endoscopy and overall in interventional procedures, the most widely used detector is the human eye; a process that has remained unchanged for thousands of years.

The human eye is an excellent detector of anatomical features and it can also detect certain major discolorations. Overall however, human vision is not appropriate for visualizing many molecular diseases. Limitations are typically related with inabilities in visualizing molecules and genes, small clusters of cells and cellular function or seeing under the surface. Radiological approaches have been considered for improving on the visual examination during interventions. Magnetic resonance imaging (MRI),11,40,45,137,142,201 positron emission tomography (PET),5,13,32,80,115,133 ultrasound (US)61,151,155,163 and X-ray computed tomography (CT)20,26,65,149 offer several promising characteristics but also come with limitations. Most whole-body imaging methods are of substantial size and cost that limits them from gaining wide dissemination in the operating room and they do not typically offer imaging resolution that matches that of the human eye (i.e., ~50 μm, or better when using surgical microscopes). US imaging and portable gamma cameras are potent methods for imaging during interventional procedures as they offer practical implementations; in particular they are portable and can be built in small form factors and in a cost-efficient manner. Despite their advantages over larger radiological systems, clinical decision making during surgery or endoscopic intervention still relies primarily on the surgeon’s ability to distinguish cancerous from healthy tissue under visual examination, palpation and corresponding histopathological confirmation.

Advanced optical imaging approaches have been considered as alternatives that can enhance interventional imaging as they directly relate to surgical vision and can offer resolutions that match or exceed that of the human eye. Of particular importance is the ability to generate contrast from disease that goes beyond sensing discolorations or the anatomical features that can be perceived by the human eye so that additional diagnostic or theranostic information is provided during the imaging session. Improved contrast can be achieved by using exogenous contrast agents tailored to highlight certain biological events or functions, or by detecting disease-related endogenous optical contrast that goes beyond the discrimination sensitivity of human vision. The ability to integrate optical methods in existing endoscopes could further allow a seamless operation in the operating suite.

This review discusses optical imaging approaches developed to enhance interventional imaging. While particular concentration is given to cancer imaging approaches, the methodologies developed apply more generally to detecting a larger range of diseases spanning from cardiovascular disease to infection or nerve imaging. This is because the improvement of optical identification of cancer during surgery closely relates to the ability for early diagnosis during screening endoscopy or for identifying inflammation or infection during interventional procedures in cardiovascular applications, anastomosis procedures and so on. The optical imaging approaches considered for interventional imaging are divided herein based on their operational characteristics into two main categories, i.e., optical imaging implementations that offer large field of view inspection (several cm × cm) and methods that operate with smaller field of view but offer higher resolution and three-dimensional tissue visualization through several hundreds of microns to centimeters.

Large Field of View Optical Methods

Narrow Band Imaging

Narrow band imaging (NBI) is a method primarily used in endoscopic systems of the digestive tract to increase/enhance local changes in tissue absorption using spectral differentiation in epi-illumination (reflectance) geometry. Typically systems capture images at spectral bands around 415, 540 nm and possibly 600 nm and offer processed images that are a fusion of the spectral images captured. The principle of operation relies on the differences in light penetration for blue light (415 nm) vs. green light (540 nm). In response, narrow band blue light images are sensitive to superficial blood vessels, while green light images contain contrast from deeper (sub-epithelial) blood vessels. NBI can therefore highlight superficial microvasculature differences and the underlying mucosal tissue of superficial suspicious lesions up to 200 μm depth. The irregular shape, size and pattern of local microvasculature (vessels and capillaries) and disrupted tissue structure associated with pre-cancer or early cancer development can thus be visualized with increased sensitivity and contrast from the surrounding tissues, compared to white light endoscopy (WLE). Studies report that NBI has been found superior for the detection of colorectal,46,86,87,208 gastric,23,98 lung221 cancers or precancers and Barrett’s esophagus214 with enhanced sensitivity and specificity compared to WLE. Figure 1 illustrates the contrast increase when using NBI compared to WLE in an intestinal metaplasia case. Kato et al.98 demonstrated the performance of NBI in 201 lesions from 111 patients with a high risk of gastric cancer. Using a magnifying endoscopy NBI system, they achieved a sensitivity and specificity of 92.9 and 94.7%, respectively in identifying gastric cancer, compared with 42.9 and 61.0% of WLE.

Figure 1
figure 1

White light endoscopy (left) observation of intestinal metaplasia at the angulus, and contrast improvement from NBI (right) of the same region

Despite the simplicity of implementation and use, NBI is still under evaluation as many studies suggest that although NBI offers an alternative, ‘contrast-enhanced’ visualization of the tissues, this technology has limited specificity,23,96 moderate inter-observer agreement,36,86 requires considerable learning experience46 or does not present a comparable advantage compared to the latest high resolution white light endoscopy (HR-WLE)35,157 or even conventional colonoscopy2 systems. To overcome these limitations some approaches adopted a multimodal solution by combining NBI with other techniques (auto-fluorescence, chromoendoscopy or HR-WLE)37 to improve the overall system performance.

Auto-Fluorescence Imaging (Endogenous Contrast)

Auto-fluorescence, i.e., the fluorescence emitted by endogenous tissue molecules, may vary between different tissues, for example during tissue progression from healthy to malignant. Collagen, nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD) are predominant intrinsic tissue molecules that emit fluorescence upon excitation with UV or blue light.42,63 Alterations in the biochemical composition of tissue during the early stages of carcinogenesis can be observed as a change in the fluorescence signals captured indicating cancer-associated tissue changes. Various wavelengths for both the excitation (300–480 nm) and emission (500–700 nm) of intrinsic fluorophores have been utilized (especially for auto-fluorescence spectroscopy), but typically blue light excitation (~450 nm) is preferred to UV light, due to the low UV transmittance of conventional optics. The fluorescence emission captured then at lower energies, typically in the green but also in the red spectral regions. Tumor appearance is manifested as a reduction of the green fluorescence (due to collagen cross-links breakdown) and an increase in the red fluorescence (due to an increase of local porphyrins).100,116,121 Auto-fluorescence imaging is technically easy to implement during patient screening or surgery and therefore it does not require expensive equipment or complex regulatory approvals. Endoscopic and wide-field auto-fluorescence imaging has been considered for screening oral,38,83,90,117,161,175 laryngeal,8,112,220 cervical,54,84,193 gastric55,72 and colorectal88,186 cancer. Lane et al.117 reported a 98 and 100% increase of sensitivity and specificity using autofluorescence imaging in 44 patients screened for oral dysplasia or cancer, while Kraft et al.112 observed a superior performance of autofluorescence (sensitivity 91%, specificity 84%) endoscopy compared to white light (sensitivity 73%, specificity 79%) and even aminolevulinic acid (ALA) induced fluorescence (sensitivity 95%, specificity 62%) endoscopy. Auto-fluorescence images of a papillary tumor showcasing the contrast improvement achieved are shown in Fig. 2.

Figure 2
figure 2

Respiratory papilloma images obtained from WLE (a), demonstrating a papillary tumor of the left vocal fold; Autofluorescence Endoscopy (b), showing a marked loss of autofluorescence; Induced Fluorescence Endoscopy (c), displaying a strong protoporphyrin IX fluorescence

Currently, however, opinions are divided as to the specificity of the technique. Huh et al.84 reported a specificity of 50% in a 604-patient study for cervical neoplasia, while Jayaprakash et al.90 presented similar specificity in 60 high risk patients with suspicious oral lesions. It is hypothesized that biological mechanisms other than tumor-related biochemical changes can also cause a reduction in green fluorescence, presenting, thus, false positives. Inflammation, pigmentation, hyperkeratosis and ulceration are all conditions under which a decrease in green fluorescence levels can be observed, rendering thus the identification of malignant lesions challenging. In addition, as discussed in the following, fluorescence imaging depends on many instrumentation and tissue parameters and data currently are not performed using standardized performance; a topic that is not accurately considered yet in clinical studies but can lead to data variability.

Hybrid approaches have been also considered for bypassing these limitations, for example the combination of autofluorescence imaging and NBI have been proposed97,161 for utilizing narrow band images to reduce the false positive rate of the autofluorescence modality (from 40% to only 10%).

Epi-Illumination Fluorescence Imaging (Exogenous Contrast)

In addition to utilizing endogenous tissue contrast, the field of targeted fluorescent agents for exogenous administration has exploded in the last decade. The administration of fluorescent contrast agents, i.e., non-specific dyes that distribute primarily in the vascular system, has been considered since the 1940s using fluorescein.114 Fluorescein is a nonspecific organic fluorophore with an absorption maximum at 494 nm emission maximum at ~521 nm. Due to the high absorption of visible light by hemoglobin, fluorescein can be detected generally in superficial tissues (<5 mm). Fluorescein is primarily used in eye diagnostics and retinal angiography studies,25,57,191 including measurements of retinal perfusion177 but its use in other applications has been considered for example in cranial base surgery,152 thoracoscopy,143 glioma surgery,114 colorectal cancer124 or to monitor aortic valve replacement surgery.159

In addition to fluorescein, indocyanine green (ICG)168 has also been considered as an alternative organic dye for optical imaging applications operating in the near-infrared. Since operation in the near-infrared allows imaging deeper than in the visible, ICG has been more recently employed in several interventional applications, for example for sentinel lymph node (SLN) identification in breast,202 gastric,101 cervical34 and vulvar33 cancer, for angiography during reconstructive surgery,39,73,113,185,210 skin flap perfusion,78 lymphatic drainage mapping1,172 and for imaging of glioma70 and liver cancer66,89 and metastases.203 Results obtained with ICG during a hepatectomy procedure can be seen in Fig. 3.

Figure 3
figure 3

Hepatocellular carcinoma cases observed using fluorescence imaging and confirmed by H&E staining. (a, b) Hepatectomy procedure in a 59 years old patient suffering from HCC. a Intraoperative ICG fluorescence imaging technique shows bright signals in the primary HCC nodule (thin arrow) and a new single nodule (arrowhead). White broken lines show the planned liver resection line. (b) A photomicrograph of the new single nodule section stained by H&E (original magnification, ×100). (c, d) Hepatectomy of a case of HCC in a 75 years-old patient. (c) Intraoperative ICG fluorescence imaging technique shows bright signals in the primary HCC nodule (thin arrow) and new multiple nodules (arrowheads). A white broken line shows the planned resection line of liver. (d) A photomicrograph of the new multiple nodules (arrowheads) section stained by H&E (original magnification, ×10)

Both fluorescein and ICG are organic dyes with the ability to visualize the vascular system. Their consideration for cancer imaging relates to possible retention of the dye by tumors, due to the enhanced permeability and retention (EPR) effect of tumor vasculature.66,89,147 In particular for ICG, Ishizawa et al.89 noted that after i.v. injection ICG was retained in hepatocellular carcinoma (HCC) as a result of biliary excretion disorders in cancerous tissues. The subsequent study included 49 patients and confirmed that tumor imaging of HCC and of metastases of colorectal carcinoma is feasible with this—otherwise tumor unspecific-technique. In another study, Adams et al.1 successfully tracked the lymphatic propulsion of an ICG micro-dose in patients suffering lymphedema and could evaluate treatment efficacy.

Despite the relatively wide use of organic dyes for in vivo staining of tissue vasculature, their use in detecting major diseases has not yet shown of high clinical relevance. To differentiate genetic diseases based on more specific biochemical signals, photosensitizers have been considered for intra-operative detection. In particular, administration of 5-aminolevulinic acid (ALA) induces tissue fluorescence by enabling the production of porphyrins in cancer cells, a process that has been studied for intra-operative identification of gliomas of grade III or IV.52,85,136,181,199,211 Stummer et al.181 in a randomized multicenter clinical trial included 322 patients and reported a 65% complete tumor resection for the ALA administered patients compared with only 36% tumor removal under white light guidance. In addition to glioma surgery, ALA-induced protoporphyrin IX (PpIX) has been also shown potent in the detection of bladder cancer85,93 and of basal cell carcinoma.59,166,215 Remarkably, the hexyl-ester of ALA, hexaminolevulinate, has shown an interesting impact on fluorescence cystoscopy in patients presenting non-muscle invasive bladder cancer, increasing the detection and surgical outcome of bladder cancer resections.130,212,213

On top of the detection improvement induced by the use of such compounds, they enable in situ photodynamic therapy. Such a strategy of using intense light to destroy cancerous cells with photosensitizers is currently used in several cancers (head, neck, esophagus, skin, lung, biliary tract) 43,53,64,91,222 with a variety of photosensitizers, mostly porphyrin derivatives, such as Photofrin134 of Foscan,160 and porphyrin precursors such as the previously cited ALA.127

Finally, an important class of exogenous fluorescent agents relates to targeted agents with specificity to cellular and molecular targets. A common form of such agents relates to labeled antibodies and anti-body fragments but significantly more elaborate chemistries have been considered, including activatable probes, i.e., probes that are dark in their base state but fluorescence after interaction with a molecular target. These probes can reveal specific characteristics of complex disease such as cancer, atherosclerosis and other multi-factorial diseases, by illuminating changes in the expression profiles, signaling pathways and other molecular abnormalities. A significant advantage of this approach is that contrast can be engineered so as to reveal specific tissue and disease bio-markers. In that respect it can capitalize on “-omics” approaches and large scale discovery of disease profiles at the systems level. Despite however of the strong potential of the approach, an important limitation concerns the navigation of the corresponding regulatory procedures and ultimately the safety of these agents for human use. The particular issue in this case is the risk associated with the clinical translation of such agents to human. Conversely with the first clinical translation of a folate targeting agent has been now achieved204 and demonstrated on surgery of ovarian cancer patients, a significant milestone associated with the clinical translation of targeted agents has been reached, hopefully serving as a stepping stone to the translation of additional agents enabling a larger disease targeting ability.

Quantitative Fluorescence Imaging and Data Standardization

Fluorescence imaging systems typically comprise of an illumination source emitting continuous wave (CW) light at one or multiple wavelengths (or spectral bands) and a lens-based charge-couple device (CCD) camera using appropriate filters for imaging the fluorescence emitted from tissue.18,146,194 The most common operation mode is epi-illumination, whereby the source and the detector are on the same side of the object imaged. Several variations of this approach have been considered, a common approach being the combination of a fluorescence sensing camera and a color camera.194

Although planar fluorescence imaging appears straightforward to implement and interpret, preclinical and clinical studies have demonstrated several limitations, which compromise accurate fluorescence quantification in common epi-illumination implementations.18,146,179 Perhaps the most important aspect is that the fluorescence images captured do not reflect the true fluorochrome bio-distribution, but rather a mixed contribution of fluorochrome bio-distribution and tissue optical properties, i.e., absorption and scattering. In this way, variations of tissue optical properties for example will modify the fluorescence image and may lead to artifacts and interpretation errors. Tissue absorption variations typically relate to oxygenated and deoxygenated hemoglobin (HbO2 and Hb) concentration differences between tissues or lesions whereas tissue scattering variations relate to cellular density and morphology diversity among them. For example a typical problem during surgery is the presence of blood on the lesion that attenuates the fluorescence signal, but such effects can more generally occur when imaging lesions of different vasculature and cellular density are imaged, which implies different amounts of hemoglobin and scattering within the lesion.

Overall methods that overcome these limitations, for example ratiometric fluorescence imaging approaches15,194 are important to offer accurate performance and enable data standardization, an important aspect for multi-center clinical trials. These approaches are implemented as multi-spectral methods, simultaneously collecting in different spectral bands and processing the data, preferably in real-time, in order to provide accurate fluorescence images. A three-imaging-channels approach, where color, fluorescence and attenuation measurements are simultaneously captured and processed was introduced for correcting for absorption variations of tissue in the surgical field.194 In this case the light-absorption correction is based on a ratio of fluorescence over light attenuation images and the system performance was illustrated in in vivo studies. Figure 4 illustrate this three channels method using an exogenous contrast agent to highlight the lymph nodes in a mouse model. Other methods have considered a double ratiometric technique, where attenuation and fluorescence images were captured for two different excitation wavelengths and were subsequently utilized in phantoms and clinically in patients undergoing prostatectomy.16 The ongoing work pioneered by Andersson-Engels and co-workers in the field of fluorescence data processing is currently in the process of clinical translation for real time monitoring of cancer treatment.94,184

Figure 4
figure 4

Quantitative fluorescence modality. As shown on the left hand side, the apparatus encompasses three cameras to acquire, respectively color, intrinsic fluorescence and excitation fluorescence epi-illumination images using adequate filters for the beam and provide through data processing normalized fluorescence images (where the instrinsic signal is removed from the fluorescence signal) in real time.194 Postmortem imaging of the surgically exposed mouse abdominal area in epi-illumination mode allows the simultaneous acquisition of: (a) a color image showing the lumbar lymph nodes around the inferior vena cava; (b) an image obtained at the excitation wavelength (intrinsic image), showing absorption differences between the two lymph nodes due to the differential injection of an absorber; (c) a conventional fluorescence image showing low signal intensity from the lymph nodes compared to bright background signals; (d) normalized image displaying markedly improved fluorescence quantification, correctly resolving the underlying fluorescence activity in the nodes

It is also possible to highlight the localization of fluorescent probes using fluorescence lifetime-resolved imaging.29 Recent advances have even pushed this method to in vivo samples, with an emphasis on intra-operative methods, albeit with a restricted field of view.74,176,182 It capitalizes on the different of lifetime of the fluorescent probe with the background fluorescence, as well as the difference of lifetime of the contrast agent depending on its immediate surroundings which is interesting in the case of fluorescent metabolites and activatable probes.197 Time-gated spectroscopy, aiming at separating photon based on their scattering pathway, is also being investigated by different groups in vivo and presents interesting results.12,183,200

Overall, the development of methods that account for photon tissue interactions in real-time are increasingly considered, especially in association with a broader acceptance of fluorescence imaging in intra-operative procedures.

Opto-acoustic Imaging

Opto-acoustic imaging is a hybrid imaging modality, using light as a source and acoustics for signal detection. Its major advantage is that image formation is not limited by photon scattering, since no photons but US waves are detected. Therefore the resolution of this optical imaging approach is defined by the US diffraction limit and can reach a few tens of microns or better. When a molecule is illuminated with light pulses (typically in the nanosecond range), the transient temperature increase causes a thermoelastic expansion. This in turns generates ultrasonic waves, ranging from a few kHz to several MHz, which can be detected by ultrasonic detectors After data treatment by mathematical procedures, an image of optical absorption inside tissue can be obtained.144 If a single wavelength is used, mapping of strong endogenous absorbers is possible, such as blood or melanin. The detection of exogenously administered chromophores is also possible especially with multi-spectral illumination techniques. Multispectral opto-acoustic tomography (MSOT) collects data after illuminating tissue at multiple wavelengths, and by applying spectral processing techniques it can then derive images of the distribution of different optical dyes or targeted probes in vivo through tissues at US resolution.145 It has been shown that detection of optical contrast can be imaged with resolutions of 30–40 μm through at least 4–6 mm of tissue.81 Deeper penetration through 1.5–2 cm has been also achieved with resolutions at the 100–200 μm range. Video rate imaging has also shown possible and has been demonstrated in visualizing kidney perfusion in mice,21 as shown in Fig. 5 or in studying tumor vascularization.95 High-resolution video rate optical imaging through several millimeters to centimeters of tissue penetration are the characteristics of a method with great potential toward interventional imaging.41 Several clinical implementations can be foreseen following the general hardware characteristics of corresponding ultrasonic probes with the addition of light generation and guiding components to impart opto-acoustic responses. Handheld devices for opto-acoustic imaging have been described so far, also in setups that also implement conventional US imaging.141 Endoscopic imaging for the upper and lower gastro-intestinal tract, vascular imaging and in general hollow organ imaging have also been proposed.217 Overall, the imaging and clinical outcome performance of these approaches needs to be established.

Figure 5
figure 5

Perfusion of a mouse kidney with indocyanide green imaged by opto-acoustic tomography. On the left panel (a), images obtained by opto-acoustic tomography at 800 nm illumination wavelength. On the right panel (b), image difference at different wavelength to highlight ICG distribution overlayed on the image acquired a 800 nm

Opto-acoustic interventional imaging can prove useful in a number of applications, such as micro-vasculature characterization,82 blood oxygen saturation and volume mapping and imaging of the bio-distribution of optical agents employed to outline anatomical, functional or molecular disease biomarkers. A wide array of optical agents can be employed spanning from carbon or gold nano-particles to organic dyes or even fluorochromes.104 These reporter molecules can be considered as stand-alone or as labels of molecules, antibodies or antibody fragments in order to target specific cellular and sub-cellular characteristics of a disease. For example organic dyes can be used to outline reconnected vessels and visualize flow and perfusion. Functionalized dyes or nano-particles can also label moieties that bind specific cellular receptors, in order to target a set of biomarkers. Nanoparticles, such as gold nanorods198 and carbon nanotubes,132 are known to be good light absorbers in the near infra-red region of the light spectrum. They can, similarly to purely organic contrast agent, be chemically modified to display specific properties such as a targeting capacity. On top of that, they can also be used as nanocarriers for drugs,164 or as photodynamic therapy agents in cancer treatment.27

Opto-acoustic imaging is a highly scalable approach since depth can be exchanged for resolution: the more superficial the imaging the higher the resolution achieved and vice versa. Microscopic implementations using opto-acoustic have been described since the early 70s. However microscopic implementation also come with reduced fields of view; therefore in principle opto-acoustic imaging can also be considered in the context of the microscopic methods that follow in the next section.

Microscopic Optical Methods

Optical Coherence Tomography (OCT)

Optical coherence tomography, developed during the last two decades, is an optical approach that maps local changes in the refractive index of a tissue in 2D and 3D. It is usually exemplified as the optical version of US, whereby it detects depth-resolved local reflectivity of biological samples by utilizing interferometry. OCT provides real-time imaging of tissue structures (morphology) with high resolution (1–15 μm), while its penetration depth, limited by the scattering properties of the tissue and the wavelength used, in biological samples is typically around 1–2 mm. Many novel implementations of OCT systems are developed in order to incorporate functional properties to the technique. Thus, novel OCT systems are able to distinguish between polarization (PS-OCT)106 and phase differences190 or feature Doppler capabilities178,218 for blood flow imaging allowing, thus, the detection of more disease-related tissue properties. OCT has found diverse applications in clinical research and is widely used predominantly in ophthalmology for the screening of various diseases,10,105,188,219 providing an useful non-contact diagnosis tool, but also in dermatology67,77 and cardiology6,126,156,189,216 where clinical translation is ongoing.79 Figure 6 shows images of the human retinal micro-capillary network obtained by classical fluorescein angiography and phase-variance OCT for comparison purpose.

Figure 6
figure 6

In vivo human retinal micro-capillary network images. Image size is 1.5 × 1.5 mm2. (a) Fluorescein angiography. (b) Projection view of retinal contributions to phase-variance (pvOCT). (c) RGB depth color-coded projection view of retinal pvOCT data. Upper right shows depth scale bar with red color denoting a top layer, green showing capillaries as an intermediate vascular bed, and finally blue represents micro-capillaries as a deeper vascular plexus layer. The imaging acquisition time of pvOCT is 3.5 s. Scale bar: 250 μm

Optical coherence tomography endoscopes have emerged, based on optical fibers for light delivery and detection. OCT endoscopes have been described in combination with imaging of the gastro-intestinal tract51,106,218,227 or the larynx22,173 for tumor detection as well as in cardiac intra-vascular imaging for imaging atherosclerotic plaques or stents,19,227 the latter being particularly useful for post-implantation monitoring of stent status. OCT has also been considered in the context of surgical procedures with applications to lymph node and tumor imaging.138 Gallwas et al.58 investigated the capability of OCT to detect and differentiate between various stages of cervical intraepithelial neoplasia (CIN) using colposcopy for the OCT guidance. OCT imaging results for 120 patients compared with histology, showed 86% sensitivity and 64% specificity using CIN grade 2 as a threshold. A larger number of applications have been reviewed by Marschall et al.131 Multi-modal implementations using OCT have also been suggested for improving imaging performance, for example the combination of OCT with multi-photon tomography,109,207 Raman spectroscopy, or opto-acoustic imaging.123

In the recent years, in vivo optical coherence microscopy (OCM) started appearing3,28,120 and making their way into hand-held probes,4,110 to try to overcome to transverse resolution limit of classical OCT, albeit with a more limited field of view. In that respect, it becomes complementary to confocal microscopy (CM) since it relies on different contrast sources and presents different imaging properties depending on sample thickness.

Endoscopic Confocal Microscopy

Confocal microscopy provides optical diffraction-limited resolution imaging even when imaging thick tissues; it is therefore appropriate for in vivo imaging applications, in contrast to conventional microscopy which requires tissue slices of 5–10 μm in order to offer high resolution imaging. CM rejects scattered photons by means of a small pinhole aperture placed in front of the detector. The implementation of CM around flexible fiber-based endoscopes has enabled the clinical propagation of the technology, since the flexible endoscope can be brought in contact with tissue to investigate the area under it. Although limited field of view can be achieved (<mm), endogenous and exogenous contrast agents can be visualized in high resolution (<1 μm lateral; <10 μm axial), within the first 100 μm of tissue. Typical frame rates achieved is in the few Hz. CM makes use of autofluorescence,102,195 or extrinsically administered agents including fluorescein,75,103,206 ALA-induced PpIX,196,224 the fluorescent agent acriflavine187 for confocal fluorescence imaging, or acetic acid24,30,129,187 has been locally applied to cause “brightening” of the cell nuclei which, in turn, enables the visualization of changes in nuclear size and density, and N/C ratio (nuclei to cytoplasm).128 Typical images obtained by this method are presented and compared to classical H&E staining in Fig. 7. Collier et al.30 reported 100% sensitivity and specificity for 28 fresh samples from patients with cervical intra epithelial neoplasia using acetic acid for the contrast build up and a processing algorithm for the scoring. Furthermore they introduced and evaluated various metrics for the classification of tissue samples based on nuclear size and cell separation distance.

Figure 7
figure 7

(a, c) Confocal reflectance and (b, d) H&E images of tissue slices from oral mucosa biopsies of the same patient. Confocal images were taken after applying 6% acetic acid to the tissue slices. The white line in image (a) defines the basement membrane. White circular structures in the confocal images are cell nuclei. Scale bar: 200 μm

Confocal imaging has been considered in many clinical applications, including the detection of oral,24,75,129,196,224 cervical30,187 or skin62 cancer, Barrett’s esophagus103,154 and in ophthalmology.48,102 Haxel et al.75 used a flexible confocal endoscope to image the oral region after i.v. injection of fluorescein and could successfully discriminate cancerous and healthy tissue, while Kiesslich et al.103 reported a sensitivity and specificity of over 95% in 63 patients screened with confocal fluorescence microscopy for Barrett’s esophagus. Finally, others followed dual-contrast approaches for real-time histology, by simultaneously using fluorescein and acriflavine206 or acriflavine and ex vivo specific molecular agents24 and compared them to histological staining.

Confocal microscopy is limited to imaging only a few tens to a few hundred of microns deep. This reduces the applicability of the method in imaging large volumes. CM can nevertheless provide a useful tool for high-resolution local imaging of suspicious areas and has been proposed for defining resection perimeters in oncology in real time,139 also in combination with the use of exogenous contrast agents, for example in neurological oncology.165 The clinical interventional application of the method may be significantly augmented when applied in combination with large field of view methods, in particular fluorescence imaging, the latter used to guide CM imaging to suspicious tissue segments.

Nonlinear Microscopy (2-Photon Microscopy, Second Harmonic Generation, CARS)

Nonlinear microscopy offers several attractive imaging characteristics and is considered as an alternative to confocal imaging. 2-photon microscopy (2PM) and second harmonic generation (SHG) are nonlinear phenomena of photon-tissue interaction. 2PM refers to the simultaneous absorption of two photons of long wavelengths that causes a fluorescence emission of one photon of shorter wavelength. SHG describes the characteristic ability of some molecules to absorb two photons and subsequently scatter a photon of exactly double frequency (SHG) without energy loss.76 Higher order interactions are also possible, such as in multi-photon absorption techniques. Although the nonlinear excitation process has been documented since the 1930 s, the first practical implementations developed with the advent of femtosecond (fs) lasers and became popular over the last two decades. As nonlinear phenomena are inherently rarer compared with linear phenomena (i.e., fluorescence), nonlinear imaging necessitates a photon flux concentrated in space and time. The major advantage of 2PM is that the typical excitation range for intrinsic fluorophores (and extrinsic that fluoresce in the green (i.e., GFP) wavelengths) is between 700 and 900 nm, allowing, thus, a deeper penetration inside the tissue (~1 mm) due to reduced absorption and scattering.44,192 Although safety concerns need to be considered, in relation to the use of highly concentrated laser powers for signal generation, multi-photon microscopy is broadly applied to imaging cell cultures,158 animals68,99,169 or ex vivo human samples, in particular the skin.7,69,92,103,108,162,205 It will however hinder its clinical in situ use. Ericson et al. also investigated drug delivery in the skin using 2PM.47 Jiang et al.92 investigated ex vivo healthy human samples from colon, stomach and skin tissue and imaged 2-photon excited autofluorescence and SHG-excited collagen using a 800 nm excitation. Additionally Rogart et al.162 showcased the ability of multiphoton microscopy to image the autofluorescence of fresh, unstained colonic mucosa samples with higher resolution compared to CM. In the last couple of years, endoscopes for in vivo imaging based on the two-photon imaging technique have started to emerge, and have to ability to display sub-cellular details.118 However, their use is currently limited to research, typically in ophthalmology.223

Coherent anti-Stokes Raman scattering (CARS) microscopy, the nonlinear analog of Raman spectroscopy, is a multiphoton tissue-photon interaction capable of imaging tissue-intrinsic molecular vibrations. Two picosecond lasers of different frequencies (‘pump’ and ‘Stokes’) interact with tissue to give rise to vibrations in resonant molecules resulting in a strong anti-Stokes signal. Being a nonlinear effect, the CARS signal is similarly confined in space and allows high-resolution three-dimensional imaging with sub-micron spatial resolution. It can be combined with other nonlinear techniques (2PM, SHG) and provide additional chemical information. So far, CARS has been used to image lipids and exogenous mineral oils49,225 in cell cultures,31,150,153,167,209 ex vivo and in vivo in small animals49,50,56,119,125,153,225,226 or ex vivo human samples.60 Recently Evans et al.49 reported a CARS microscope capable of real time video rate imaging (20 fps) in small animals in vivo. They successfully visualized the lipid structure in mouse ear and recorded the perfusion of extrinsically administered mineral oil. Lim et al.125 investigated the effects of high fat diet on atherosclerotic plaque formation in ApoE- deficient mice, using CARS, SHG and 2PM autofluorescence microscopies to image lipids, collagen and elastin, respectively. Figure 8 summarizes the theory behind non-linear optical microscopy and provides an example of a mouse aorta imaged by 2PM, SHG and CARS.

Figure 8
figure 8

Two-photon absorption, SHG and CARS. Top: Jablonski diagram of the different modalities. Note that in second-harmonic generation and Raman scattering no actual electronic excitation takes place. Bottom: Representative composite images of the entire descending aorta in ApoE-deficient mice fed a Western diet. Images from three separate channels—Two-photon excitation fluorescence (a), SHG (b), and CARS (c)—were acquired simultaneously, with all three channels visible in (d). Each image consists of several individual image tiles, cropped and stitched together (approximately 8 × 10 tiles per composite image). Scale bar, lower left of each image = 500 μm

Ongoing research in the field of CARS microscopy aims to identify the suitable vibrational spectral regions that exhibit additional, medically relevant chemical and molecular signatures. Such generalization of the approach could open diagnostic and therapeutic applications, including applications to interventional imaging. Interestingly, CARS has been very recently combined with non-linear excitation of auto-fluorescence for application in dermatology.107

Point Spectroscopy (Reflectance, Fluorescence, Raman)

The term ‘point spectroscopy’ refers to fiber-optic implementations that measure spectra from confined “point” volumes on the surface of healthy and diseased tissues. Although not a purely ‘imaging’ technique, point spectroscopy provides detailed information regarding the back-reflected, fluorescence and vibrational (Raman) spectral properties of tissues. The acquired spectra are typically mathematically processed to reveal the exogenous122 or endogenous tissue fluorescence, measure the local tissue absorption and scattering coefficients (reflectance spectroscopy) or provide information about cell nuclei size (light scattering spectroscopy) and lipids (Raman). These tissue parameters represent biochemical and structural information and can offer information that can differentiate disease from healthy tissues. Point spectroscopy techniques have been used to detect breast,71 oral,122,135,170,171 brain,17 cervical63 and colon174 cancer and atherosclerotic plaques.167 Müller et al.135 employed a tri-modal spectroscopy approach combining autofluorescence, diffuse reflectance and light scattering spectroscopy to distinguish normal from dysplastic and cancerous oral tissue. The study included 23 patients and showed that the method could differentiate between normal and malignant tissue with 96% sensitivity and 96% specificity and between cancerous and dysplastic tissue with 64% sensitivity and 90% specificity.

Discussion and Conclusion

The above presentation of bio-photonic technologies is not exhaustive of the existing literature on novel implementations, alternatives to these techniques and proof of principle studies. The methods presented however define major directions that have either considered in clinical studies or bring imaging performance that attains great promise for clinical application. Regardless, the clinical propagation and acceptance of optical technologies has been slow. One factor that affects this is that there is a large variety of photonic technologies, many of them developed with a technology-drive or a biological application drive but not around a clinical application. Therefore it becomes challenging to select the appropriate technology, match it with the complex and diverse requirements of a clinical need, and prove clinical impact. An additional limiting factor of clinical propagation is naturally the performance and accuracy of the method in resolving disease specific biomarkers.

Many researchers are increasingly orientating towards multimodal approaches in a ‘best of both worlds’ attempt to bridge the limitations and extend the applicability of each technique. An important aspect in the multi-modal combination is the ability to reveal different contrast from each modality in order to reveal disease specific biomarkers. This however is not straightforward; the ability to combine multiple information collected from tissue in order to devise an approach with detection sensitivity and specificity is often confronted with limitations associated with the heterogeneity of disease from patient to patient or even within the same patient. While a large number of optical methods are geared towards detecting endogenous contrast (vascularization, hypoxia, auto-fluorescence, presence of lipids or NADH with CARS, 2PM, etc.), having the obvious advantage of not requiring injection of specific probes, the utilization of induced contrast, in particular contrast with biochemical or molecular specificity may prove an important parameter in improving diagnostics and theranostics in interventional procedures. For example, the emergence of 5-ALA fluorescence for glioma surgery180 or the more recent use of a targeted fluorochrome agent in ovarian cancer surgery204 are cases that combine biochemical/molecular contrast with wide field of view visualization capacity that can lead to practical implementations for enhancing interventional vision. Such methods can be further used to guide small field of view methods such as interventional CM or opto-acoustic imaging, in order to offer increased visualization capacities. Similarly, a promising alternative is the combination of macroscopic fluorescence imaging for biopsy guidance. Furthermore, ‘dual modality’ tumor targeted contrast agents are currently investigated9,14,111,140,148 for the visualization of cancer with multiple modalities (MRI, PET, fluorescence) so that interventional imaging can be combined with pre-operative imaging.

Even though wide-field fluorescence imaging appears straightforward to implement, the approach is still limited by photon scattering. The approach cannot visualize with the resolution of microscopy methods and the fluorescence intensity is affected by tissue optical properties. This can lead to false negatives and false positives, unless special care is given to normalize the images for the effects of optical property variation, for accurate “medical-grade” images. Quantitative (normalized) planar fluorescence imaging, implemented in real-time mode can provide instant feedback to the surgeon and importantly, directly relate to his surgical or endoscopic vision, offering a natural extension of imaging in the operating room. With the addition of disease targeted fluorescence probe, the clinical sensitivity and specificity of the technique can be improved beyond the metrics achieved when visualizing intrinsic contrast. This “engineered” contrast can lead to significant flexibility in identifying disease but also important anatomical features, such as nerves.

In some applications, such as imaging of the upper and lower gastrointestinal tract, modalities that can be included in an endoscopic probe present a clear advantage. Tumor detection by means of autofluorescence imaging, confocal endo-microscopy, NBI, two photon microscopy and optical coherence imaging is becoming more and more widely accepted as a more reliable diagnosis or surgical guidance tool and should ultimately replace, or at the very least compliment, classical WLE in selective applications. Relying on various contrast sources for each methods (be it endogenous or exogenous), they possess physical characteristics (penetration depth, resolution, field of view) that make them also suitable to replace classical tissue sectioning methods, with which they also share the need of a healthy reference to give definitive medical information and avoid false positive and false negative results. In that respect, using a multimodal approach to combine data, ideally anatomical and functional, should provide the clinician with the right tool to make the right decisions at all times.

In addition, opto-acoustic imaging is emerging as a modality with multiple advantages since it can provide anatomical, physiological and molecular high-resolution images from several millimeters to centimeters deep in tissue. The combination with wide field-of-view methods and the use of multi-spectral illumination for biomarker quantification comes with significant potential for interventional application as well. It is also foreseen that integration of opto-acoustic imagers in endoscopic probes, similar to the one commercially available for ultrasonic applications, will soon be performing well enough to be a credible alternative to other endoscopic modalities.

In conclusion, a variety of optical imaging techniques have emerged with the potential to provide intra-operatively guidance for cancer surgery and trustable evaluation of tumor margins. Since the most common surgical ‘modality’ is the visual inspection of tissue, epi-illumination fluorescence imaging is perhaps the most straightforward way to enhance the surgeon’s vision by providing contrast between malignant and benign tissue. Featuring various sources of intrinsic and extrinsic tissue contrast, optical modalities have shown great potential for the future. With the increasing development of endoscopic implementations and minimal invasive surgery approaches, advanced optical imaging technologies are expected to play an increasingly important role in the future with the promise to improve overall survival and quality of life.