Confocal Endomicroscopy

Christina E. Sarris and Nader Sanai

Abstract
Fluorescence image-guided surgery is rapidly demonstrating its ability to improve brain tumor resections. However, there are still a number of limitations inherent to all wide-field fluorescence imaging techniques, such as limited sensitivity to detect glioma infiltration at the margins and ambiguous image contrast. Confocal endomicroscopy is a complimentary technology to targeted fluorescence in tumor resection, and it is certain that as our technology for fluorescent probes continues to evolve, the confocal microscope will continue to be refined. Recent work suggests that intraoperative high-resolution microscopy, a real-time alternative to invasive biopsy and histopathology, has the potential to better quantify tumor burden at the final stages of surgery and ultimately to improve patient outcomes when combined with wide-field imaging approaches. Additional studies are needed to further elucidate the clinical benefits of these new technologies for brain tumor patients.

Keywords: confocal endomicroscopy, high-resolution microscopy, fluorescence, extent of resection, histopathology, brain tumor

15.1 Introduction

Postoperative patient survival for brain tumor surgery is largely dependent on extent of resection. However, surgeons continue to face challenges in intraoperative differentiation of healthy brain parenchyma from pathologic tissue. Improvements in microscope technology, along with development of novel fluorescent agents, are helping neurosurgeons overcome these challenges as we become better able to characterize these boarder zones in tumor resection. Wide-field fluorescence image-guided surgery (FIGS) has been reported to have many benefits in glioma resection, but it does have several challenges that limit its utility. High-resolution confocal endomicroscopy has recently been introduced to the neurosurgical armamentarium and is demonstrating benefit in brain tumor resection. As this technology continues to improve and neurosurgeons become more comfortable with its incorporation into operative workflow, confocal endomicroscopy may contribute significantly to better tumor resections and patient outcomes.

15.2 Wide-Field Microscopy and Tumor Surgery

Visualization in fluorescence microscopy relies on fluorescence and phosphorescence rather than reflection of light as in conventional white-light microscopy. In fluorescence microscopy, the specimen is instead illuminated by a particular wavelength of light. The development of fluorescence microscopy has been motivated by the observation that reflectance imaging does not provide sufficient and reliable image contrast to distinguish between healthy and pathologic tissues. It has been well established that fluorescence can aid in this distinction.

One such type of fluorescence microscopy is wide-field fluorescence imaging. Utilization involves a low-powered surgical microscope, which continuously acquires the entire specimen on the microscope stage through the use of an eyepiece, and/or at a rapid frame rate with a digital detector array. Innately, the resolution of this technology is assumed to be low, as the field of view ranges from tens to hundreds of millimeters along any given lateral dimension, with a spatial resolution of tens to hundreds of microns. The most popular commercially available surgical microscopes for wide-field FIGS have been the Zeiss Pentero BLUE400 and the Leica FL400 microscopes, both utilized in numerous clinical studies and approved for routine use in the United States and Europe.

FIGS, however, suffers from a number of challenges 1:

• Angle and working distance: When utilizing wide-field microscopy, one should attempt to maintain a perpendicular angle to the field and a constant working distance between the microscope and the tissue. This becomes challenging as tumor cavities are often out of range for the fixed working distance of the standard fluorescence microscopes used for FIGS, and often side walls are difficult to visualize because of their steep angle with respect to the illumination source. Finally, tumors located within sulci and behind bends are inaccessible without a miniature imaging probe.²

• Ambiguity at margins: Wide-field imaging does not allow for qualitatively or reproducibly defining margins for gliomas. For example, visualization of 5-aminolevulinic acid (5-ALA) induced protoporphyrin IX (PpIX) contrast is effective at revealing regions of bulk tumor (deep red), but fluorescence intensities decay near the margins (lighter pink) and vanish entirely as the tumor cell density continues to decline. On histopathological examination, however, all three territories demonstrate glioma cell infiltration.³

• Sensitivity: Unfortunately, visible fluorescence is not always captured with wide-field FIGS. For low-grade gliomas (LGGs), wide-field FIGS with 5-ALA remains ineffective, as it does not produce visible fluorescence for the majority of low-grade tumors. In some cases, heterogeneous fluorescence has been noted in focal areas of anaplastic transformation. However, the vast majority of LGGs are invisible with 5-ALA.^{4,5,6,7,8,9,10} Interestingly, PpIX fluorescence can be measured ex vivo in LGG tissue following 5-ALA administration.^7,8 In these analyses, the resultant fluorescence intensity of the tumor tissue is significantly higher than in similarly treated normal tissue⁸ and increases with both tumor grade and proliferative index.^3,7,11

• Quantification: The ultimate goal of these novel technologies is to reliably distinguish between pathologic and normal tissues via use of a contrast agent and to accurately quantify the concentration of that contrast agent at each resolvable tissue location. This is not, however, easily done as the image pixel is complicated by background and variations in tissue optical properties. Thus, recent research efforts have been spent with attempts to advance the FIGS systems and correct these issues such that image intensities correlate directly with fluorophore concentration ( Fig. 15.1). For example, advanced FIGS systems are now utilizing multispectral imaging. This involves the illumination and/or detection of fluorophores and backscattered illumination light at various optical wavelengths, which helps correct for tissue optical properties as well as nonspecific accumulation of molecular contrast agents.12,13,14 These methods typically utilize proportional imaging (normalization), either between fluorescence measurements^12,13,14,15 or a combination of fluorescence and light attenuation images.^16,17 The SurgOptix T3-platform is based on one such multispectral imaging technology. Another system under development is the FLARE system,^18,19,20 which has incorporated color imaging with dual-band fluorescence imaging. This system utilizes an automated background subtraction algorithm in its operation, similar to the Zeiss OPMI Pentero system. This allows for isolation of fluorescence from other sources of background light in the operating room.

Fig. 15.1 Comparison of traditional hematoxylin and eosin stained tumor section (left) and confocal view (right) of an anaplastic oligodendroglioma. Both images demonstrate hypercellular tumor with pleomorphic nuclei. Prominent vascular channels (arrows) are seen with both techniques.

While improvements in FIGS technology are ongoing, its innate limitations have allowed clinician scientists to explore other means by which to improve tumor surgery.

15.3 High-Resolution Intraoperative Microscopy

The reported rates of gross total tumor resection continue to remain low despite use of imaging technologies such as intraoperative MRI and wide-field FIGS. As discussed earlier, a major reason for this is that these wide-field imaging techniques lack the resolution, and hence the sensitivity, to detect the disseminated tumor cells at the margins of such diffuse tumors. Thus, the neurosurgeon is faced with the challenge of interpreting subjective image intensities to determine an appropriate surgical margin. This process is neither quantitative nor reproducible for one surgeon or between multiple surgeons. For some extra-axial lesions, the gross appearance of the tumor is sufficient to establish the tumor-brain tissue planes with microdissection techniques. Other lesions, however, are less easily distinguished, particularly in the setting of prior treatment effect, cerebral edema, or microscopic infiltration. This is particularly true for gliomas and higher-grade meningiomas, where defining extent of resection on the basis of gross tissue characteristics is insufficient and neuronavigation can be unreliable due to brain shift.

Beyond identifying tumor margins, the opportunity to intraoperatively define tumor grade and histologic subtype is of critical importance, particularly for intracranial gliomas, where tumor grade is not reliably predicted with either preoperative MRI 21 or stereotactic biopsy.²² Some have thus advocated for the use of frozen-section pathology to confirm tissue status during glioma resection,² which unfortunately is time consuming and relies on the invasive removal of biopsies. Intraoperative frozen-section analysis can be misleading or nondiagnostic, particularly in cases of mechanical tissue disruption from the resection process.^23,24 Such diagnostic unpredictability is further complicated by the inherent heterogeneity of gliomas, which can contain high-grade populations nested within a lowgrade stroma.²⁵ Others have suggested the need for quantitative measurements of PpIX fluorescence, such as with spectral measurement probes.^4,8,26

To overcome these persistent challenges in the resection of complex intra- and extra-axial brain tumors, recent work has been directed toward adapting routine postoperative neuropathology methods into a real-time intraoperative technique. Overwhelmingly, this has led to development of a potentially powerful complement to wide-field FIGS–intraoperative high-resolution optical-sectioning microscopy, or confocal microscopy. Confocal microscopy is an optical imaging technique that uses point illumination and a spatial pinhole to eliminate out-of-focus light in specimens that are thicker than the focal plane, thereby enhancing optical resolution beyond light microscopy and detecting light produced by fluorescence very close to the focal plane. Intraoperative confocal microscopy has miniaturized this approach to enable visualization of live tissue cytoarchitecture with spatial resolution on a cellular level.^27,28,29,30 Ultimately, this allows the physician to have biopsy images displayed in real time to aid in immediate operative decision-making. This microscope can be placed directly in contact with tissues to quantify the presence of labeled cells in real time without the need for excisional biopsy and time-consuming histopathology. While late to become part of the neurosurgical armamentarium, the technology has been widely utilized and has proved feasible in various bodily regions outside the central nervous system, including the colon, pancreas, stomach, and alveoli.^{31,32,33,34,35,36} In neurosurgery, the ability to resolve and detect sparse subpopulations of labeled cells, such as tumors cells at the diffuse margins of a glioma, could provide a standardized quantitative metric by which neurosurgeons may eventually be able to optimize their resections as well as to objectively determine an unambiguous “extent of resection” for their surgeries ( Fig. 15.1).

15.3.1 Confocal Microscope Evolution

Marvin Minsky, American scientist and inventor, first introduced confocal microscopy in 1955 with demonstration that it was possible to obtain optical sections via aid of a pinhole and detector combination.³⁷ Since that time, the last sixty years have seen development of numerous confocal microscopes based off the principle that a pinhole equivalent blocks out-of-focus light.³⁸ The usability of confocal endomicroscopy has been limited by instrument size and ease of surgical incorporation for the neurosurgeon. Until recently, the size of the requisite apparatus limited the technology to examination of excised tissue samples or isolated cells in a bench-top setting. Newer technology, however, features fiberoptic and microscopic miniaturization, substantially expanding its portability and applicability in an in vivo clinical setting.^39,40,41,42 These systems now consist of a miniature handheld probe and movable workstation with an LCD screen. Using a single optical fiber as both the illumination point source and detection pinhole, high-resolution images are acquired and combined with miniaturized scanning and optical systems.⁴³ Thus, the widespread use of confocal endomicroscopy in the gastrointestinal tract has been largely because of the ease of incorporating the technology into the distal tip of conventional video endoscopes. The bladder mucosa, skin, and eye have similarly been studied with in vivo confocal microscopy.^44,45,46,47 Even more recently, in vivo confocal microscopy has been utilized in robotic-assisted radical prostatectomy.⁴⁸

Microscope designs have undergone numerous modifications to achieve optical sectioning such as multiphoton excitation,⁴⁹ single-axis confocal microscopy,^50,51 dual-axis confocal (DAC) microscopy,⁵² and structured illumination.^53,54,55 Designs have also differed in their scanning mechanisms, including proximally scanned coherent fiber bundles,^{56,57,58,59,60,61,62,63,64,65} distally scanned fiber tips,^{66,67,68,69,70} and microelectromechanical systems (MEMS) scanners.^{52,71,72,73,74,75,76,77,78} There are several commercially available confocal microscopy systems, including the Cellvizio (Mauna Kea Technologies, Paris, France) and Optiscan FIVE 1—both of which have been utilized in neurosurgery.^79,80 The first microscope used in vivo for brain tumor resection in humans was the Optiscan system utilized by researchers at the Barrow Neurological Institute (BNI) in Phoenix, AZ.^5,81,82 The Optiscan has a 475 × 475 µm field of view and focal plane depth to 250 µm. As a result of the unique resonant-scanning mechanism used for imaging, this microscope is somewhat limited by a slow frame rate (0.8 frames/s) that leads to motion artifacts and makes the clinical use of the device less effective. The Cellvizio (Mauna Kea Technologies) is a miniature microscope based on coherent fiber-bundle technologies.^{57,58,60,61,62,64,83} These confocal microscopes treat each fiber within the bundle as a separate confocal pinhole for spatial filtering of out-of-focus and scattered light for high-contrast imaging of tissues at modest depths. Proximal scanning allows the distal tip of the device to be extremely small (0.5–3 mm) and flexible. One disadvantage of these technologies is that they often do not allow for axial adjustment of the focal plane since the mechanisms for doing so would significantly increase the size of the distal tip of these devices. While the ability to image deeply is not a fundamental necessity for intraoperative determinations of tissue status, there are practical advantages for being able to adjust the focal plane of an optical-sectioning device during surgery. For example, adjustment of the axial imaging depth allows the surgeon to search for an optimal imaging plane in which the tissues show minimal signs of surgical disruption, and at which signal levels and contrast are optimal. Another limitation to fiber-bundle-based approaches is that current fiber-bundle manufacturers utilize ion-doped glass fibers that create large autofluorescence backgrounds when excited at 405 nm.⁸⁴ This is less of an issue at other excitation wavelengths, such as at 488 nm, but the autofluorescence background limits the ability of these technologies to be utilized for imaging 5-ALA-induced PpIX, in which the optimal absorption peak is at 405 nm. Both the Optiscan and Cellvizio use a 488-nm excitation light, and Cellvizio additionally has a 660-nm single-band excitation light.⁸⁰ The EndoMAG1 by Karl Storz Company has also been evaluated for use in neurosurgery, with a circular scanning field covering 300 × 300 µm and an 80-µm scanning depth.⁸⁵

In addition to miniature optical-sectioning devices that utilize conventional single-axis confocal approaches, recent efforts have been made to develop intraoperative microscopes using an alternative confocal architecture called a DAC microscope or a divided-pupil confocal microscope.^{52,86,87,88,89,90,91,92,93} In the DAC architecture, the illumination and collection beam paths are spatially separated, as opposed to the common-path configuration of typical microscopes. Simulations and experiments have shown that the DAC configuration provides certain benefits in terms of optical-sectioning contrast in tissues, including the ability to image at deeper depths compared to conventional single-axis confocal microscopes.^88,91,94,95 In addition, dual-axis designs, which utilize low-numerical-aperture (NA; weakly focusing) beams as opposed to the high-NA beams preferred for conventional confocal microscopy, have been shown to be scalable in portable devices with diameters ranging from 3 to 10 mm.^90,92,96 These devices have utilized miniature MEMS scanning mirrors to scan an image within tissues. MEMS-scanned microscopes have been shown to enable high frame rate imaging (up to 30 Hz),⁹² which is beneficial in clinical settings to reduce motion artifacts and image blur during handheld use.

15.3.2 Feasibility and Results in Neurosurgery

There are numerous reasons for introducing confocal endomicroscopy to neurosurgery. Aggressive resection with cleaner, near-normal tissue margins is often acceptable in areas that are not adjacent to eloquent brain. However, in certain localized regions near eloquent brain, a real-time high-resolution “optical biopsy” could potentially provide the surgeon with an accurate calibration measurement that, in combination with intraoperative neuronavigation and other surgical cues, may allow significant improvement in extent of resection. In LGGs, the need for intraoperative optical-sectioning microscopy is especially compelling due to the fact that 5-ALA-induced PpIX fluorescence is often not detectable with wide-field FIGS.^4,5,7,8,9,10 Since the metabolic conversion of 5-ALA to fluorescent PpIX is associated with highly proliferative and mitotic tumor cells, only a small subset of tumor cells are fluorescent in LGGs. As a result, wide-field imaging, in which each resolvable pixel represents an average signal from thousands of cells, lacks the sensitivity to detect the presence of these sparse tumor cell populations. On the contrary, it has recently been demonstrated that intraoperative high-resolution microscopy is able to visualize and quantify the presence of sparsely scattered fluorescent cells in LGGs.5 Recent work illustrates the ability of intraoperative high-resolution microscopy to detect and quantify sparse cell populations where wide-field imaging fails.

Bringing the confocal microscope to neurosurgery remained challenging until recent years because of the difficulty of safely introducing such instrumentation to the brain in vivo. The first results of use of confocal endomicroscopy in neurosurgery were published in 2010.³⁰ Using the Optiscan FIVE 1 system, this study demonstrated the feasibility of performing in vivo confocal microscopy in an animal brain tumor model and compared this imaging with conventional histologic images from the same tissue. Using intravenous fluorescein and topical acriflavine dye, handheld confocal imaging produced suitable images that correlated well with corresponding histologic sections, effectively distinguishing tumor versus nontumor with readily identifiable margins by observers without prior neuropathology training ( Fig. 15.2).

The first human feasibility study of confocal microscopy for brain tumor resection was published shortly after.⁸² Thirty-three patients with brain tumors received intravenous sodium fluorescein, followed by confocal imaging with the Optiscan FIVE 1 system probe. Initial examination of tissue integrity during confocal visualization demonstrated good tissue preservation, with intact parenchyma and vasculature on the surface and as deep as 500 µm. Also, areas of neovascularization from tumor-induced angiogenesis could distinguish pathologic from normal parenchyma. Specific confocal features of high-grade glioma—dense cellularity, irregular cellular phenotypes, and neovascularization—were defined. This study established the technology’s potential value during the microsurgical resection of both intra- and extra-axial tumors. For a variety of tumor histologies, including gliomas, meningiomas, hemangioblastomas, and central neurocytomas, the handheld device generated a real-time, fluorescein-enhanced pathological image of sufficient resolution for a neuropathologist to establish a preliminary diagnosis. This preliminary analysis of intraoperative confocal microscopy not only suggested a correlation between imaging and tumor grade, but also demonstrated a capacity to distinguish tumor margins from adjacent parenchyma.

Subsequent work in a series of 50 microsurgical tumor resections with confocal microscopy⁸¹ compared confocal microscopy images with corresponding biopsy samples. The pathologist in a blinded fashion reviewed the images to evaluate their usefulness as a diagnostic tool, and ultimately had a diagnostic accuracy rate of 92.9%. This study further characterized the confocal microscopic features of various tumor types, including gliomas of various grades, meningiomas, schwannomas, and a hemangioblastoma. The fluorescein-contrasted pathological images were found to be of sufficient resolution for a neuropathologist to establish a preliminary diagnosis without a frozen-section sample, as well as the technology’s potential for in vivo identification of infiltrating tumor edge to help maximize resection. A later fluorescein study of intracranial neoplasms in 74 patients demonstrated confocal endomicroscopy specificity and sensitivity were, respectively, 94 and 91% for gliomas and 93 and 97% for meningiomas.⁹⁷