|Home | About | Journals | Submit | Contact Us | Français|
Advances in optical designs are enabling the development of miniature microscopes that can examine tissue in situ for early anatomic and molecular indicators of disease, in real time, and at cellular resolution. These new devices will lead to major changes in how diseases are detected and managed, driving a shift from today’s diagnostic paradigm of biopsy followed by histopathology and recommended therapy, to non-invasive point-of-care diagnosis with possible same-session definitive treatment. This shift may have major implications for the training requirements of future physicians to enable them to interpret real-time in vivo microscopic data, and will also shape the emerging fields of telepathology and telemedicine. Implementation of new technologies into clinical practice is a complex process that requires bridging gaps between clinicians, engineers and scientists. This article provides a forward-looking discussion of these issues, with a focus on malignant and pre-malignant lesions, by first highlighting some of the clinical areas where point-of-care in vivo microscopy could address unmet needs, and then by reviewing the technological challenges that are being addressed, or need to be addressed, for in vivo microscopy to become a standard clinical tool.
Microscopic observation of tissue specimens—prepared according to established protocols in histology for fixation, embedding, sectioning and staining—is regarded by the medical community as a highly reliable method for the diagnosis of disease and is the current standard of care. The science and art of histopathology is based upon the observation of deviations from normal cellular and nuclear morphology, cellular orientation (polarity) and density, as well as an assessment of overall tissue architecture and cellular composition. Immunohistochemistry and in situ hybridization enable high-resolution localization of protein or nucleic acid markers of disease, respectively, with reasonable specificity and sensitivity. While molecular testing in the absence of morphology, via PCR or flow cytometry, for example, also has value in the detection, diagnosis, and monitoring of human malignancy, morphologic visualization of human tissue will continue as the clinical standard for diagnosis of cancer and pre-cancerous conditions for the foreseeable future.
Despite its acceptance as the standard-of-care, there are limitations to conventional histopathologic diagnosis. First is the need to perform an invasive biopsy or resection procedure. Second is its dependence on the ability of the human observer (typically, the pathologist) to make reliable judgments, both as a single observer over time (“intra-observer variation”) and between different observers (“inter-observer variation”). A third issue is the sampling of tissue itself, which depends on removing the appropriate tissue from the patient and then selecting and microscopically reviewing the most relevant portions of the biopsied material. A fourth concern relates to possible artifacts introduced during processing that can impair morphological and/or molecular interpretation. These artifacts include sample desiccation, shrinkage caused by fixation and embedding in non-aqueous materials, and differential loss or alteration of molecular constituents. In short, devitalized tissue sections mounted on glass slides provide pathologists with a limited two-dimensional cross-section of what was once living three-dimensional tissue.
In contrast, we will describe the potential advantages, challenges, and implications of what can be called “point-of-care pathology (POCP),” which consists of real-time morphologic examination, at the cellular level, of living tissues in their native context. Due to space constraints, we will not discuss non-image-based techniques, such as varieties of light-scattering characterization, or reflectance and fluorescence spectroscopy.
In situ examination circumvents the need for ex vivo tissue processing, a key source of variability in the era of molecular diagnostics. While in situ sampling error remains a potential problem, and in vivo molecular imaging has its own set of challenges, POCP has the potential to provide immediate assessment of disease status, improve diagnostic accuracy, guide tissue biopsies, and accelerate diagnostic and therapeutic processes. This development may even permit identification of new molecular and functional signatures of cancer. For example, real-time microscopic visualization of blood flow or the transport of contrast agents could provide early physiological evidence (functional biomarkers) of malignant transformation prior to the onset of altered morphology . In addition, techniques that require no exogenous molecular probes— relying only on the intrinsic optical properties of tissue such as autofluorescence, light-scattering, or fluorescence lifetime—are highly sensitive to environmental properties such as hydration, pH, oxygenation, and electrolyte concentration, and are ideally performed in situ [2,3,4,5,6].
If exogenous agents are to be used for molecularly specific diagnosis, staging, and typing of diseases, they can be chosen from a large number of probes under development [7,8,9]. Advanced, activatable probes have the desirable property of only switching “on” (i.e., generating a signal) when they have reached their target, thereby improving signal-to-background ratios [7,10,11,12,13,14]. In addition, optical imaging is ideally suited for the simultaneous detection of multiple probes, each emitting and/or absorbing light at a different wavelength . These advanced agents can provide optical contrast and report on both the molecular content and environment of living tissue, and thus have the unique potential to advance the field of point-of-care pathology. However, given the enormous scope of this specific topic, including the challenges posed by obtaining regulatory approval, and the fact that such reagents have recently been reviewed elsewhere [7,8,9,10,11,12,13], we will not discuss them further.
The simplest view of in vivo pathology is that the proceduralist (endoscopist, surgeon, or other) obtains diagnostic morphologic information in real time and relies only on the in vivo morphologic findings to make on-the-spot decisions about patient management and intervention. In vivo microscopy would thus move morphologic assessment from post-procedure to intra-procedure, fundamentally changing the interactions of the healthcare team with the patient. In the case of endoscopy, either the endoscopist (e.g., gastroenterologist or surgeon) and the pathologist would have to interact during the procedure in ways that do not impose delays, or the endoscopist would have to make her/his own assessments:
Major efforts are underway to develop automated image-processing algorithms for the analysis of digital pathology images, including those obtained on fresh tissues with miniature microendoscopes [18,19,20]. Such approaches have the benefit of being able to integrate multiple parameters, either contained within a given image, or accessed across multiple imaging modalities (Fig. 1). Adding to the potential value of an automated evaluation, clinical and biochemical data can also be folded into decision-support algorithms . Part of the motivation for these efforts is the concept that in vivo pathology may contain morphologic and functional information beyond what one individual trained in traditional morphologic methods can interpret. In addition, integrating multiple information sources quickly enough to guide intra-procedural decision making could be taxing even for two collaborating physicians (proceduralist and pathologist) on a case. With that in mind, it is worth noting that relevant real-time optical information obtained during the procedure should be captured for possible post-review, either as “snap-shots” and/or as a continuous digital record.
Regardless of clinical practice model, applications in which in vivo pathology could add value may be broadly categorized as either image-guided biopsy or image-guided therapy:
However, it is important to consider the potential limitations of in-vivo microscopy when medical decisions are made in the absence of confirmatory histologic evaluation of tissue. For example, if the intra-procedural decision is not biopsy but immediate treatment (e.g., through localized freezing or heating), the imaged tissue may be obliterated. Intra-procedural diagnostic error may thus remain undetected, leading either to inappropriate or unnecessary therapeutic intervention, and/or the failure to follow a different management path of greater benefit to the patient. Even with the development of highly reliable and validated technologies for in vivo microscopy, rigorous quality management must be instituted, to ensure both appropriate use of the technology, and the competence of the healthcare practitioners (proceduralist and pathologist alike) in an ongoing fashion. Second, the inherent artifacts of histopathology, such as tissue shrinkage, are an almost unconscious part of the knowledge foundation for pathologists. Morphologic assessment of living tissues will require establishment of a new interpretive skill set, particularly since the artifacts introduced by devitalization and fixation will be absent. Until new image-to-disease correlations are established, there is potentially enhanced risk of interpretive error of in vivo microscopic images.
There is often a disconnect between the engineers and scientists who develop new medical technologies – and who are often largely technically driven – and clinical end-users, whose adoption of new methods will be guided by very different incentives. Therefore, for the remainder of this article, we seek to do two things. First, we attempt to survey and project forward, based on organ sites, particular applications that could benefit from point-of-care pathology. Second, we discuss the technical advances in optical hardware that are needed to translate in vivo pathology into the clinic. Many of these advances leverage the application of lasers to various forms of imaging (and simultaneous treatment), and thus exploit novel or unfamiliar optical contrast mechanisms. Possibly the furthest along into the clinic is optical coherence tomography (OCT) and its relative, optical coherence microscopy [23,24]. These reflectance microscopy techniques have overcome the engineer-clinician chasm, achieving enormous penetration in select areas. As they have been thoroughly reviewed in the literature, we will not discuss them in depth in this review. Photo-acoustic methods for deep-tissue imaging, which rely upon the detection of acoustic signals generated by optical absorption of brief laser pulses [25,26], is another extremely promising technology being translated into clinical use, but will for the same reason not be covered here.
Additional novel microscope technologies being applied clinically, but not for in vivo use, are also worthy of mention. First, there are efforts to develop ex vivo three-dimensional pathology devices for rapid bedside microscopic imaging of fresh biopsy samples. Such devices could significantly improve clinical care, for example, by expediting the performance of Mohs micrographic resection of skin tumors [27,28] or the detection of residual disease at the surgical margins of breast cancer patients . Efforts are also ongoing to develop technologies for off-site pathology diagnostics, with the potential to influence global health challenges in resource-limited regions. Examples include the utilization of cell-phone cameras and mobile networks for diagnoses of diseases  such as malaria [31,32,33] and cervical cancer .
Numerous academic and corporate research groups are advancing the field of in vivo microscopy to address point-of-care clinical needs, each with unique challenges in terms of imaging hardware, contrast agents, regulatory approval, integration of these technologies into the clinical and reimbursement workflow, and patient acceptance. Such potential applications can be roughly grouped into three categories: imaging of accessible surfaces, endoscopic imaging of hollow organs, and imaging during surgery or through incisions. Table 1 provides examples of each, along with a list of technical requirements that are relevant for each application. Note that ophthalmic applications have deliberately been left off of this list, as techniques such as scanning laser ophthalmoscopy (SLO) have enjoyed great success in the clinic and are not considered here as an “unmet need” but rather as a model for translating technologies to the clinic.
Common themes appear in Table 1 in terms of both clinical and technical needs. As summarized in the introduction, the predominant clinical paradigms include the detection of diseases, image-guided biopsy, and image-guided surgery or therapy. An additional aim is to monitor responses to therapy, and to track the recurrence of disease, through morphological and/or molecular imaging of targeted contrast agents.
A major technical challenge is the miniaturization of the devices to fit within small hollow organs, within endoscope instrument channels, trocars (for laparoscopy or core-needle biopsy), and within surgical cavities. As far as detection is concerned, large fields of view are always desired, either in individual image frames, or through the stitching together (mosaicing) of overlapping image frames. In certain cases, deep imaging depths may be desired – for example, to monitor the physical penetration of a topically applied chemical therapeutic, or to visualize the entire mucosal (epithelial) layer of a hollow organ. The technical challenges listed in Table 1 are being addressed in various ways, as described in the following section. Figures 2 through through44 show examples of images obtained in clinical and preclinical settings, from accessible organs (Fig. 2), hollow organs (Fig. 3), and during surgical procedures (Fig. 4).
Technical improvements specifically in microscopy-based approaches will be required to advance in vivo pathology, with requirements dictated by the geometric constraints imposed by the organ or tissue site, as well as the distinguishing characteristics of the disease targets. Important parameters include the dimensions of the optical head, the flexibility of the tether, the tissue imaging depth, the frame rate, spatial resolution, field of view, detection modality, and the capability to handle individual or multiple wavelengths. The following section outlines various design considerations along with some current and future approaches towards addressing unmet clinical needs. This section assumes a basic understanding of confocal and multiphoton microscope technologies, as described in numerous introductory and review articles [35,36,37,38,39,40,41]. While multiphoton approaches have been popular for pre-clinical investigations, confocal approaches have been pursued more aggressively for clinical applications. This is in part due to the fact that confocal designs tend to be simpler, utilizing inexpensive low-power lasers as opposed to the high-peak-power pulsed lasers and dispersion-compensating optics and fibers required for multiphoton-based systems.
In terms of in vivo pathology, the specific application will dictate the requirement for spatial resolution. For example, if detecting nuclear atypia is required, then achieving sub-cellular resolution will be important. Alternatively, if molecular imaging of cell-surface proteins with targeted reagents is the goal, then cellular-level resolution may be sufficient. As with resolution, image contrast is a critical variable to consider. Reflectance imaging often provides limited contrast, since only structures with large differences in refractive index compared to the surrounding medium, such as cell membranes and nuclei, can be easily visualized. Fluorescence imaging of exogenously applied fluorophores can generate enhanced contrast of selected targets, provided that they bind specifically to those targets and with high affinity, and that unbound probes in the background are effectively cleared, or compensated for . As mentioned previously, activatable probes that emit optical signals only when they bind to or find themselves in the target region have great potential for increasing achievable signal-to-background [7,10,11,12,13,14]. However, in general, enthusiasm for exogenous labeling approaches have to be tempered by considerations of cost and time required to achieve FDA approval and the difficulties of obtaining adequate return on investment for their developers.
In certain clinical applications, imaging below the surface may be necessary. For example, in tumor resection, molecular reagents may label subsurface cells more accurately than surface cells that have been surgically perturbed. Another example would be to visualize dysplastic changes throughout the epithelial layer of organs such as the colon, where the mucosa is ~ 500 μm thick. Multiphoton microscopy, in particular, has demonstrated an ability to image relatively deeply in a variety of tissues types. Where deep optical sectioning is not required, however, conventional confocal microscopy has been utilized to acquire high-contrast microscopic images near the tissue surface (< 100 μm deep). A variant of conventional confocal microscopy that utilizes low-NA lenses to image in both reflectance and fluorescence modes is termed dual-axis confocal (DAC) microscopy. This technique has demonstrated improved optical sectioning with penetration depths approaching that of multiphoton microscopy [43,44,45]. The DAC architecture can enable laser-scanned imaging over relatively large fields of view with the use of biaxial-scanning MEMS mirrors [46,47,48]. An alternative approach for imaging at shallow depths is to use structured illumination to distinguish between signal generated at the focal plane vs. signal generated elsewhere [49,50,51,52,53]. This is a full-field imaging approach that does not require beam scanning and is thus able to achieve high frame rates, albeit at limited imaging depths in tissues due to issues related to dynamic range and shot noise.
The greatest constraint in the design of in vivo microscopes is size. Optical-sectioning microscopes, which are generally necessary for high-contrast high-resolution cellular imaging in thick tissues, are mature technologies as embodied in large bench-top laboratory instruments. However, adaptation of optical-sectioning microscopes for use in clinical settings has necessitated creative approaches towards miniaturization. In general, to achieve high spatial resolution in microscopy, light must be focused sharply into specimens using high-numerical-aperture (NA) objectives, which are engineered to enable aberration-free imaging over a field of view (FOV). Multi-element compound miniature optics have been developed, using conventional glass lenses as well as injection-molded aspheric lenses, to enable uniform telecentric imaging [54,55,56]. Gradient-index (GRIN) optics, which are cylindrical components with small diameters on the order of 1 mm, have also been utilized for confocal endomicroscopy and multiphoton imaging but are currently limited in NA to about 0.6 [46,57,58,59,60,61,62]. Hybrid designs utilizing conventional optics and GRIN lenses have been developed with larger NAs of up to 0.8 , enabling both greater light collection and higher spatial resolution, while still enabling a small overall probe size.
In terms of packaging, a number of design approaches have been explored to bring miniature microscopes into the clinic, some of which rely on optical fiber components.
One approach towards in vivo microscopy has been to retrofit large, non-fibered systems with long-working-distance objectives or miniature optics, such as GRIN lenses, for imaging in animals and humans. This approach, commonly termed intravital microscopy, has been instrumental in enabling a number of landmark studies in immunology, stem-cell biology, cancer, and other diseases [61,64,65,66,67,68]. Surgically implanted window models have been developed in animals to provide long-term optical access to tissues; alternatively, miniature GRIN-lens microendoscopes have been inserted into deep tissues for long-term imaging studies. One recent application of intravital microscopy has been to perform line scans across individual vessels, or groups of blood vessels, to monitor the passage of fluorescently labeled cells over time [65,69,70,71]. This “in vivo flow cytometry” technique has been used for multicolor investigations of various cell types, including circulating tumor cells, immune cells, and hematopoietic stem cells. The large size and reduced flexibility of these non-fiber-based approaches has made the clinical translation of such devices challenging, except for certain applications in which optical access is unrestricted, such as in the skin and eyes.
Two major forms of miniature flexible microscope designs have emerged: those that utilize an imaging fiber-bundle approach and those that utilize a single optical fiber for delivering laser illumination and for signal collection. Fiber bundles are useful in that the optical scanning mechanism can be located at the proximal end of the system and thus outside of the patient or animal subject. Each individual fiber in the bundle is imaged to a unique point within the sample. By illuminating and collecting light from one fiber at a time, a 2D image may be reconstructed. The fiber-bundle approach has the advantage of requiring few or no moving parts within the distal optical head, which allows for impressive miniaturization of the endomicroscope to sub-millimeter dimensions in some cases [1,5,54,55,72,73,74,75,76,77]. However, such devices generally image at a single depth, since axial scanning would require the addition of a linear actuator within the distal scan head to adjust the position of the focusing optics, thus limiting the extent of miniaturization [56,78]. Several non-confocal versions of this technology have been developed for rapid full-field (non-scanned) camera-based imaging of tissue surfaces [5,19,79,80]. This approach is a favorable choice if high-speed, compact, and inexpensive imaging of a tissue surface is preferred rather than high-contrast subsurface imaging.
In contrast, single-fiber approaches rely upon micro-optics and scanning mechanisms integrated into the distal optical head of the endomicroscope to enable 2D or even 3D imaging. Two major strategies have been used to accomplish this: micro-electro-mechanical systems (MEMS) scanning mirrors [46,48,81,82,83,84,85,86,87], and mechanical fiber/optical component deflectors [88,89,90,91,92]. An alternative method for achieving distal scanning has been achieved in a miniature microscope through the use of spectral encoding, in which the wavelength components of a broadband light source are mapped along one dimension or two dimensions with a diffractive optical element such as a prism or grating [93,94,95,96]. This eliminates the need for scanning along those dimensions. However, it has been largely limited to reflectance imaging since fluorescence emission is broadband, where spatial positions cannot efficiently be encoded by wavelength components [79,97].
The ability to simultaneously image at multiple wavelengths is extremely important in many applications. Multi-color fluorescence microscopy allows for the observation of interactions and correlations between labeled cells and tissue components 64,65,66,67,68,98,99,100,101] as well as for the ratiometric quantification of image intensities [42,102,103,104]. Color-corrected micro-objectives have been developed for in vivo devices [54,55,56]. Diffractive optical elements may also be used to compensate for the effects of chromatic dispersion and are being explored for the development of achromatic GRIN lenses [B. Messerschmidt, GRINTECH GmbH, personal communication] and microscope systems. Often, these approaches may require a sacrifice in diffraction-limited performance and resolution in order to achieve effective color correction.
Clinical considerations such as ease of use and ergonomics cannot be overlooked in the design of instruments for in vivo pathological assessment. As these technologies continue to mature, various refinements will be necessary to encourage clinical adoption, such as the integration of POC microscopes into existing devices. For example, in image-guided surgery applications, miniature microscopes could be integrated with therapeutic tools such as tumor suction catheters for co-localized image detection and treatment. In other cases, POC microscopes could be designed to be deployed through existing platforms, such as the instrument channels in conventional endoscopes and laparoscopes. Additional clinical considerations include the ease of sterilization, the lifetime of the device, and the availability of disposable components. Robotic-assisted surgery, although not yet shown to improve outcome, is maturing and is predicted to increase in clinical use. Integration of in vivo microscopes with surgical robots would be especially beneficial since motion-induced imaging artifacts could be greatly attenuated.
While the subject of software development is beyond the scope of this review, key components in the clinical translation of any in vivo imaging device will include optimizing what is presented to the proceduralist and pathologist (if involved), and the development of image-processing algorithms to assist in interpretation. Among the greatest challenges that will confront imaging scientists, particularly in biomedicine and regardless of modality or dimensional scale, will be to understand how best to combine the virtues of automated analysis with the power of a trained human visual observer. In a field as variable and complex as pathology, for instance, our goal, in the short- to medium-term, should not be to entirely replace human image interpretation, but rather to find ways to optimize the presentation of image data to reduce observer bias and to increase the accuracy and efficiency of analysis.
Optical coherence tomography (OCT) is an example of the successful translation of a non-invasive optical technology to broad application in the medical practice market [105,106]. This market extends beyond the field of ophthalmology into gastroenterology, dermatology, urology, otolaryngology, and others. Lessons learned from its successful implementation include: (a) approval by the Food and Drug Administration (FDA) for clinical utilization; (b) inclusion of the FDA-approved technology in Medicare and Medicaid reimbursement schedules; (c) ease of use and incorporation into the clinical workflow; (d) development of professional and technical expertise amongst healthcare providers; (e) development of a strong evidence base for successful clinical management; and (f) robust interest from the commercial sector in continued investment in this technology.
Any novel technique of in situ microscopy will encounter four major challenges :
Extensive effort is currently underway to answer the first three questions of the “validation” issue. The overall translational challenges are discussed by Wells et al. , who noted that, until primary evidence from the field can be developed, optical technologies must depend on histopathologic evaluation as a “gold standard” for the validation of analytical and clinical utility. Hence, the potentially flawed technology of histopathology remains as a validation portal through which optical technologies, including point-of-care pathology, must pass.
The field of in vivo pathology, enabled by advances in optics, sensors and electronics, is rapidly evolving and being translated into various clinic arenas. Since there is often a disconnect between the clinical end users and the engineers and scientists developing the core technologies, we have attempted to do two things: to outline some of the major clinical needs and scenarios for which in vivo pathology can be effective, as well as to summarize the major technical approaches and challenges in meeting those needs. Each specific clinical application will necessitate its own set of design choices that will require extensive communication between physicians and those developing the technologies. Since complex issues are not easily summarized, this article discusses a limited set of challenges, including the integration of these techniques into the clinical workflow, the use of contrast agents, regulatory approval, reimbursement, robotic aids, and an infrastructure for telepathology. Nonetheless, we hope that discussions such as this can provide a useful framework for progress along this exciting new frontier.
The authors would like to acknowledge research support, in part, from the following sources: NCI - U54CA136465 (C. H. Contag); NCI - P50CA114747 (S. S. Gambhir); and NIBIB - R00EB008557 (J. T. C. Liu).