A roadmap for the clinical implementation of optical-imaging biomarkers

Abstract

Clinical workflows for the non-invasive detection and characterization of disease states could benefit from optical-imaging biomarkers. In this Perspective, we discuss opportunities and challenges towards the clinical implementation of optical-imaging biomarkers for the early detection of cancer by analysing two case studies: the assessment of skin lesions in primary care, and the surveillance of patients with Barrett’s oesophagus in specialist care. We stress the importance of technical and biological validations and clinical-utility assessments, and the need to address implementation bottlenecks. In addition, we define a translational roadmap for the widespread clinical implementation of optical-imaging technologies.

Main

Optical-imaging biomarkers (OIBs), which rely on the interactions of tissue and non-ionizing optical radiation (with typical wavelengths in the range of 400–1,000 nm), can be used for the non-invasive detection and characterization of disease states. OIBs enable the real-time analysis of tissue biochemistry and the use of compact point-of-care and low-cost imaging devices (when compared to radiological imaging), and can operate across ranges of resolutions and depths spanning over four orders of magnitude1.

Across the visible and near-infrared spectrum, light undergoes a range of complex interactions with tissue (Fig. 1). Conventional photographic methods that aim to replicate human vision2 discard most of the information obtained from these interactions and only capture reflected light across three channels (red, green and blue). Over the past decade, a wide range of promising OIBs that extract in-depth information provided by the different light–tissue interactions have emerged. However, for any new imaging biomarker to be deployed in a clinical setting, detailed validation is required. Technical validation defines the precision and accuracy with which the biomarker can be measured, whereas biological validation establishes the association between the biomarker and the underlying physiological, anatomical or pathological process. Clinical validation can then establish whether the biomarker does indeed identify, measure or predict the clinical outcome of interest. To achieve clinical validation, the imaging device needs to conform to clinical performance and safety specifications, and be approved for use in patients.

Fig. 1: Visible and near-infrared light–tissue interactions.
figure1

Visible and near-infrared light experiences a wide range of complex interactions with tissue constituents. OIBs exploit these interactions in clinical applications. The interactions illustrated are: reflection, absorption and scattering (occurring, for example, in white-light imaging, diffuse reflectance spectroscopy, elastic scattering spectroscopy, narrow-band imaging, and multispectral or hyperspectral imaging), phase (in, for example, OCT and angle-resolved low-coherence interferometry), polarization (in, for example, polarimetry and holography), nonlinear effects (in Raman spectroscopy, coherent anti-stokes Raman spectroscopy, multiphoton fluorescence imaging and multi-harmonic generation imaging), fluorescence (in autofluorescence intensity or lifetime imaging of endogenous fluorophores, and in optical molecular imaging of exogenous fluorophores), and the photoacoustic effect (in photoacoustic, or optoacoustic, microscopy and tomography). Line colours represent wavelengths of light. Multicoloured lines represent broadband light. Changes in colour represent changes in wavelength when light interacts with tissue (from green, to orange, to red, to dark red, increasing in wavelength). Perpendicular lines represent wavefronts and thus indicate the optical coherence of phase. Arrows with perpendicular lines represent the orientation of polarization. Curved lines represent the emission of acoustic waves. ICG, indocyanine green; IV, intravenous; NAD(P)H, nicotinamide adenine dinucleotide (phosphate).

With standard radiological imaging—such as computed tomography (CT) or magnetic resonance imaging (MRI)—the imaging device required to measure a novel imaging biomarker is already clinically approved for use in humans and is widely available across radiology departments3. In contrast, for OIBs it is uncommon that a clinically approved imaging device (alongside its associated specialist data-acquisition and data-interpretation methods) is available for clinical validation. Therefore, biological validation may be restricted to testing ex vivo samples, such as histopathological sections. Compared to the in vivo setting, these can be prone to bias and generate a different range of optical interactions, ultimately resulting in misleading conclusions as to the potential clinical utility of the OIB (refs. 4,5). Furthermore, OIBs may be deployed in a range of settings during the patient-management pathway, spanning primary care (such as a family physician) and specialist care (specialist practice, referral or medical centre). Hence, even a well-defined OIB with promising performance in an experimental setting may not receive approval if it fails to adequately address the specific diagnostic question at a defined point in the patient-management pathway.

Given the promise of emerging OIBs, in this Perspective we discuss challenges and opportunities for their clinical translation, from development to implementation in healthcare systems. To this end, we selected two distinct case studies in early cancer detection, and identified a number of OIBs that have reached maturity in clinical use (Table 1). The first case study covers primary care in the context of the assessment of melanocytic skin lesions, and the second covers specialist care in the context of the surveillance of patients with Barrett’s oesophagus. We use trends identified from both cases to define a set of translational characteristics that contribute to the likelihood of a new OIB being incorporated into healthcare, accelerating translation during both technical and biological validation through to the assessment of clinical utility. Taking a page from the recent consensus roadmap for radiological imaging biomarkers3, we propose an OIB roadmap that links the translational characteristics of OIBs and defines the key bottlenecks that must be overcome in order to facilitate a smoother path for their clinical translation.

Table 1 Optical-imaging modalities and selected OIBs currently in development for early cancer detection

Early detection of melanoma in primary care

Although malignant melanoma represents less than 5% of cutaneous malignancies, it causes the majority of skin-cancer deaths6. In the United Kingdom, the current management of patients presenting with a pigmented skin lesion involves a primary-care general practitioner assessing the patient’s history and performing a naked-eye visual inspection of the lesion, guided by a seven-point checklist: lesion size (diameter larger than 7 mm), change in lesion size, and the presence of irregular pigmentation, an irregular border, inflammation, itch or altered sensation, and of oozing or crusting of the lesion. If the lesion appears suspicious on the basis of this visual assessment, the patient is referred to specialist care via an urgent skin-cancer pathway7.

The potential benefit that population-screening programmes could have in malignant melanoma management is evidenced by the improved survival of patients for whom melanomas are detected by physicians rather than by patients or family members8. This occurs because lesion thickness is a key determinant of patient outcome (the five-year survival rate improves to 91% for lesions smaller than 1-mm thick from a mere 46% for lesions larger than 4-mm thick9), and because physicians can typically recognize thinner lesions. A large population study suggested that an almost 50% reduction in mortality rates is possible by adopting full-body skin examinations performed by dermatologists10,11. However, owing to methodological limitations of the study12, it did not provide sufficient evidence to recommend a national population-screening programme.

Because the skin provides an easily accessible surface for optical imaging, OIBs can improve the triage of suspected malignant melanoma in primary care, and provide sensitivity and specificity for diagnosis in specialist care, thus avoiding unnecessary biopsies. Optical-imaging approaches can be broadly categorized as methods that enable visualization of the tissue with spatial resolution (imaging), and methods that map point-based biochemical information (spectroscopy). A few multispectral-imaging techniques are, however, capable of combining both types of information. In what follows, we review devices that have received regulatory approval for use in humans in each of these categories.

Imaging

Dermatologists use a handheld dermoscope (a magnifying optical-imaging instrument) in about 78% of examinations of suspected malignant melanoma13. According to a meta-analysis of 8,487 suspicious skin lesions, a dermoscope improves the detection sensitivity from 71% to 90%, and the specificity from 81% to 90% (ref. 14), with respect to examination solely by the naked eye. Multispectral-imaging methods employ handheld devices to illuminate the tissue with broadband white light, and measure the reflected light at several different wavelengths (typically, up to ten). The resulting spectral images are then processed, often using reference spectral properties of prominent tissue absorbers, such as melanin, haemoglobin, water and lipids. For example, SIAscopy (MedX, Canada) measures the reflected light of the skin at eight different wavelengths. The data are then used to generate maps of melanin, dermal melanin, haemoglobin and collagen in the skin using model-based fitting15. The MoleMate system (MedX, Canada) combines SIAscopy with a primary-care scoring algorithm. The algorithm interprets these maps and classifies lesions as ‘suspicious’, ‘not suspicious’, ‘haemangioma’ or ‘seborrheic keratosis’16. A large primary-care randomized controlled trial17 of the MoleMate system in the United Kingdom showed that incorporating the use of the device into existing best-practice clinical guidelines (including the seven-point checklist) led to similar diagnostic accuracy for ‘suspicious’ lesions as that of current practice18. Furthermore, clinicians and patients rated the device more highly for ‘reassuring and thorough care’18. However, as its use resulted in more referrals from primary care, it was not recommended for use in routine primary care.

Two additional multispectral-imaging methods have been examined for the inspection of malignant melanoma. SkinSpect (Spectral Molecular Imaging, USA) produces melanin and haemoglobin maps similar to those of MoleMate19,20,21,22 also using model-based fitting, so it may be possible to include SkinSpect into established scoring algorithms. MelaFind (Strata Skin Science, USA; formerly known as Mela Science), which has been discontinued, took a different approach. Instead of using automatic segmentation, it performs feature extraction and classification23 to translate the multispectral-imaging data into a binary biomarker classification: ‘biopsy’ or ‘no-biopsy’24. In a study comparing German and United States dermatologists, German dermatologists were less likely to incorporate MelaFind into their decision-making. It was suggested that the German physicians’ older age, more extensive dermoscopy training and unfamiliarity with MelaFind25 contributed to the less-frequent use of the technology, thus indicating that MelaFind may find a niche in assisting less-experienced dermatologists or those without training in dermoscopy. Despite the widespread availability of multispectral-imaging methods for the assessment of suspected malignant melanoma, none of these methods have been incorporated into routine use or into decision-making algorithms.

Reflectance confocal microscopy (RCM) captures high-resolution images of the epidermis and papillary dermis with almost cellular resolution26. These images are interpreted via a feature-based scoring algorithm that classifies a binary biomarker: ‘non-melanoma’ and ‘melanoma’27. RCM has been implemented in the VivaScope imaging systems (Caliber I.D., USA), which were found to significantly increase specificity when compared to dermoscopy in a recent meta-analysis26 of eight studies28,29,30,31,32,33,34,35. However, national authorities, such as the National Institute for Health and Care Excellence in the UK, do not recommend its use because of insufficient evidence of a clinically relevant advance36. High-resolution images can also be obtained with DermaInspect (Jenlab, Germany), a multiphoton-tomography–fluorescence-lifetime-imaging device that measures time-resolved tissue autofluorescence37,38 to detect metabolic and structural molecules in tissue. In comparison to simple handheld devices conventionally used by dermatologists, the device is much larger and more expensive. Preliminary results for diagnosing malignant melanoma using DermaInspect are promising39, but further studies in larger patient cohorts are needed to evaluate whether DermaInspect can improve the sensitivity or specificity for the diagnosis of malignant melanoma.

Optical coherence tomography (OCT) captures high-resolution cross-sectional images of the skin by a method that is similar to ultrasound40. OCT enables the imaging of skin layers deeper than 1 mm (that is, deeper than dermoscopy or confocal techniques). Pilot studies have demonstrated the potential of OCT for the detection of malignant melanoma41,42. Two devices are currently in clinical use with similar technical specifications43: Skintell (Agfa Healthcare, Belgium and Germany) and Vivosight (Michelson Diagnostics, UK). Since cross-sectional OCT images are similar to histological images, and en-face OCT images are similar to confocal-microscopy images, OCT is able to use feature descriptors from pathologists, including architectural patterns and cytological features of pigmented cells in the epidermis, dermo–epidermal junction and dermis, providing a starting point for feature-based algorithms for disease classification and biomarker development41. The widespread adoption of OCT in ophthalmology44,45 may contribute to the successful application of OCT in the diagnosis of malignant melanoma, but additional clinical trials are needed to evaluate whether there is sufficient added value in depth-resolved OCT images.

Spectroscopy

In spectroscopy methods, the tissue is illuminated, and changes in wavelength or intensity that result from light absorption, light scattering or fluorescence interactions of light and tissue46,47 are measured. These methods differentiate malignant from benign tissue on the basis of the properties of the bulk chemical constituents of tissue—such as water, lipids, proteins, RNA, DNA, haemoglobin and melanin—or of specific structural (in particular, collagen) or metabolic (such as NADH) molecules46,47. Spectroscopy-based modalities are typically limited to a narrow field of view or to even mapping with a single point, yet acquire data at high spectral resolution (often measuring the intensity at hundreds or thousands of different wavelengths). The high dimensionality of spectral data prohibits interpretation by the human eye, so this data is usually analysed via multivariate statistical techniques or, increasingly, by machine-learning approaches to reach a binary biomarker classification, with histopathology providing the ground truth for training these algorithms48. Perhaps owing to the complexity of the data, most spectroscopy methods remain firmly in the preclinical phase and have not achieved regulatory approval49,50,51,52,53,54,55,56,57,58,59.

Raman spectroscopy is one optical-spectroscopy modality that has received regulatory approval. It probes the primary chemical constituents of tissue via inelastic scattering. The acquired spectra are classified using principal component analysis and discriminant analysis, both based on a training set of known spectra for which the classifications are known a priori. Several large clinical studies have demonstrated the potential of Raman spectroscopy in skin imaging. One of these studies55 has resulted in a commercial clinical system (Aura; Verisante, Canada)60 that has received regulatory approval. Nonetheless, despite over two decades of research, Raman spectroscopy is not routinely used in the clinic for the evaluation of malignant melanoma. The slow uptake might be due to challenges in obtaining an accurate ground truth for training the classifiers61,62, and to challenges of quality assurance and control in spectral calibration and spectra interpretation48. Recent developments have significantly shortened the signal acquisition times (<1s in endoscopy), which may solve some of these problems by allowing faster and more robust signal acquisition from tissues that are inherently heterogeneous, allowing for better delineation of the rich spectral features enabled by this imaging modality.

Considerations for technology adoption in primary care

One key to improving the outcomes of malignant melanoma will be earlier disease detection. To achieve this, detection methods have to minimize false-negative rates and be easily implementable while maintaining a high negative predictive value. Afterwards, cost-effectiveness can be considered via an evaluation of the entire impact of the technology with regards to complete patient-management.

Another matter is whether primary-care physicians are willing to use new imaging methods. Although there is little doubt as to whether microscopic tools such as RCM provide better information than macroscopic imaging systems (and then naked-eye examination), the added time required to image entire macroscopic fields of view, or to choose which microscopic fields to image, is a challenge for the adoption of imaging technology in primary care. Methods for automated mosaicking with microscopy hold substantial promise to improve this, but the logistical problem of mapping out a macroscopic field of view with a microscopic tool is one that may need to be solved in order to facilitate wide adoption of the technology.

Early detection of dysplasia in Barrett’s oesophagus

Barrett’s oesophagus (BO) is an acquired condition in which columnar epithelium replaces the stratified squamous epithelium of the distal oesophagus. Patients with BO have an elevated risk of developing oesophageal adenocarcinoma63,64, which increases in the presence of dysplasia65,66. As a result, patients with BO are often recommended to undergo routine surveillance with high-definition white light endoscopy (HD-WLE) on a three-to-five year basis67,68,69,70,71. Targeted and random biopsies are taken during endoscopy, and analysed by a pathologist for the identification of cellular changes associated with dysplasia.

Although several studies have shown that a surveillance regime detects cancers at an earlier stage and increases survival72,73,74,75,76, its sensitivity remains as low as 40% (ref. 77), thus resulting in a high rate of missed dysplastic lesions78. The main reasons for this low sensitivity are that dysplasia can be difficult to spot on HD-WLE (ref. 68) and that taking random biopsies is time consuming, costly and prone to sampling errors68. New developments in optical imaging to address these limitations fall into two main categories: advances in wide-field imaging aim to provide better visualisation of dysplasia and to provide a ‘red flag’ for the endoscopist to target the biopsy; and advances in narrow-field imaging provide an ‘optical biopsy’ of the suspicious areas that could ultimately reduce or replace tissue biopsy. The combination of a high sensitivity ‘red flag’ approach with a high specificity ‘optical biopsy’ approach may prove to be a useful strategy in endoscopic imaging for BO (and for other cancers in the gastrointestinal tract). According to the guidelines of the American Society for Gastrointestinal Endoscopy (ASGE), a new technology needs to have 90% sensitivity, 80% specificity and 98% negative predictive value to be recommended for targeted biopsy79. To date, only three technologies have met these thresholds: narrow band imaging (NBI), acetic acid chromoendoscopy and endoscope-based confocal laser endomicroscopy (eCLE).

Wide-field imaging

In chromoendoscopy, dyes are applied to enhance the contrast in tissues so as to improve the detection of early lesions. The most common dyes are methylene blue80,81 (a vital stain that is absorbed by the tissue) and acetic acid78,82,83 (a weak acid that causes a transient whitening of the tissue and accentuates the mucosal pit pattern). The optical biomarker in these cases is the presence or absence of a pattern of dye staining, which is normally agreed by expert consensus. Although chromoendoscopy helps to visualise the tissue structure without the need for specialized equipment, application of the dye requires additional tissue preparation, adding time and costs to the procedure that are not widely reimbursed5. Chromoendoscopy using acetic acid has been clinically recommended to increase the yield of targeted biopsy5, but phototoxicity of methylene blue remains a concern84,85,86. Numerous research groups are developing targeted optical molecular-imaging dyes that enable the visualisation of complex biochemical processes involved in disease87,88. These dyes bind disease-specific cell-surface receptors or detect other disease-specific molecular changes, revealing particular pathologies as regions of altered fluorescence. As with traditional chromoendoscopy, the use of these dyes would increase the time and cost of an endoscopy procedure; furthermore, the cost of synthesising new targeted dyes at clinical grade, and of performing clinical trials to prove efficacy, can be prohibitive. Thus, despite promising results89,90,91,92, a convincing demonstration of a significant improvement in clinical performance is still required to justify the added complexity of using optical molecular imaging in routine endoscopic surveillance.

To avoid the need for dyes, virtual (or electronic) chromoendoscopy (VC) can be employed to enhance the contrast by illuminating the tissue via restricted bandwidths of light, as in narrow-band imaging (NBI, Olympus) and blue-laser imaging (BLI, Fuji)93,94. Alternatively, post-processing software can be used to enhance the images, as in intelligent colour enhancement (FICE, Fuij)95 and iSCAN (Pentax)96. Post-processing software has been shown to be equivalent or superior to HD-WLE or NBI in several gastrointestinal indications97,98,99,100,101,102,103,104,105,106, with reasonable inter-observer agreement107,108. VC requires only minor hardware or software modifications to the standard endoscopy equipment and does not require any dye that could create toxicology concerns. Hence, VC has received regulatory approval and is widely available. NBI using a feature-based classification109 has also met the ASGE threshold for clinical recommendation in targeting biopsy5.

Exploiting a different contrast mechanism, autofluorescence imaging (AFI) relies on the intrinsic fluorescence emission of structural (such as collagen) and metabolic (such as NADH) molecules in tissue110 to detect dysplasia as dark purple patches of reduced AFI signal111. Although AFI initially showed promising results, its low specificity111 prevented it from replacing WLE in the clinic. Instead, AFI has been combined with HD-WLE and VC in standard commercial forward-facing endoscopes to create trimodal-imaging systems112,113,114.

Endomicroscopy and spectroscopy

Once suspicious areas have been identified via wide-field imaging, a tissue biopsy is taken for histopathological analysis. Given the time and cost required for pathological assessments, several high-resolution endoscopy (or endomicroscopy) modalities have been developed with the goal of replacing tissue biopsies with in situ optical biopsies. In these approaches, high-resolution images of the epithelium are captured and interpreted using a set of consensus-feature-based criteria to classify pathological states115,116. Usually, an exogenous contrast agent, such as fluorescein, is added to enhance the contrast.

Owing to the need for confocal laser excitation and image collection optics, confocal laser endomicroscopy (CLE) is performed over a narrow field of view by introducing a probe into the accessory channel of a standard endoscope (probe-based, pCLE) or by replacing the standard endoscope with a standalone endomicroscope (eCLE). pCLE has the advantage of being compatible to the standard endoscopy procedure and has received regulatory approval (Cellvizio, Mauna Kea). eCLE (FIVE 1, Optiscan; ISC 1000, Pentax) has also received regulatory approval and meets the standards for use in targeting biopsy, but is no longer commercially available.

Volumetric optical coherence tomography (vOCT; also known as volumetric laser endomicroscopy, VLE) is an alternative endomicroscopy technique able to capture high-resolution depth-resolved cross-sectional images of the entire distal oesophagus117,118,119 through a balloon-based system (NVisionVLE, Ninepoint Medical)120,121,122,123,124. As with pCLE, the balloon is introduced via the working channel of a standard endoscope. vOCT images have been shown to correlate with histological images125,126,127 and can be interpreted by a trained image interpreter using a set of feature-based criteria to perform disease classification128. However, the dissimilarity between cross-sectional OCT images and standard HD-WLE images limits their co-registration and makes it challenging for endoscopists to re-locate suspicious areas identified on OCT for WLE-guided intervention; cautery marking is currently being investigated to solve this problem129. A further consequence of the lack of familiarity of the cross-sectional OCT images is the need for advanced training of expert image interpreters, although it is possible for these experts to identify dysplasia130 with good inter-observer agreement128. Another barrier that limits widespread translation is that manual image analysis takes several hours. In this regard, automated segmentation algorithms may help131. Automated algorithms may also be combined with the emerging tethered capsule OCT endoscopy systems118 that can be swallowed un-sedated under the supervision of a nurse in primary care132. As cross-sectional information is particularly important in the assessment of submucosal invasions when staging oesophageal adenocarcinoma, further developments to improve imaging at depth, for example, by using photoacoustic imaging133, are still needed.

In addition to endomicroscopy imaging, Raman spectroscopy has been applied endoscopically to obtain biochemical information of the tissue so as to distinguish dysplasia from surrounding Barrett’s tissue on the basis of least-squares discriminant analysis of the Raman spectra. But low reproducibility, expensive equipment and the need for long integration times, have limited clinical translation. However, recent improvements in integration times134,135 and in classification algorithms to provide the endoscopist with auditory feedback of a diagnosis in real time (0.2 seconds) led to in vivo trials134. Larger-scale in vivo trials and calibration methods will be needed to improve reproducibility and to ensure that promising ex vivo results apply in vivo61.

Considerations for technology adoption in specialist care

In primary care the consequences of false negatives are particularly acute, since no further follow up will be provided to the patient. In specialist care, however, the number of patients is lower than in primary care, and the trade-off between acceptable levels of false-positive and false-negative results depends on the next step of disease management. Ideally, both should be minimized. As such, the cutting edge of new technology is often found in specialist care, where novel techniques could add value to the decision-making process. Defining the value added of a given technology requires rigorous testing for diagnostic accuracy, acceptability, effectiveness and cost-effectiveness in the intended setting, by using case-controlled, randomized and double-blinded clinical trials to assess the technology’s real value to the healthcare system. If the diagnosis is more accurate or the amount of time a specialist is required is reduced, then the technology may add value. Regulatory approvals for devices, particularly with predicates, often lack clinical evidence from such trials136, so it is common for approved technologies to fail to achieve widespread adoption, because they do not change clinical practice. Recent changes137 to the European Union medical device legislative framework may mitigate this challenge in the future. Thus, the development pipeline of optical-imaging devices and their associated OIBs continues well beyond regulatory approval. Most commonly, consensus studies from specialist societies are required to reach the recommendation for use in routine patient management.

Translational characteristics of successful OIBs

The two case studies discussed reveal a wide range of promising optical methods that have been established in a research setting and that could potentially improve early cancer detection. However, few of these methods have successfully received regulatory approval, and even fewer have progressed to recommendation in healthcare systems.

Cost-effectiveness is a key consideration for adoption into healthcare systems, and it applies to one-off costs (such as the purchase of instrumentation, and the training of operators and interpreters) and to ongoing costs (such as the employment of operators, image interpreters and support staff, the purchase of disposable materials, cleaning or sterilization, and annual maintenance). In private healthcare systems, the development of new reimbursement codes can promote translation, as has been the case with OCT in the context of opthalmology138. Ultimately, acceptable costs would be determined on the basis of the overall performance of the OIB and whether it increases quality-adjusted life years (QALYs) or it reduces other healthcare costs. Examples of healthcare costs that could be alleviated with successful translation of a new OIB for early cancer detection include those costs associated with treating more advanced disease (approximately a further £10,000 per patient for late-stage colorectal cancer with respect to patients with the early stage of the disease139) or those associated with exhaustive physical biopsies (for example, targeting biopsies in BO surveillance can reduce per-patient biopsy costs in the United Kingdom from approximately £1,000 to £30; ref. 78; such costs depend considerably on country). Assessments of OIB cost-effectiveness should therefore be made in light of the intended diagnostic indication and of the potential change in the current clinical patient pathway that could be achieved.

Having achieved cost-effectiveness, OIBs still face a wide range of hurdles in clinical translation. In reviewing the two case studies, we identified several translational characteristics (detailed in the following sections) that are common among widely adopted OIBs. Most of the identified characteristics fall into three main categories: device and methodology, image acquisition and handling, and image interpretation.

Contrast mechanism underlying the OIB

Exploiting endogenous contrast (imaging without application of dyes) to detect the OIB is favourable for a clinical application, although exogenous contrast agents can improve the contrast of cancer tissue compared to healthy tissue, both as non-specific stains140 and as targeted molecular-imaging agents87. However, contrast agents require synthesis with good manufacturing practice (GMP) standards, toxicology studies, specific instrumentation and additional procedure time and cost77,141. Several aspects of contrast-agent chemistry increase the likelihood of their successful clinical translation: having a validated target (structural or molecular) increases confidence in the reproducibility of the results; administering agents topically rather than intravenously limits agent exposure to the tissue of interest and speeds up procedures; and using formulations of agents with long-term stability is favourable.

Instrumentation for OIB measurement

If a new device is required to measure the OIB, it must be approved by regulatory bodies such as the United States Food and Drug Administration136,142,143,144. For clinical implementation, compact, robust and transportable optical-imaging devices are highly desirable and more likely to succeed in clinical translation, while complex or delicate optics are less amenable to clinical implementation. Furthermore, devices that are compatible with existing systems, or that include current standard-of-care methods for reference, are more likely to reach deployment in healthcare systems, as they do not require a complete overhaul of existing equipment and facilities. The compatibility with current systems also facilitates head-to-head trials and allows a new modality to be introduced to the clinic as an adjunct to an existing one. Compatibility with the existing clinical workflow, and the avoidance of major changes to procedure times or costs, are additional advantages.

Device-operator expertise

The potential operators of optical instruments for measuring OIBs may be lay people, professionals working in primary care or highly specialized individuals working in a specialist care centre. If a new approach gives similar clinical results to an existing approach, yet requires less expertise, the likelihood of clinical translation is higher. Furthermore, if the need for training is sufficiently reduced, the approach may be translated from an expert to a generalist setting, reducing the burden on specialist care centres and the cost of running a high-volume imaging suite118. Conversely, if a high level of specialist knowledge is required to measure an OIB compared to the existing standard-of-care, the OIB is unlikely to be widely adopted, unless evidence can be provided that it contributes to improved clinical outcomes. Excessive complexity in the name of additional device performance may deter clinical adoption145. Clear standard operating procedures should be determined and adhered to if they are to improve ‘ease of use’ for operators and the increase likelihood of translation.

Some devices may be clinically translated because they reassure patients and clinicians of improved outcomes, whether or not they do actually improve the outcome. One example is the MoleMate skin-imaging system that, despite not being more accurate than best-practice for detecting melanoma, reassured patients and clinicians that they were receiving or providing thorough care18.

OIB interpreter expertise

Converting the raw-image data acquired using the optical modality into clinically relevant OIB information involves establishing image-interpretation criteria needed to deliver high sensitivity, specificity, inter-interpreter agreement and short training times. Criteria can include a binary decision, a library-based classification, the presence of specific image patterns109,146, or a change in signal intensity relative to a defined threshold. Several caveats can complicate this process. For example, increased expertise does not necessarily imply better inter-interpreter agreement147,148, and when applied to videos149 the criteria often result in lower performance than originally reported in still-image interpretation150,151,152. Establishing these OIBs is a time-consuming task, often requiring international consensus across multiple centres after regulatory approval of a device. Determining their clinical performance then requires randomized controlled trials in the appropriate settings, which are expensive and difficult to implement in a diagnostic setting for screening and surveillance programmes. For more complex biomarkers, expert image interpreters need to be trained, which further adds to the cost and time for adoption and makes the biomarker difficult to standardize across centres. Simplifying the output of the biomarker or transforming it to familiar images reduces the need for retraining and may enable non-experts to make diagnoses, as well as potentially smoothing the translational pathway.

Automated analysis that provides clear feedback to the interpreter of the biomarker can minimize the need for expert image interpreters. For many optical-imaging approaches, data reduction is essential because the dimensionality of the data is beyond interpretation by the human image interpreter. It has the potential to be objective, standardized and cheaper than employing human expertise, but has yet to mature to a stage where it is fully capable of operating in real time with sufficient performance to replace the human image interpreter153,154. If the technology is to be adopted in primary care, it will be important to use the device in a way that provides expert-level diagnostics with a high negative predictive value.

Repeatability and reproducibility of OIB measurements

Both repeatability and reproducibility across devices, operators and image interpreters must be assessed to evaluate the achievable precision for measurement of the OIB. Although inter-analyst agreement (encompassing operators and image interpreters) is often assessed, intra-analyst, intra-device and inter-device variability are often overlooked, making the comparative evaluation of OIBs challenging. To maximize the opportunity for translation to the intended setting, studies should be designed to enable the comparison of results obtained across multiple centres.

Co-registration of OIB information

The spatially resolved information obtained via imaging is often useful during a later surgical procedure for guiding tumour resection. However, optical-imaging modalities do not always provide sufficient anatomical information to guide an intervention. Strategies to overcome this (with increasing levels of complexity that could hinder translation) are co-registration with an existing modality that is compatible with surgical treatment155, application of laser cautery marks to highlight target areas129, and projection of the image data onto the patient or into the surgeon’s field of view via augmented-reality technology156,157.

The aforementioned wide-field endoscopy techniques produce 2D images that are easily integrated into standard equipment for HD-WLE, facilitating the widespread uptake of OIBs in endoscopy (as compared to the requirement for new equipment in many of the dermoscopy-based approaches). Acetic acid chromoendoscopy and NBI were recently reported to meet the thresholds required for recommendation5, and this milestone was achieved in part because of their deployment at many sites internationally, enabling the extensive development of image-classification criteria and publication of consensus statements109,150,151,152,158,159.

A roadmap for OIBs

We have formulated an OIB roadmap (Fig. 2), adapted from the international consensus ‘Imaging Biomarker Roadmap’ created for use in cancer studies3. Domain 1, ‘Discovery’, may be driven by a ‘technology push’, where an existing imaging technology is applied to an unmet clinical need, or by a ‘clinical pull’, where the OIB and/or technology are developed in response to a clearly defined, unmet clinical need160.

Fig. 2: Roadmap for OIBs.
figure2

The proposed roadmap differs from the radiological-imaging-biomarker roadmap3 by considering the intrinsic coupling of OIBs to the development of the optical-imaging modality. Red feedback loops decelerate clinical translation. Purple circles represent points at which careful consideration of translational characteristics (1–6) can accelerate translation. Technical and biological validations occur in parallel, and are closely linked because of the interplay between device, operator, interpreter and OIB. The technical validation occurs throughout the roadmap to ensure availability and precision in all settings. Cost effectiveness is omitted here yet it impacts the roadmap at every stage, owing to the equipment and personnel costs of performing imaging studies3. Translational characteristics: (1) contrast mechanism, (2) instrumentation, (3) operator expertise, (4) image-interpreter expertise, (5) repeatability and reproducibility, and (6) co-registration. SOP, standard operating procedure.

In all domains, optical-device development plays a larger role in translation than in standard radiological imaging, where a device to measure the imaging biomarker is usually already clinically approved. The OIB roadmap further emphasizes the concurrency of technical and biological validation. Technical validation seeks to define the precision and accuracy with which a given OIB can be measured, whereas biological and clinical validation seek to connect the OIB with underlying pathological processes and define clinical-performance characteristics, such as negative predictive value or specificity. Throughout the OIB roadmap, there is a complex interplay of limitations imposed by the device, the contrast mechanism and the OIB definition, meaning that technical and biological validation cannot be considered in isolation. For example, the precision (repeatability and reproducibility) of measuring a given OIB defined with respect to a perfect test-target will differ considerably from the precision defined with respect to a biological measurement made in a patient, which is directly relevant in the clinical application.

The OIB roadmap emphasizes the feedback loops that often inhibit translation, and highlights the key translational characteristics that can be exploited to accelerate translation. Also, our case studies revealed five key translational barriers that need to be addressed to facilitate a smoother path to clinical translation. We discuss each of them in the following subsections.

Lack of defined safe exposure limits for diagnostics

Safe and effective illumination levels that allow accurate measurement of a given biomarker must be established before in vivo trials take place. Visible and near-infrared light is not ionizing, but can lead to thermal damage, photosensitization and photoallergic reactions under certain conditions161. Molecularly damaging radiation extends further towards the visible spectrum162. The published guidelines for optical-radiation exposure limits only concern the retina and skin, and are limited in their application to workers rather than patients, which presents a challenge for those using incoherent light and internal diagnostic-imaging devices161,163. Rather, safety for internal exposure is defined in terms of thermal and photobiological consequences, with the assumption that the gastrointestinal tract and skin respond similarly. There is an unmet need for experimentally validated exposure limits for tissues beyond the retina and skin, so that safe patient exposures for optical imaging can be clearly defined. This would also address frequently raised concerns by ethical review boards of clinical trials. Access to advisors of optical-radiation protection with training in clinical optical imaging would increase the understanding of the capabilities, strengths and weaknesses of optical-imaging systems and their biomarkers.

Lack of standardized quality assurance

The uncertainty regarding optical-exposure limits is illustrative of a more general lack of standardization in optical imaging. Defining the performance of a device at the point of manufacture and throughout lifetime is crucial for comparing different devices (by measuring the same source of optical contrast) and for debugging device-specific problems.

Whilst performance standards (documents that suggest relevant performance characteristics and test methods) exist for the established non-optical-imaging methods, such as the NEMA (National Electrical Manufacturers Association) imaging standards for positron emission tomography (PET)164, optical imaging currently lacks published standards142. As a result, optical-imaging contrast agents and devices must often be approved in tandem. This does not occur with PET, for which contrast agents are usually approved for use in a range of devices (and PET devices can also be approved for use with a range of contrast agents), because sufficient evidence supporting similarities across devices exists142. Unfortunately, the consequence for optical imaging is that healthcare providers must purchase specific devices when working with specific optical-imaging agents, resulting in prohibitive costs and hindering the translation of new contrast agents.

In the United States and Europe, medical devices may be approved based on similarity to a predicate device, gaining 510(k) clearance in the United States and CE marking in the European Union136. Yet without standardization, the similarity to a predicate device may be difficult to define. With imaging standards, device performance could be compared to an appropriate standard, allowing inferiority to be spotted at an earlier stage in the translational roadmap. Additionally, if device-performance standards were established it would help separate device approval from agent approval in cases where exogenous agents are being imaged.

Calibration standards exist for all standard diagnostic imaging tests165. Stable luminescent phantoms, calibrated to the International System of Units (ISU) unit of radiance, with an effective shelf-life of more than two years, have now been developed to assess some optical diagnostic-imaging tests based on reflectance and fluorescence imaging166. Future work in other modalities, such as OCT and photoacoustic imaging, should aim to create similar standards for assessing intra-device and inter-device variability, thus helping to achieve improved repeatability and reproducibility for multicentre use.

Lack of validated representative ex vivo models

When tissue is excised from the body, several OIBs change irreversibly. For example, the lack of active blood flow changes the spectrum of tissue by reducing blood oxygenation to 0%, which consequently alters the haemoglobin absorption spectrum. Furthermore, tissue autofluorescence can be modified on exposure to ambient light, and the tissue structure may be distorted by surgical trauma or by positioning the tissue on a rigid surface. Ultimately, tissues will degrade unless they are frozen or fixed, which further alters tissue properties61. The gap between data acquired with ex vivo imaging and data acquired with in vivo imaging is therefore large, and data acquired from ex vivo tissues may contain insurmountable artefacts if the tissues are not properly handled.

Challenges with using ex vivo tissue for validation raise the need for new model systems, which may arise through improved tissue-mimicking phantoms, or from bioengineered tissues. Whichever route is chosen, to avoid disappointing in vivo results later in the roadmap, it is crucial to ensure that the ex vivo model is as representative as possible of the in vivo situation. For example, modalities that are sensitive to blood oxygenation should never use excised tissue for validation unless it is perfused, modalities sensitive to preparation artefacts should avoid chemically fixed or frozen tissue61, and dyes that incorporate labelled human antibodies should use models that try to replicate human biology, such as human-derived tumour xenografts, rather than animal tumour models where the binding affinity may be different.

Lack of accurate and validated clinical gold-standards

The clinical gold-standards used as comparators for evaluating the diagnostic accuracy, sensitivity and specificity of new OIBs must be well validated. Most commonly, gold-standard diagnosis is determined via the assessment of stained tissue sections by a pathologist. To achieve the most accurate gold standard for validation of OIBs, consensus of several independent pathologists is needed, as they are not always in agreement167. In the best-case scenario, only unanimous decisions would be accepted, although this is difficult to achieve in practice as it significantly decreases the number of samples that can be incorporated into any performance analysis. In practice, decision algorithms are used to combine contrasting histologic diagnoses of several pathologists to reach a single final diagnosis24. In future, machine-learning algorithms trained on huge datasets may be able to provide an objective diagnosis, but this is still some way from being realized. Another confounding factor is the transition from a biopsy read to in vivo imaging, which requires an appreciation of nuances associated with the fact that a sample of tissue is typically a mix of pathological and healthy tissue. Each optical-imaging pixel could include contributions from one or more tissue states. Validating imaging tools in vivo thus requires high congruence in spatial alignment between the in vivo and ex vivo coordinates, which is not always achieved in basic ‘single read’ pathology studies. Once the gold standard is validated, careful thought must be applied to how these findings will be accurately co-registered with optical images, so as not to introduce further artefacts into the comparison between the novel technique and the gold standard. The different scales at which in vivo optical-image data is recorded and histopathological analysis is performed can make this particularly challenging168.

Difficulties in conducting representative single-centre trials

To maximize opportunities for validation in multicentre trials, early-stage single-centre trials should replicate the common clinical environment as much as possible. First, representative populations should be chosen to reduce spectrum effects169. For example, due to the ethical considerations of taking biopsies from healthy tissue, many skin-imaging trials have been carried out in enriched populations with high disease incidence, which has prompted regulatory bodies to disregard such results and investigators to endorse the need for a mix of lesions representative of the target population for the future testing of new approaches24. Second, standard operating procedures should be determined by a single centre and adhered to in multicentre trials in order to prevent bias. For example, inspecting images for minutes, when the eventual standard operating procedure in a clinical setting would require inspection of videos in real time, results in misleading performance evaluations149,150,151,152. If image data were collected according to an appropriate standard operating procedure, collation of data into an online repository would assist in the validation process for a new OIB. Third, the expertise of the operator and image interpreter should be representative of the expertise realistically available in routine care. Endoscopic trimodal imaging provides a cautionary example: promising results in a tertiary referral centre114 were not reproduced in a community practice setting112, which could have been due to a combination of spectrum effects, different standard operating procedures or different expertise.

Outlook

Many of the OIBs reviewed here are still in development, yet several are already impacting patient care. Owing to their relatively low-cost implementation, lack of ionizing radiation and potential for real-time analysis across a range of scales, OIBs bring new opportunities for changing healthcare practice in the two case studies discussed in this Perspective, as well as in many other areas170,171.

Traditionally, the most advanced optical-imaging methods, and almost all radiological-imaging methods, have been confined to an expert setting due to the expense and size of these devices and the need for on-site expertise. The emergence of inexpensive and compact optical-detector technology and communications networks capable of broadcasting images means that image acquisition and interpretation no longer need to be co-localized. The development of cheap, robust and simple-to-use devices makes OIB data acquisition potentially feasible in primary care, for example with tethered capsule endoscopes or handheld skin-imaging devices17,130. In addition, the feasibility of image acquisition being performed in one centre, with experts interpreting images remotely, has already been demonstrated in endomicroscopy172. OIBs could, in future, be implemented in the home. The rapidly improving specifications of sensors within smartphones, the growing computational power of such devices, and their widespread access to mobile communication networks, have already led to applications for the monitoring of suspicious moles, but so far without proven accuracy or applicable value173. Data captured from smartphones could ultimately be interpreted remotely via expert-image interpreters or by using automated algorithms trained on the vast database of images resulting from such widespread acquisition. Examples of how these changes might impact the case studies discussed here are shown in Fig. 3.

Fig. 3: Current and potential future clinical implementations of optical imaging in two case studies.
figure3

The current clinical pipeline (black arrows) from the home through primary care to specialist care is shown for patients with suspected Barrett’s oesophagus (green) and suspicious malignant melanoma (pink). The potential future pipelines (red lines) will be mediated by new or modified indications for imaging (red boxes) and high-speed communication networks (red dashed lines). Risk stratification in primary care reduces the number of false positives referred for urgent specialist follow-up. Boxes highlight where imaging takes place in the pipeline.

We expect that this kind of flexible application of OIBs will rapidly become more common in research. Remote image acquisition and interpretation in a range of indications and healthcare settings, from telepathology174 to teledermatology175, and from the home to specialist care, are likely to become increasingly widespread, enabling better patient triage and decreasing the costs associated with high-volume specialist-care imaging suites and expert operators. Remote interpretation will allow for greater division of expertise, reducing training costs and increasing the average expertise available for image analysis. For some indications, automated diagnoses based on centralized image databases might circumvent the need for an image interpreter entirely.

There are a number of challenges for advancing the clinical implementation of OIBs. Regulatory approvals for new optical-imaging devices tend to favour commercial vendors rather than clinical value136, and often procedures are approved with a new device where the application is not improving care but rather just changing the sub-specialty that implements the procedure. Commercially, devices may be approved on the basis of similarity to a predicate device, without proof of clinical utility136, although the regulatory landscape is currently changing in this regard. Without optical-imaging standardization, similarity to a predicate may be difficult to define. Worse still is the prospect that the predicate itself may pre-date current regulations. Since devices can be marketed without proof of clinical utility, manufacturers may lack the motivation to invest the significant time, money and expertise required to conduct trials for assessing clinical utility, despite the potential long-term benefits.

Distributing devices into the hands of the patients themselves raises questions about the appropriate nature of clinical trials for testing, and brings additional challenges in standardization. Although the individual optical hardware or software may itself be inexpensive, the final clinical test must fully represent the cost of development, meaning that the cost-per-test can in some cases be similar to or greater than the costs of the existing standard of care.

The advent of new liquid biomarkers has raised questions about the future utility of imaging in the context of early cancer detection. Imaging currently provides crucial insight for the early detection of malignant melanoma and oesophageal cancer, and endoscopy methods are also broadly applied for the identification of cancers in the aerodigestive tract, as well as in the cervix and bladder. Although liquid biomarkers have shown great promise, even a study176 that achieved good performance for the detection of late-stage cancers (73% sensitivity) reported much poorer performance for the detection of early-stage disease (40% median sensitivity in the eight cancers tested, and only 20% sensitivity in oesophageal cancer). Moreover, while liquid biomarkers have the potential to accurately detect and even localize early-stage disease, imaging is likely to remain an important tool in screening and surveillance programmes for early cancer detection for many years to come. Also, imaging is the only test capable of providing the detailed spatially resolved information that is ultimately required for the resection of early cancers, and has shown great promise in this regard177,178. It is therefore likely that liquid biomarkers could be implemented as tools complementary to imaging, perhaps as a low-cost early-stage option to pre-screen patients prior to imaging, should a sufficiently high specificity be achieved.

Advanced optical-imaging research requires the technical development of optical devices as well as the technical and biological validation of the resulting OIBs. In many cases, the discovery of promising OIBs using optical devices dedicated to ex vivo imaging can motivate the development of devices able to measure the same biomarker in vivo. An understanding of the strengths and weaknesses of these optical devices is needed to assess the feasibility of measuring certain biomarkers in a clinical scenario, with a clear appreciation for the biological heterogeneity of the biomarker. For example, it may be the case that a multimodal approach is needed to achieve the required performance through the combination of two or more OIBs (ref. 179). These aspects demand a multidisciplinary research effort that combines clinical expertise of an unmet need with physics and engineering knowledge of the optical interactions in order to specifically tailor devices for the detection of known biomarkers. The sharing of intellectual-property ownership, or of appropriate licencing agreements, may also be required in order to combine imaging modalities. Furthermore, collaboration with computer scientists may help to better the design of interpretation algorithms to yield useful clinical biomarkers. Embracing this multidisciplinary approach is necessary to promote the translation of OIBs.

Imaging is currently the only medical tool capable of providing detailed, immediate and spatially resolved biological information in vivo, which is especially relevant for diagnosis of the earliest stages of cancer. Despite these promises, few OIBs have been translated to routine clinical use. The translational characteristics and barriers discussed here complement the OIB roadmap, which we hope will improve the chances of new optical-imaging approaches achieving widespread clinical implementation. The vast array of complex tissue–light interactions and the equally diverse arsenal of optical devices for detecting these interactions should help to improve the current standard-of-care for cancer patients.

References

  1. 1.

    Wax, A. et al. Angle-resolved low coherence interferometry for detection of dysplasia in Barrett’s esophagus. Gastroenterology 141, 443–447 (2011).

    PubMed  PubMed Central  Google Scholar 

  2. 2.

    Imamoto, Y. & Shichida, Y. Cone visual pigments. Biochim. Biophys. Acta 1837, 664–673 (2014).

    CAS  PubMed  Google Scholar 

  3. 3.

    O’Connor, J. B. P. et al. Imaging biomarker roadmap for cancer studies. Nat. Rev. Clin. Oncol. 14, 169–186 (2017).

    PubMed  Google Scholar 

  4. 4.

    Sharma, P. et al. White Paper AGA: Advanced Imaging in Barrett’s Esophagus. Clin. Gastroenterol. Hepatol. 13, 2209–2218.

    PubMed  Google Scholar 

  5. 5.

    Thosani, N. et al. ASGE Technology Committee systematic review and meta-analysis assessing the ASGE Preservation and Incorporation of Valuable Endoscopic Innovations thresholds for adopting real-time imaging–assisted endoscopic targeted biopsy during endoscopic surveillance. Gastrointest. Endosc. 83, 684–698.

  6. 6.

    Nikolaou, V. & Stratigos, A. J. Emerging trends in the epidemiology of melanoma. Br. J. Dermatol. 170, 11–19 (2014).

    CAS  PubMed  Google Scholar 

  7. 7.

    Suspected Cancer: Recognition and Referral: NICE Guideline (NG12) (NICE, 2015); https://www.nice.org.uk/guidance/ng12

  8. 8.

    Terushkin, V. & Halpern, A. C. Melanoma Early Detection. Hematol. Oncol. Clin. North Am. 23, 481–500 (2009).

    PubMed  Google Scholar 

  9. 9.

    Lindholm, C. et al. Invasive cutaneous malignant melanoma in Sweden, 1990–1999: A prospective, population-based study of survival and prognostic factors. Cancer 101, 2067–2078 (2004).

    PubMed  Google Scholar 

  10. 10.

    Breitbart, E. W. et al. Systematic skin cancer screening in Northern Germany. J. Am. Acad. Dermatol. 66, 201–211 (2012).

    PubMed  Google Scholar 

  11. 11.

    Katalinic, A. et al. Does skin cancer screening save lives?: an observational study comparing trends in melanoma mortality in regions with and without screening. Cancer 118, 5395–5402 (2012).

    PubMed  Google Scholar 

  12. 12.

    Bibbins-Domingo, K. et al. Screening for Skin Cancer. Jama 316, 429–435 (2016).

    PubMed  Google Scholar 

  13. 13.

    Stratigos, A. J. et al. Euromelanoma: a dermatology-led European campaign against nonmelanoma skin cancer and cutaneous melanoma. Past, present and future. Br. J. Dermatol. 167, 99–104 (2012).

    PubMed  Google Scholar 

  14. 14.

    Vestergaard, M. E. et al. Dermoscopy compared with naked eye examination for the diagnosis of primary melanoma: A meta-analysis of studies performed in a clinical setting. Br. J. Dermatol. 159, 669–676 (2008).

    CAS  PubMed  Google Scholar 

  15. 15.

    Moncrieff, M. et al. Spectrophotometric intracutaneous analysis: a new technique for imaging pigmented skin lesions. Br. J. Dermatol. 146, 448–457 (2002).

    CAS  PubMed  Google Scholar 

  16. 16.

    Emery, J. D. et al. Accuracy of SIAscopy for pigmented skin lesions encountered in primary care: development and validation of a new diagnostic algorithm. BMC Dermatol. 10, 9 (2010).

    PubMed  PubMed Central  Google Scholar 

  17. 17.

    Walter, F. M. et al. Protocol for the MoleMate UK Trial: a randomised controlled trial of the MoleMate system in the management of pigmented skin lesions in primary care. BMC Fam. Pract. 11, 36 (2010).

    PubMed  PubMed Central  Google Scholar 

  18. 18.

    Walter, F. M. et al. Effect of adding a diagnostic aid to best practice to manage suspicious pigmented lesions in primary care: randomised controlled trial. Br. Med. J. 345, e4110 (2012).

    Google Scholar 

  19. 19.

    Vasefi, F. et al. Multimode optical dermoscopy (SkinSpect) analysis for skin with melanocytic nevus. Proc. of SPIE 9711, 971110 (2016).

    Google Scholar 

  20. 20.

    Vasefi., F. et al. Polarization-sensitive hyperspectral imaging in vivo: a multimode dermoscope for skin analysis. Sci. Rep. 4, 4924 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  21. 21.

    Vasefi, F. et al. Separating melanin from hemodynamics in nevi using multimode hyperspectral dermoscopy and spatial frequency domain spectroscopy. J. Biomed. Opt. 21, 114001 (2016).

    PubMed  PubMed Central  Google Scholar 

  22. 22.

    MacKinnon, N. et al. In vivo skin chromophore mapping using a multimode imaging dermoscope (SkinSpect). Proc. SPIE 8587, (1–13 (2013).

    Google Scholar 

  23. 23.

    Elbaum, M. et al. Automatic differentiation of melanoma from melanocytic nevi with multispectral digital dermoscopy: a feasibility study. J. Am. Acad. Dermatol. 44, 207–218 (2001).

    CAS  PubMed  Google Scholar 

  24. 24.

    Monheit, G. et al. The performance of MelaFind: a prospective multicenter study. Arch. Dermatol. 147, 188–194 (2011).

    PubMed  Google Scholar 

  25. 25.

    Hauschild, A. et al. To excise or not: impact of MelaFind on German dermatologists’ decisions to biopsy atypical lesions. J. Dtsch Dermatol. Ges. 12, 606–614 (2014).

    PubMed  Google Scholar 

  26. 26.

    Xiong, Y.-Q. et al. Comparison of dermoscopy and reflectance confocal microscopy for the diagnosis of malignant skin tumours: a meta-analysis. J. Cancer Res. Clin. Oncol. Springer 1627–1635 (2017).

  27. 27.

    Segura, S. et al. Development of a two-step method for the diagnosis of melanoma by reflectance confocal microscopy. J. Am. Acad. Dermatol. 61, 216–229 (2009).

    PubMed  Google Scholar 

  28. 28.

    Guitera, P. et al. Surveillance for treatment failure of lentigo maligna with dermoscopy and in vivo confocal microscopy: new descriptors. Br. J. Dermatol. 170, 1305–1312 (2014).

    CAS  PubMed  Google Scholar 

  29. 29.

    Guitera, P. et al. In vivo reflectance confocal microscopy enhances secondary evaluation of melanocytic lesions. J. Invest. Dermatol. 129, 131–138 (2009).

    CAS  PubMed  Google Scholar 

  30. 30.

    Stanganelli, I. et al. Integration of reflectance confocal microscopy in sequential dermoscopy follow-up improves melanoma detection accuracy. Br. J. Dermatol. 172, 365–371 (2015).

    CAS  PubMed  Google Scholar 

  31. 31.

    Witkowski, A. M. et al. Non-invasive diagnosis of pink basal cell carcinoma: how much can we rely on dermoscopy and reflectance confocal microscopy? Ski. Res. Technol. 22, 230–237 (2016).

    CAS  Google Scholar 

  32. 32.

    Venturini, M. et al. Reflectance confocal microscopy allows in vivo real-time noninvasive assessment of the outcome of methyl aminolaevulinate photodynamic therapy of basal cell carcinoma. Br. J. Dermatol. 168, 99–105 (2013).

    CAS  PubMed  Google Scholar 

  33. 33.

    Langley, R. G. B. et al. The diagnostic accuracy of in vivo confocal scanning laser microscopy compared to dermoscopy of benign and malignant melanocytic lesions: a prospective study. Dermatology 215, 365–372 (2007).

    PubMed  Google Scholar 

  34. 34.

    Moscarella, E. et al. The role of reflectance confocal microscopy as an aid in the diagnosis of collision tumors. Dermatology 227, 109–117 (2013).

    PubMed  Google Scholar 

  35. 35.

    Alarcon, I. et al. Impact of in vivo reflectance confocal microscopy on the number needed to treat melanoma in doubtful lesions. Br. J. Dermatol. 170, 802–808 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  36. 36.

    VivaScope 1500 and 3000 imaging systems for detecting skin cancer lesions - guidance (DG19) (NICE, 2015).

  37. 37.

    Konig, K. & Riemann, I. High-resolution multiphoton tomography of human skin with subcellular spatial resolution and picosecond time resolution. J. Biomed. Opt. 8, 432–439 (2003).

    PubMed  Google Scholar 

  38. 38.

    Manfredini, M. et al. High-resolution imaging of basal cell carcinoma: a comparison between multiphoton microscopy with fluorescence lifetime imaging and reflectance confocal microscopy. Ski. Res. Technol. 19, 433–443 (2013).

    Google Scholar 

  39. 39.

    Seidenari, S. et al. Multiphoton laser tomography and fluorescence lifetime imaging of melanoma: morphologic features and quantitative data for sensitive and specific non-invasive diagnostics,”. PLoS ONE 8, e70682 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  40. 40.

    Ulrich, M. et al. Dynamic optical coherence tomography in dermatology. Dermatology 232, 298–311 (2016).

    PubMed  Google Scholar 

  41. 41.

    Boone, M. A. L. M. et al. High-definition optical coherence tomography imaging of melanocytic lesions: a pilot study. Arch. Dermatol. Res. 306, 11–26 (2014).

    PubMed  Google Scholar 

  42. 42.

    Gambichler, T. et al. High-definition optical coherence tomography of melanocytic skin lesions. J. Biophotonics 8, 681–686 (2015).

    PubMed  Google Scholar 

  43. 43.

    Schuh, S. et al. Comparison of different optical coherence tomography devices for diagnosis of non-melanoma skin cancer. Ski. Res. Technol. 22, 395–405 (2016).

    CAS  Google Scholar 

  44. 44.

    Nolan, R. C. et al. Optical coherence tomography for the neurologist. Semin. Neurol. 35, 564–577 (2015).

    PubMed  Google Scholar 

  45. 45.

    Tan, A. C. S. et al. An overview of the clinical applications of optical coherence tomography angiography. Eye 32, 262–268 (2017).

    PubMed  PubMed Central  Google Scholar 

  46. 46.

    Liu, Q. Role of optical spectroscopy using endogenous contrasts in clinical cancer diagnosis. World J. Clin. Oncol. 2, 50–63 (2011).

    PubMed  PubMed Central  Google Scholar 

  47. 47.

    Kendall, C. et al. Vibrational spectroscopy: a clinical tool for cancer diagnostics. Analyst 134, 1029–1045 (2009).

    CAS  PubMed  Google Scholar 

  48. 48.

    Pence, I. & Mahadevan-Jansen, A. Clinical instrumentation and applications of Raman spectroscopy. Chem. Soc. Rev. 45, 1958–1979 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  49. 49.

    Brancaleon, L. et al. In vivo fluorescence spectroscopy of nonmelanoma skin cancer. Photochem. Photobiol. 73, 178–183 (2001).

    CAS  PubMed  Google Scholar 

  50. 50.

    Garcia-Uribe, A. et al. Skin cancer detection by spectroscopic oblique-incidence reflectometry: classification and physiological origins. Appl. Opt. 43, 2643–2650 (2004).

    PubMed  Google Scholar 

  51. 51.

    Rajaram, N. et al. Pilot clinical study for quantitative spectral diagnosis of non-melanoma skin cancer. Lasers Surg. Med. 42, 716–727 (2010).

    PubMed  PubMed Central  Google Scholar 

  52. 52.

    Thompson, A. J. et al. In vivo measurements of diffuse reflectance and time-resolved autofluorescence emission spectra of basal cell carcinomas. J. Biophotonics 5, 240–254 (2012).

    PubMed  Google Scholar 

  53. 53.

    Garcia-Uribe, A. et al. In vivo diagnosis of melanoma and nonmelanoma skin cancer using oblique incidence diffuse reflectance spectrometry. Cancer Res. 72, 2738–2745 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  54. 54.

    Upile, T. et al. Elastic scattering spectroscopy in assessing skin lesions: an ‘in vivo’ study. Photodiagnosis Photodyn. Ther. 9, 132–141 (2012).

    PubMed  Google Scholar 

  55. 55.

    Lui, H. et al. Real-time raman spectroscopy for in vivo skin cancer diagnosis. Cancer Res. 72, 2491–2500 (2012).

    CAS  PubMed  Google Scholar 

  56. 56.

    Zhao, J. et al. Real-time Raman spectroscopy for non-invasive skin cancer detection - preliminary results. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2008, 3107–3109 (2008).

    PubMed  Google Scholar 

  57. 57.

    Lim, L. et al. Clinical study of noninvasive in vivo melanoma and nonmelanoma skin cancers using multimodal spectral diagnosis. J. Biomed. Opt. 19, 117003 (2014).

    PubMed  PubMed Central  Google Scholar 

  58. 58.

    Schleusener, J. et al. In vivo study for the discrimination of cancerous and normal skin using fibre probe-based Raman spectroscopy. Exp. Dermatol. 24, 767–772 (2015).

    PubMed  Google Scholar 

  59. 59.

    Zakharov, V. P. et al. Combined Raman spectroscopy and autofluoresence imaging method for in vivo skin tumor diagnosis. Proc. SPIE 9198, (919804 (2014).

    Google Scholar 

  60. 60.

    Stamnes, J. J. et al. Optical detection and monitoring of pigmented skin lesions. Biomed. Opt. Express 8, 2946–2964 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  61. 61.

    Shim, M. G. & B. C. Wilson, B. C. The effects of ex vivo handling procedures on the near-infrared Raman spectra of normal mammalian tissues. Photochem. Photobiol. 63, 662–671 (1996).

    CAS  PubMed  Google Scholar 

  62. 62.

    Kim, S., Byun, K. M. & Lee, S. Y. Influence of water content on Raman spectroscopy characterization of skin sample. Biomed. Opt. Express 8, 1130–1138 (2017).

    PubMed  PubMed Central  Google Scholar 

  63. 63.

    Hvid-Jensen, F. et al. Incidence of adenocarcinoma among patients with Barrett’s esophagus. N. Engl. J. Med. 365, 1375–1383 (2011).

    CAS  PubMed  Google Scholar 

  64. 64.

    Gatenby, P. et al. Lifetime risk of esophageal adenocarcinoma in patients with Barrett’s esophagus. World J. Gastroenterol. 20, 9611–9617 (2014).

    PubMed  PubMed Central  Google Scholar 

  65. 65.

    Duits, L. C. et al. Barrett’s oesophagus patients with low-grade dysplasia can be accurately risk-stratified after histological review by an expert pathology panel. Gut 64, 700–706 (2015).

    PubMed  Google Scholar 

  66. 66.

    Singh, S. et al. Incidence of esophageal adenocarcinoma in Barrett’s esophagus with low-grade dysplasia: a systematic review and meta-analysis. Gastrointest. Endosc. 79, 897–909 (2014).

    PubMed  Google Scholar 

  67. 67.

    Fitzgerald, R. C. et al. British Society of Gastroenterology guidelines on the diagnosis and management of Barrett’s oesophagus. Gut 63, 7–42 (2014).

    PubMed  Google Scholar 

  68. 68.

    Spechler, S. J. et al. American Gastroenterological Association technical review on the management of Barrett’s esophagus. Gastroenterology 140, 18–52 (2011).

    Google Scholar 

  69. 69.

    Wang., K. K. & Sampliner, R. E. Updated guidelines 2008 for the diagnosis, surveillance and therapy of Barrett’s esophagus. Am. J. Gastroenterol. 103, 788–797 (2008).

    PubMed  Google Scholar 

  70. 70.

    Evans, J. A. et al. The role of endoscopy in Barrett’s esophagus and other premalignant conditions of the esophagus. Gastrointest. Endosc. 76, 1087–1094 (2012).

    PubMed  Google Scholar 

  71. 71.

    Weusten, B. et al. Endoscopic management of Barrett’s esophagus: European Society of Gastrointestinal Endoscopy (ESGE) position statement. Endoscopy 49, 191–198 (2017).

    PubMed  Google Scholar 

  72. 72.

    Rubenstein, J. H. et al. Effect of a prior endoscopy on outcomes of esophageal adenocarcinoma among United States veterans. Gastrointest. Endosc. 68, 849–855 (2008).

    PubMed  PubMed Central  Google Scholar 

  73. 73.

    Cooper, G. S., Kou, T. D. & Chak, A. Receipt of previous diagnoses and endoscopy and outcome from esophageal adenocarcinoma: a population-based study with temporal trends. Am. J. Gastroenterol. 104, 1356–1362 (2009).

    PubMed  Google Scholar 

  74. 74.

    El-Serag, H. B. et al. Surveillance endoscopy is associated with improved outcomes of oesophageal adenocarcinoma detected in patients with Barrett’s oesophagus. Gut 65, 1252–1260 (2016).

    PubMed  Google Scholar 

  75. 75.

    Kastelein, F. et al. Impact of surveillance for Barrett’s oesophagus on tumour stage and survival of patients with neoplastic progression. Gut 65, 1–7 (2015).

    Google Scholar 

  76. 76.

    Verbeek, R. E. et al. Surveillance of Barrett’s Esophagus and mortality from esophageal adenocarcinoma: a population-based cohort study. Am. J. Gastroenterol. 109, 1215–1222 (2014).

    PubMed  Google Scholar 

  77. 77.

    Sturm, M. B. & Wang, T. D. Emerging optical methods for surveillance of Barrett’s oesophagus. Gut 64, 1816–1823 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  78. 78.

    Chedgy, F. J. Q. et al. Acetic acid chromoendoscopy: Improving neoplasia detection in Barrett’s esophagus. World J. Gastroenterol. 22, 5753–5760 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  79. 79.

    Sharma, P. et al. The American Society for Gastrointestinal Endoscopy PIVI (preservation and incorporation of valuable endoscopic innovations) on imaging in Barrett’s Esophagus. Gastrointest. Endosc. 76, 252–254 (2012).

    PubMed  Google Scholar 

  80. 80.

    Lim, C. H. et al. Randomized crossover study that used methylene blue or random 4-quadrant biopsy for the diagnosis of dysplasia in Barrett’s esophagus. Gastrointest. Endosc. 64, 195–199 (2006).

    PubMed  Google Scholar 

  81. 81.

    Horwhat, J. D. et al. A randomized comparison of methylene blue-directed biopsy versus conventional four-quadrant biopsy for the detection of intestinal metaplasia and dysplasia in patients with long-segment Barrett’s esophagus. Am. J. Gastroenterol. 103, 546–554 (2008).

    PubMed  Google Scholar 

  82. 82.

    Coletta, M. et al. Acetic acid chromoendoscopy for the diagnosis of early neoplasia and specialized intestinal metaplasia in Barrett’s esophagus: a meta-analysis. Gastrointest. Endosc. 83, 57–67 (2016).

    PubMed  Google Scholar 

  83. 83.

    Tholoor, S. et al. Acetic acid chromoendoscopy in Barrett’s esophagus surveillance is superior to the standardized random biopsy protocol: results from a large cohort study. Gastrointest. Endosc. 80, 417–424 (2014).

    PubMed  Google Scholar 

  84. 84.

    Olliver, J. R. et al. Chromoendoscopy with methylene blue and associated DNA damage in Barrett’s oesophagus. Lancet 362, 373–374 (2003).

    CAS  PubMed  Google Scholar 

  85. 85.

    Sturmey, R. G., Wild, C. P. & Hardie, L. J. Removal of red light minimizes methylene blue-stimulated DNA damage in oesophageal cells: implications for chromoendoscopy. Mutagenesis 24, 253–258 (2009).

    CAS  PubMed  Google Scholar 

  86. 86.

    Kovacic, P. & Somanathan, R. Toxicity of imine-iminium dyes and pigments: electron transfer, radicals, oxidative stress and other physiological effects. J. Appl. Toxicol. 34, 825–834 (2014).

    CAS  PubMed  Google Scholar 

  87. 87.

    James, M. L. & Gambhir, S. S. A molecular imaging primer: modalities, imaging agents, and applications. Physiol. Rev. 92, 897–965 (2012).

    CAS  PubMed  Google Scholar 

  88. 88.

    Lee, J. H. & Wang, T. D. Molecular endoscopy for targeted imaging in the digestive tract. Lancet Gastroenterol. Hepatol. 1, 147–155 (2016).

    PubMed  Google Scholar 

  89. 89.

    Sturm, M. B. et al. Targeted imaging of esophageal neoplasia with a fluorescently labeled peptide: first-in-human results. Sci. Transl. Med. 5, 184ra61 (2013).

    PubMed  Google Scholar 

  90. 90.

    Joshi, B. P. et al. Multimodal endoscope can quantify wide-field fluorescence detection of Barrett’s neoplasia. Endoscopy 48, 1–13 (2015).

    CAS  Google Scholar 

  91. 91.

    Bird-Lieberman, E. L. et al. Molecular imaging using fluorescent lectins permits rapid endoscopic identification of dysplasia in Barrett’s esophagus. Nat. Med. 18, 315–321 (2012).

    CAS  PubMed  Google Scholar 

  92. 92.

    Nagengast, W. B. et al. Near-infrared fluorescence molecular endoscopy detects dysplastic oesophageal lesions using topical and systemic tracer of vascular endothelial growth factor A. Gut 68, 7–10 (2017).

    PubMed  Google Scholar 

  93. 93.

    Kaneko, K. et al. Effect of novel bright image enhanced endoscopy using blue laser imaging (BLI). Endosc. Int. Open 2, E212–E219 (2014).

    PubMed  PubMed Central  Google Scholar 

  94. 94.

    Yoshida, N. et al. Ability of a novel blue laser imaging system for the diagnosis of colorectal polyps. Dig. Endosc. 26, 250–258 (2014).

    PubMed  Google Scholar 

  95. 95.

    Miyake, Y. et al. Development of new electronic endoscopes using the spectral images of an internal organ. In Proc. IS&T/SID’s Thirteen Color Imaging Conf. 261–269 (Society for Imaging Science and Technology, 2005).

  96. 96.

    Kodashima, S. & Fujishiro, M. Novel image-enhanced endoscopy with i-scan technology. World J. Gastroenterol. 16, 1043–1049 (2010).

    PubMed  PubMed Central  Google Scholar 

  97. 97.

    Li, C. Q. et al. Magnified and enhanced computed virtual chromoendoscopy in gastric neoplasia: a feasibility study. World J. Gastroenterol. 19, 4221–4227 (2013).

    PubMed  PubMed Central  Google Scholar 

  98. 98.

    Osawa, H. et al. Diagnosis of depressed-type early gastric cancer using small-caliber endoscopy with flexible spectral imaging color enhancement. Dig. Endosc. 24, 231–236 (2012).

    PubMed  Google Scholar 

  99. 99.

    Osawa, H. et al. Diagnosis of endoscopic Barrett’s esophagus by transnasal flexible spectral imaging color enhancement. J. Gastroenterol. 44, 1125–1132 (2009).

    PubMed  Google Scholar 

  100. 100.

    Aminalai, A. et al. Live image processing does not increase adenoma detection rate during colonoscopy: a randomized comparison between FICE and conventional imaging (Berlin Colonoscopy Project 5, BECOP-5). Am. J. Gastroenterol. 105, 2383–2388 (2010).

    PubMed  Google Scholar 

  101. 101.

    Chung, S. J. et al. Efficacy of computed virtual chromoendoscopy on colorectal cancer screening: a prospective, randomized, back-to-back trial of Fuji Intelligent Color Enhancement versus conventional colonoscopy to compare adenoma miss rates. Gastrointest. Endosc. 72, 136–142 (2010).

    PubMed  Google Scholar 

  102. 102.

    Basford, P. J. et al. High-definition endoscopy with i-Scan for evaluation of small colon polyps: the HiSCOPE study. Gastrointest. Endosc. 79, 111–118 (2014).

    PubMed  Google Scholar 

  103. 103.

    Hong, S. N. et al. Prospective, randomized, back-to-back trial evaluating the usefulness of i-SCAN in screening colonoscopy. Gastrointest. Endosc. 75, 1011–1021 (2012).

    PubMed  Google Scholar 

  104. 104.

    Lee, C. K., Lee, S. H. & Hwangbo, Y. Narrow-band imaging versus I-Scan for the real-time histological prediction of diminutive colonic polyps: A prospective comparative study by using the simple unified endoscopic classification. Gastrointest. Endosc. 74, 603–609 (2011).

    PubMed  Google Scholar 

  105. 105.

    Yoshida, Y. et al. A randomized crossover open trial of the adenoma miss rate for narrow band imaging (NBI) versus flexible spectral imaging color enhancement (FICE). Int. J. Colorectal Dis. 28, 1511–1516 (2013).

    PubMed  Google Scholar 

  106. 106.

    Chung, S. J. et al. Comparison of detection and miss rates of narrow band imaging, flexible spectral imaging chromoendoscopy and white light at screening colonoscopy: a randomised controlled back-to-back study. Gut 63, 785–791 (2014).

    PubMed  Google Scholar 

  107. 107.

    Masci, E. et al. Interobserver agreement among endoscopists on evaluation of polypoid colorectal lesions visualized with the Pentax i-Scan technique. Dig. Liver Dis. 45, 207–210 (2013).

    PubMed  Google Scholar 

  108. 108.

    Pigò, F. et al. I-Scan high-definition white light endoscopy and colorectal polyps: prediction of histology, interobserver and intraobserver agreement. Int. J. Colorectal Dis. 28, 399–406 (2013).

    PubMed  Google Scholar 

  109. 109.

    Sharma, P. et al. Development and validation of a classification system to identify high-grade dysplasia and esophageal adenocarcinoma in Barrett’s Esophagus using narrow band imaging. Gastroenterology 150, 591–598 (2016).

    PubMed  Google Scholar 

  110. 110.

    Haringsma, J. & Tytgat, G. N. J. Fluorescence and autofluorescence. Clin. Gastroenterol. 13, 1–10 (1999).

    CAS  Google Scholar 

  111. 111.

    Kara, M. A. et al. Endoscopic video autofluorescence imaging may improve the detection of early neoplasia in patients with Barrett’s esophagus. Gastrointest. Endosc. 61, 679–685 (2005).

    PubMed  Google Scholar 

  112. 112.

    Curvers, W. L. et al. Endoscopic trimodal imaging versus standard video endoscopy for detection of early Barrett’s neoplasia: a multicenter, randomized, crossover study in general practice. Gastrointest. Endosc. 73, 195–203 (2011).

    PubMed  Google Scholar 

  113. 113.

    Boerwinkel, D. F. et al. Endoscopic TriModal imaging and biomarkers for neoplasia conjoined: a feasibility study in Barrett’s esophagus. Dis. Esophagus 27, 435–443 (2014).

    CAS  PubMed  Google Scholar 

  114. 114.

    Curvers, W. L. et al. Endoscopic tri-modal imaging is more effective than standard endoscopy in identifying early-stage neoplasia in Barrett’s esophagus. Gastroenterology 139, 1106–1114 (2010).

    PubMed  Google Scholar 

  115. 115.

    Wallace, M. et al. Miami classification for probe-based confocal laser endomicroscopy. Endoscopy 43, 882–891 (2011).

    CAS  PubMed  Google Scholar 

  116. 116.

    Kiesslich, R. et al. In vivo histology of Barrett’s esophagus and associated neoplasia by confocal laser endomicroscopy. Clin. Gastroenterol. Hepatol. 4, 979–987 (2006).

    PubMed  Google Scholar 

  117. 117.

    Vakoc, B. J. et al. Comprehensive esophageal microscopy by using optical frequency-domain imaging. Gastrointest. Endosc. 65, 898–905 (2007).

    PubMed  PubMed Central  Google Scholar 

  118. 118.

    Gora, M. J. et al. Tethered capsule endomicroscopy enables less invasive imaging of gastrointestinal tract microstructure. Nat. Med. 19, 238–240 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  119. 119.

    Yun, S. H. et al. Comprehensive volumetric optical microscopy in vivo. Nat. Med. 12, 1429–1433 (2006).

    CAS  PubMed  PubMed Central  Google Scholar 

  120. 120.

    Wolfsen, H. C. et al. Safety and feasibility of volumetric laser endomicroscopy in patients with Barrett’s esophagus. Gastrointest. Endosc. 82, 631–640 (2015).

    PubMed  Google Scholar 

  121. 121.

    Swager, A. F. et al. Detection of buried Barrett’s glands after radiofrequency ablation with volumetric laser endomicroscopy. Gastrointest. Endosc. 83, 80–88 (2016).

    PubMed  Google Scholar 

  122. 122.

    Swager, A. et al. Volumetric laser endomicroscopy in Barrett’s esophagus: a feasibility study on histological correlation. Dis. Esophagus 29, 505–512 (2015).

    PubMed  Google Scholar 

  123. 123.

    Trindade, A. J. et al. Volumetric laser endomicroscopy can target neoplasia not detected by conventional endoscopic measures in long segment Barrett’s esophagus. Endosc. Int. 4, 318–322 (2016).

    Google Scholar 

  124. 124.

    Leggett, C. L. et al. Comparative diagnostic performance of volumetric laser endomicroscopy and confocal laser endomicroscopy in the detection of dysplasia associated with Barrett’s esophagus. Gastrointest. Endosc. 83, 880–888 (2015).

    PubMed  PubMed Central  Google Scholar 

  125. 125.

    Poneros, J. M. et al. Diagnosis of specialized intestinal metaplasia by optical coherence tomography. Gastroenterology 120, 7–12 (2001).

    CAS  PubMed  Google Scholar 

  126. 126.

    Evans, J. A. et al. Optical coherence tomography to identify intramucosal carcinoma and high-grade dysplasia in Barrett’s esophagus. Clin. Gastroenterol. Hepatol. 4, 38–43 (2006).

    PubMed  PubMed Central  Google Scholar 

  127. 127.

    Evans, J. A. et al. Identifying intestinal metaplasia at the squamocolumnar junction by using optical coherence tomography. Gastrointest. Endosc. 65, 50–56 (2007).

    PubMed  Google Scholar 

  128. 128.

    Sauk, J. et al. Interobserver agreement for the detection of barrett’s esophagus with optical frequency domain imaging. Dig. Dis. Sci. 58, 2261–2265 (2013).

    PubMed  PubMed Central  Google Scholar 

  129. 129.

    Suter, M. J. et al. Esophageal-guided biopsy with volumetric laser endomicroscopy and laser cautery marking: a pilot clinical study. Gastrointest. Endosc. 79, 886–896 (2014).

    PubMed  PubMed Central  Google Scholar 

  130. 130.

    Gora, M. J. et al. Imaging the upper gastrointestinal tract in unsedated patients using tethered capsule endomicroscopy. Gastroenterology 145, 723–725 (2013).

    PubMed  Google Scholar 

  131. 131.

    Ughi, G. J. et al. Automated segmentation and characterization of esophageal wall in vivo by tethered capsule optical coherence tomography endomicroscopy. Biomed. Opt. Express 7, 660–665 (2016).

    Google Scholar 

  132. 132.

    Gora, M. et al. Tethered capsule endomicroscopy : from bench to bedside at a primary care practice tethered capsule endomicroscopy : from bench to bedside at a primary care practice. J. Biomed. Opt. 21, 104001 (2016).

    PubMed  PubMed Central  Google Scholar 

  133. 133.

    Ntziachristos, V. ESOTRAC https://www.esotrac2020.eu/ (2018).

  134. 134.

    Bergholt, M. S. et al. Fiberoptic confocal raman spectroscopy for real-time in vivo diagnosis of dysplasia in Barrett’s esophagus. Gastroenterology 146, 27–32 (2014).

    PubMed  Google Scholar 

  135. 135.

    Almond, L. M. et al. Endoscopic Raman spectroscopy enables objective diagnosis of dysplasia in Barrett’s esophagus. Gastrointest. Endosc. 79, 37–45 (2014).

    PubMed  Google Scholar 

  136. 136.

    Van Norman, G. A. Drugs and devices: comparison of European and U. S. approval processes. JACC Basic Transl. Sci. 1, 399–412 (2016).

    PubMed  PubMed Central  Google Scholar 

  137. 137.

    Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017. Official J. European Union L117, 1–175, https://eur-lex.europa.eu/legal-content/ENG/TXT/PDF/?uri=CELEX:32017R0745 (2017).

  138. 138.

    Beswick, D. M., Kaushik, A. & Beinart, D. Biomedical device innovation methodology: applications in biophotonics. J. Biomed. Opt. 23, 1–7 (2017).

    PubMed  Google Scholar 

  139. 139.

    Laudicella, M. et al. Cost of care for cancer patients in England: evidence from population-based patient-level data. Br. J. Cancer 114, 1286–1292 (2016).

    PubMed  PubMed Central  Google Scholar 

  140. 140.

    Trivedi, P. J. & Braden, B. Indications, stains and techniques in chromoendoscopy. QJM 106, 117–131 (2013).

    CAS  PubMed  Google Scholar 

  141. 141.

    Sevick-Muraca, E. M. et al. Advancing the translation of optical imaging agents for clinical imaging. Biomed. Opt. Express 4, 160–170 (2013).

    PubMed  Google Scholar 

  142. 142.

    Tummers, W. S. et al. Regulatory aspects of optical methods and exogenous targets for cancer detection. Cancer Res. 77, 2197–2206 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  143. 143.

    US FDA. Overview of Device Regulation. US Department of Health and Human Services https://www.fda.gov/MedicalDevices/DeviceRegulationandGuidance/Overview/default.htm (2017).

  144. 144.

    European Commission. Medical devices. European Union https://ec.europa.eu/growth/sectors/medical-devices_en (2017).

  145. 145.

    Marcu, L. et al. Biophotonics: the big picture. J. Biomed. Opt. 23, 1–7 (2017).

    PubMed  Google Scholar 

  146. 146.

    Boerwinkel, D. F. et al. Third-generation autofluorescence endoscopy for the detection of early neoplasia in Barrett’s esophagus: a pilot study. Dis. Esophagus 27, 276–284 (2014).

    CAS  PubMed  Google Scholar 

  147. 147.

    Alvarez Herrero, L. et al. Zooming in on Barrett oesophagus using narrow-band imaging: an international observer agreement study. Eur. J. Gastroenterol. Hepatol. 21, 1068–1075 (2009).

    PubMed  Google Scholar 

  148. 148.

    Curvers, W. L. et al. Mucosal morphology in Barrett’s esophagus: interobserver agreement and role of narrow band imaging. Endoscopy 40, 799–805 (2008).

    CAS  PubMed  Google Scholar 

  149. 149.

    Silva, F. B. et al. Endoscopic assessment and grading of Barrett’s esophagus using magnification endoscopy and narrow-band imaging: accuracy and interobserver agreement of different classification systems. Gastrointest. Endosc. 73, 7–14 (2011).

    PubMed  Google Scholar 

  150. 150.

    Kara, M. A. et al. Detection and classification of the mucosal and vascular patterns (mucosal morphology) in Barrett’s esophagus by using narrow band imaging. Gastrointest. Endosc. 64, 155–166 (2006).

    PubMed  Google Scholar 

  151. 151.

    Singh, R. et al. Narrow-band imaging with magnification in Barrett’s esophagus: validation of a simplified grading system of mucosal morphology patterns against histology. Endoscopy 40, 457–463 (2008).

    CAS  PubMed  Google Scholar 

  152. 152.

    Sharma, P. et al. The utility of a novel narrow band imaging endoscopy system in patients with Barrett’s esophagus. Gastrointest. Endosc. 64, 167–175 (2006).

    PubMed  Google Scholar 

  153. 153.

    de Bruijne, M. Machine learning approaches in medical image analysis: From detection to diagnosis. Med. Image Anal. 33, 94–97 (2016).

    PubMed  Google Scholar 

  154. 154.

    Suzuki, K. Overview of deep learning in medical imaging. Radiol. Phys. Technol. 10, 1–17 (2017).

    Google Scholar 

  155. 155.

    Garcia-Allende, P. B. et al. Towards clinically translatable NIR fluorescence molecular guidance for colonoscopy. Biomed. Opt. Express 5, 78–92 (2013).

    PubMed  PubMed Central  Google Scholar 

  156. 156.

    Pelargos, P. E. et al. Utilizing virtual and augmented reality for educational and clinical enhancements in neurosurgery. J. Clin. Neurosci. 35, 1–4 (2016).

    PubMed  Google Scholar 

  157. 157.

    Nicolau, S. et al. Augmented reality in laparoscopic surgical oncology. Surg. Oncol. 20, 189–201 (2011).

    PubMed  Google Scholar 

  158. 158.

    Kandiah, K. et al. OC-054 Development and validation of a classification system to identify Barrett’s neoplasia using acetic acid chromoendoscopy: the predict classification. Gut 65(31), 1–31 (2016).

    Google Scholar 

  159. 159.

    Longcroft-Wheaton, G. et al. Duration of acetowhitening as a novel objective tool for diagnosing high risk neoplasia in Barrett’s esophagus: a prospective cohort trial. Endoscopy 45, 426–432 (2013).

    CAS  PubMed  Google Scholar 

  160. 160.

    Wilson, B. C., Jermyn, M. & Leblond, F. Challenges and opportunities in clinical translation of biomedical optical spectroscopy and imaging. J. Biomed. Opt. 23, 1–13 (2018).

    PubMed  Google Scholar 

  161. 161.

    ICNIRP. ICNIRP guidelines on limits of exposure to incoherent visible and infrared radiation. Health Phys. 71, 804–819 (2013).

    Google Scholar 

  162. 162.

    Brookner, C. K. et al. Safety analysis: relative risks of ultraviolet exposure from fluorescence spectroscopy and colposcopy are comparable. Photochem. Photobiol. 65, 1020–1025 (1997).

    CAS  PubMed  Google Scholar 

  163. 163.

    EU. Directive 2006/25/EC of the European Parliament and of the Council of 5th of April 2006. Official J. European Union L114, 38–60 (2006); https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2006:114:0038:0059:en:PDF

  164. 164.

    Performance Measurements of Positron Emission Tomographs (PETs) (NEMA, 2013); https://www.nema.org/Standards/Pages/Performance-Measurements-of-Positron-Emission-Tomographs.aspx

  165. 165.

    Boellaard, R. et al. FDG PET/CT: EANM procedure guidelines for tumour imaging: version 2.0. Eur. J. Nucl. Med. Mol. Imaging 42, 328–354 (2014).

    PubMed  PubMed Central  Google Scholar 

  166. 166.

    Zhu, B. et al. Determining the performance of fluorescence molecular imaging devices using traceable working standards with SI units of radiance. IEEE Trans. Med. Imaging 0062, 1 (2015).

    Google Scholar 

  167. 167.

    Downs-Kelly, E. et al. Poor interobserver agreement in the distinction of high-grade dysplasia and adenocarcinoma in pretreatment Barrett’s esophagus biopsies. Am. J. Gastroenterol. 103, 2333–2340 (2008).

    PubMed  Google Scholar 

  168. 168.

    O’Connor, J. P. B. et al. Imaging intratumor heterogeneity: role in therapy response, resistance, and clinical outcome. Clin. Cancer Res. 21, 249–257 (2015).

    PubMed  Google Scholar 

  169. 169.

    Usher-Smith, J. A., Sharp, S. J. & Griffin, S. J. The spectrum effect in tests for risk prediction, screening, and diagnosis. BMJ 353, i3139 (2016).

    PubMed  PubMed Central  Google Scholar 

  170. 170.

    Calin, M. A. et al. Hyperspectral imaging in the medical field: present and future. Appl. Spectrosc. Rev. 49, 435–447 (2014).

    Google Scholar 

  171. 171.

    Grosenick, D. et al. Review of optical breast imaging and spectroscopy. J. Biomed. Opt. 21, 091311 (2016).

    PubMed  Google Scholar 

  172. 172.

    Fuks, D. et al. Intraoperative confocal laser endomicroscopy for real-time in vivo tissue characterization during surgical procedures. Surg. Endosc. https://doi.org/10.1007/s00464-018-6442-3 (2018).

    PubMed  Google Scholar 

  173. 173.

    Kassianos, A. P. et al. Smartphone applications for melanoma detection by community, patient and generalist clinician users: a review. Br. J. Dermatol. 172, 1507–1518 (2015).

    CAS  PubMed  Google Scholar 

  174. 174.

    Farahani, N. et al. International telepathology: promises and pitfalls. Pathobiology 83, 121–126 (2016).

    PubMed  Google Scholar 

  175. 175.

    Tensen, E. et al. Two decades of teledermatology: current status and integration in national healthcare systems. Curr. Dermatol. Rep. 5, 96–104 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  176. 176.

    Cohen, J. D. et al. Detection and localization of surgically resectable cancers with a multi-analyte blood test. Science 359, 926–930 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  177. 177.

    Chi, C. et al. Intraoperative imaging-guided cancer surgery: from current fluorescence molecular imaging methods to future multi-modality imaging technology. Theranostics 4, 1072–1084 (2014).

    PubMed  PubMed Central  Google Scholar 

  178. 178.

    De Boer, E. et al. Optical innovations in surgery. Br. J. Surg. 102, 56–72 (2015).

    Google Scholar 

  179. 179.

    Heinzmann, K. et al. Multiplexed imaging for diagnosis and therapy. Nat. Biomed. Eng. 1, 697–713 (2017).

    PubMed  Google Scholar 

  180. 180.

    Bhat, Y. M. et al. High-definition and high-magnification endoscopes. Gastrointest. Endosc. 80, 919–927 (2014).

    Google Scholar 

  181. 181.

    Medical Systems (Olympus Technologies, 2017); https://www.olympus-europa.com/medical/en/medical_systems/technologies/narrow_band_imaging__nbi_1/technologies_nbi.jsp

  182. 182.

    FICE Dual Mode (Fujifilm Europe, 2017); https://www.fujifilm.eu/eu/products/medical-systems/endoscopy/technology/fice-dual-mode.

  183. 183.

    i-Scan Imaging (Pentax Medical, 2017); https://www.pentaxmedical.com/pentax/en/95/1/i-scan-imaging

  184. 184.

    About SIMSYS-MoleMate (MedX Health, 2017); http://medxhealth.com/Our-Products/SIAscopytrade;/overview.aspx.

  185. 185.

    MelaFind http://www.melafind.com (accessed 26 April 2017).

  186. 186.

    Manfredi, M. A. et al. Electronic chromoendoscopy. Gastrointest. Endosc. 81, 249–261 (2015).

    PubMed  Google Scholar 

  187. 187.

    Rameshshanker, R. & Wilson, A. Electronic imaging in colonoscopy: clinical applications and future prospects. Curr. Treat. Options Gastroenterol. 14, 140–151 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  188. 188.

    Lu, G. & Fei, B. Medical hyperspectral imaging: a review. J. Biomed. Opt. 19, 10901 (2014).

    PubMed  Google Scholar 

  189. 189.

    Jang, J. The past, present, and future of image-enhanced endoscopy. Clin. Endosc. 48, 466–475 (2015).

    PubMed  PubMed Central  Google Scholar 

  190. 190.

    DERMA Medical Systems https://www.dermamedicalsystems.com/index.php?menu_id=117 (accessed 2 May 2017).

  191. 191.

    Alali, S. & Vitkin, A. Polarized light imaging in biomedicine: emerging Mueller matrix methodologies for bulk tissue assessment. J. Biomed. Opt. 20, 61104 (2015).

    PubMed  Google Scholar 

  192. 192.

    Cellvizio: Our Flagship Product (Mauna Kea Technologies, 2017); http://www.maunakeatech.com/en/hospital-administrators/cellvizio-solution

  193. 193.

    OptiScan ViewnVivo. http://viewnvivo.com/ (accessed 26 April 2017).

  194. 194.

    Clinical Applications: Noninvasive cellular imaging of the skin (Caliber ID, 2017); http://www.caliberid.com/clinical.html

  195. 195.

    Wong Kee Song, L. M. et al. Autofluorescence imaging. Gastrointest. Endosc. 73, 647–650 (2011).

    Google Scholar 

  196. 196.

    DermaInspect (Jenlab, 2017); http://www.jenlab.de/DermaInspect.29.0.html

  197. 197.

    Hanson, K. M. & Bardeen, C. J. Application of nonlinear optical microscopy for imaging skin. Photochem. Photobiol. 85, 33–44 (2009).

    CAS  PubMed  Google Scholar 

  198. 198.

    Fu, L. & Gu, M. Fibre-optic nonlinear optical microscopy and endoscopy. J. Microsc. 226, 195–206 (2007).

    CAS  PubMed  Google Scholar 

  199. 199.

    Thomas, G. et al. Advances and challenges in label-free nonlinear optical imaging using two-photon excitation fluorescence and second harmonic generation for cancer research. J. Photochem. Photobiol. B Biol. 141, 128–138 (2014).

    CAS  Google Scholar 

  200. 200.

    NvisionVLE Imaging System (NinePoint Medical Inc, 2017); http://www.ninepointmedical.com/nvisionvle-imaging-system/

  201. 201.

    Vivosight (Michelson Diagnostics, 2017); https://vivosight.com/about-us/product/

  202. 202.

    A. J. Trindade, A. J., Smith, M. S. & Pleskow, D. K. The new kid on the block for advanced imaging in Barrett’s esophagus: a review of volumetric laser endomicroscopy. Therap. Adv. Gastroenterol. 9, 408–416 (2016).

    Google Scholar 

  203. 203.

    Kallaway, C. et al. Advances in the clinical application of Raman spectroscopy for cancer diagnostics. Photodiagnosis Photodyn. Ther. 10, 207–219 (2013).

    CAS  PubMed  Google Scholar 

  204. 204.

    Wang, W. et al. Real-time in vivo cancer diagnosis using raman spectroscopy. J. Biophotonics 8, 527–545 (2015).

    CAS  PubMed  Google Scholar 

  205. 205.

    Tu, Q. & Chang, C. Diagnostic applications of Raman spectroscopy. Nanomed. Nanotechnol. Biol. Med. 8, 545–558 (2012).

    CAS  Google Scholar 

  206. 206.

    SpectraScience and PENTAX Medical Announce Expanded Distribution of WavSTAT4 Optical Biopsy System in Europe. PENTAX Medical (24 April 2013).

  207. 207.

    Benes, Z. & Antos, Z. Optical biopsy system distinguishing between hyperplastic and adenomatous polyps in the colon during colonoscopy. Anticancer Res. 29, 4737–4739 (2009).

    PubMed  Google Scholar 

  208. 208.

    Boerwinkel, D. F. et al. Fluorescence spectroscopy incorporated in an Optical Biopsy System for the detection of early neoplasia in Barrett’s esophagus. Dis. Esophagus 28, 345–351 (2014).

    PubMed  Google Scholar 

  209. 209.

    Yu, G. Near-infrared diffuse correlation spectroscopy in cancer diagnosis and therapy monitoring. J. Biomed. Opt. 17, 010901 (2012).

    PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We would like to thank F. Walter of the University of Cambridge for helpful comments on our manuscript. D.J.W., C.R.M.F. and S.E.B. are financially supported by CRUK (C14303/A17197, C47594/A16267, C47594/A21102, C55962/A24669) and EPSRC (C197/A16465, EP/N014588/1).

Author information

Affiliations

Authors

Contributions

D.J.W., J.P.B.O. and S.E.B. conceived the manuscript. D.J.W. researched and wrote the manuscript together with C.R.M.F. B.W.P. reviewed and edited the manuscript. All authors discussed and agreed with the final version of the manuscript.

Corresponding author

Correspondence to Sarah E. Bohndiek.

Ethics declarations

Competing interests

S.E.B. receives research support from iThera Medical GmbH and PreXion Inc., and chairs the International Photoacoustic Standardisation Consortium (IPASC).

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Waterhouse, D.J., Fitzpatrick, C.R.M., Pogue, B.W. et al. A roadmap for the clinical implementation of optical-imaging biomarkers. Nat Biomed Eng 3, 339–353 (2019). https://doi.org/10.1038/s41551-019-0392-5

Download citation

Further reading

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing