Endoscopic resection of adenomatous colorectal polyps has been demonstrated to reduce mortality from cancer of the colorectum, the second highest cause of cancer death in the United States (1,2). The global gold standard for the diagnosis of a colorectal adenoma is ex vivo histopathologic microscopy; however, 43% of resected polyps are identified as benign and up to 1% has already undergone malignant transformation (3). Given such a range of polyp phenotypes and the risk of malignancy that they each confer, clinical management pathways are currently individualized to the risk profile of each encountered polyp post hoc, most commonly through manipulation of surveillance intervals.
There is a need for accurate, real-time, endoscopic risk stratification of colorectal polyps, which could be achieved if histologic subtype can be predicted in vivo with low levels of error. This would allow real-time tailored management approaches based on the risk, such as the “resect and discard” strategy (low-risk adenomas can be resected and discarded without histology) and the “diagnose and leave” strategy (benign hyperplastic polyps of the recto-sigmoid can be left without polypectomy) (4). Such approaches could improve the quality and safety of patient care by avoiding the risks associated with unnecessary polypectomy and by shortening procedure times. Furthermore, cost efficiencies have been predicted to exceed $1 billion annually (5).
To achieve this goal, technological innovations have been progressively integrated into the endoscopy suite over the past two decades, with the manipulation or augmentation of optical outputs showing promise (6). Optical technologies such as narrow-band imaging (NBI) can now be found in a host of endoscopy suites worldwide; however, several questions remain regarding their diagnostic accuracies and how they can be deployed effectively and safely in different clinical scenarios. Previous reviews have been limited by predefining a small number of technologies, including post hoc predictions from images or videos, methodologic incorporation of heterogeneity, and presenting incomplete measures of diagnostic accuracy (4,6,7). Furthermore, the impact of recent implementation studies needs to be assessed while identifying how such technologies have developed over time. This review uses the novel approach of chronologic meta-analyses to track whether the current rates of improvement are sufficient to exceed diagnostic thresholds required for clinical implementation and redefining of management pathways.
The aim of this study is to compare the diagnostic accuracy of optical technologies in the real-time endoscopic histologic assessment of colorectal polyps.
A systematic review with meta-analysis was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy Studies (PRISMA-DTA) statement (8). No review protocol has been published in advance of this review.
Electronic database search, appraisal of review article bibliographies, and broad Internet search were conducted to identify relevant peer-reviewed studies in humans. Medline (1946 till present), EMBASE (1947 till present), and the Cochrane Database were searched on March 1, 2018, using a strategy including but not limited to the following terms and Boolean operators: ((exp Intestine, Large/OR colon* OR rect*) AND (dysplas* OR adenoma* OR polyp* OR lesion*) OR exp Colonic Polyps/) AND (exp Colonoscopy/OR Endoscop*) AND (Real adj time OR in adj vivo OR spectroscop* OR narrow adj band OR optical) AND (diagnos* OR detect* OR classif* OR histology*) (full search strategy listed in Appendix 1, Supplementary Digital Content, http://links.lww.com/AJG/A67). Bibliographies of relevant review articles were hand-searched, and a Google Scholar search was conducted to identify any additional articles of relevance.
Criteria for study inclusion
Two authors (S.E.M. and L.P.) independently reviewed the search results and applied inclusion criteria after manuscript review. Any disagreement between the authors was resolved by the senior author (J.M.K). For inclusion, studies must have prospectively compared the real-time diagnostic accuracy of a single endoscopically deployable optical technology with the gold standard histopathologic assessment of colorectal lesions, with respect to at least one of the following outcomes: presence of adenoma, differentiation of high- and low-grade dysplasia in conventional adenomas, identification of adenoma subtype, or presence of adenocarcinoma. Formal histopathologic validation using hemotoxylin and eosin staining with microscopy was required in all cases. It was expected that diagnosis was made in accordance with internationally recognized guidelines such as the Vienna criteria (9). The intervention must have been used in at least 10 adult patients in vivo, with studies using procedural photographs or videos for post hoc interpretation excluded. Patient cohorts suffering from polyposis syndromes or inflammatory bowel disease were excluded. No limitation was placed on the language or the date of publication. Studies were excluded if more than one optical technology was used, they did not provide data suitable for the meta-analysis (components of a 2 × 2 diagnostic accuracy contingency table, or data from which this can be calculated), data were not primary (review articles, editorials), or where there was no full peer-reviewed manuscript (conference proceedings).
Data extraction and statistical analysis
Two reviewers independently extracted raw data from manuscripts as absolute values of a diagnostic contingency table, with adenocarcinoma or adenoma being defined as “positive” and benign tissues considered “negative.” Where insufficient data were presented to determine these values (such as sensitivity and specificity without a disease prevalence), the authors were contacted directly to provide the necessary data. Additional data collected included the country, technical details of the technology used, whether endoscopists were sufficiently experienced (defined as having undergone a training course or having previously performed at least 50 cases with the specific technology), whether a separate analysis is presented for polyps less than 5 mm, and diagnostic accuracies for a subgroup of predictions made with high confidence. Quality scoring was conducted for each study included in a meta-analysis using the QUADAS-2 tool (10).
Where 3 or more studies had used a specific optical technology to determine identical clinical outcomes, these underwent Bayesian bivariate meta-analysis using Laplace approximation in RStudio (v1.1.442) using the meta4diag program (11). Diagnostic accuracy was evaluated using sensitivity, specificity, and the area under a summary receiver operator curve. Heterogeneity was assessed after visual interpretation of forest plots. Publication bias was assessed by creating funnel plots. Statistical significance was defined with α = 0.05, and 95% confidence intervals (CIs) were presented where appropriate. Subgroup analyses were predefined and based on factors of clinical relevance: experience of the endoscopist, endoscope specifications, the confidence of prediction, polyp size, and recto-sigmoid location of the polyps. Studies that were not suitable for the meta-analysis underwent narrative review. The historical and predicted future progress of technological development toward meeting the adenoma negative predictive value (NPV), clinical implementation threshold of over 90% was addressed by plotting aggregate meta-analyses chronologically by the year with logarithmic trend identification.
The database and hand-searching returned 4,084 studies, which after screening identified 284 manuscripts to be assessed for eligibility in the study (Figure 1). It was not possible to source 3 manuscripts through either our institutional access or the British Library, and as a result, these studies could not be included. One hundred two studies were deemed suitable for inclusion, consisting of 129 patient cohorts analyzing a total of 33,123 colorectal polyps.
Eighty-nine percent of studies were conducted in the health care systems of developed countries, with Japan (21%) and the United States (17%) being the 2 most popular locations. China was the most prevalent developing country, conducting 6% of included studies. The indication for colonoscopy was not always defined; however, the vast majority of studies described it as for either cancer screening or polyp surveillance.
Nine different optical technologies were suitable for inclusion, falling within 5 broad categories: digital chromoendoscopy, dye-based chromoendoscopy, fluorescence analysis, microscopic imaging, and computer-aided recognition.
Digital chromoendoscopy was the most deployed type of optical technology, referring to the selection of specific wavelengths of light that are to be presented in the video output. Three constituent technologies were used on 23,099 colorectal polyps in 84 cohorts across 71 studies—narrow-band imaging (NBI; Olympus Medical Systems), Fuji Intelligent Chromo Endoscopy (FICE; Fujinon), and i-SCAN (PENTAX Medical).
NBI endoscopes only emit light with the absorption spectrum of hemoglobin (blue 415 nm and green 540 nm light), acting to increase the contrast between vessels and mucosa (12). Forty-eight studies ((13–40),(41–60)) suitable for the meta-analysis used NBI to assess 17,568 colorectal polyps, with most of them using the enhancement of visual features such as pit pattern and microvascular patterns to classify based on the Kudo, Sano, or NICE criteria (58,61,62). The diagnostic accuracy of NBI in the identification of neoplastic vs nonneoplastic polyps is shown in Table 1. The definition of neoplasia differed between studies, with the vast majority of neoplastic lesions defined as conventional adenomas or invasive adenocarcinoma. Twenty-three studies included serrated adenomas and polyps in their analysis, with an average prevalence of 2.8%. Traditional serrated adenomas or serrated polyps with cytologic dysplasia were included as neoplastic lesions in all but one study (42). Five additional studies (n = 497) used NBI in the assessment of only sessile serrated adenomas, the presence of dysplasia in sessile serrated polyps, differentiating grades of dysplasia in adenomas, and diagnosing invasive carcinoma; which could not be meta-analyzed due to insufficient numbers (63–67).
FICE and i-SCAN use spectral estimation from white light endoscopy to present images with selected wavelengths, with the latter specifically aiming to enhance structure surface, contrast, or tone (68). These findings can then be classified using similar criteria to NBI. The 12 FICE (14,49,69–78) and 9 i-SCAN (21,79–86) studies suitable for the meta-analysis have a diagnostic accuracy as shown in Table 1. Serrated adenomas were included in half of the studies with a prevalence of 2.1% and were included as neoplastic lesions in all. Serrated polyps with no mention of the presence of dysplasia were included as nonneoplastic lesions in one study (84).
There was no statistically significant difference between the diagnostic accuracy of NBI, FICE, or i-SCAN, which can be seen when the summary receiver operator curves are plotted (Figure 2). Furthermore, within each technology, subgroup analysis based on the confidence of prediction, adequate experience of the endoscopist, the use of high magnification, or the use of high definition did not significantly change the diagnostic potential. When only diminutive polyps were assessed, all technologies showed a similar diagnostic ability compared with polyps of all sizes.
A funnel plot was created for all studies using digital chromoendoscopy (Figure 3). This implies the presence of a significant publication bias, where there is a relative absence of studies with a lower diagnostic odds ratio and higher SD. This would act to overestimate the true diagnostic accuracy of digital chromoendoscopy.
A subgroup analysis was conducted of trained endoscopists using NBI to make high confidence predictions of diminutive colorectal polyps, demonstrating an NPV of 86.8% (80.9%–91.6% CI). When limited to only those on the left side, the NPV was 92.7% (88.3%–96.1% CI), and for only recto-sigmoid polyps, 7 studies demonstrated an NPV of 93.8% (90.9%–96.4% CI).
Most of the studies included in the meta-analysis were found to be of high quality, with 57% of patient cohorts scoring full marks on the QUADAS-2 tool (Appendix 2, Supplementary Digital Content, http://links.lww.com/AJG/A67). The most common potential sources of bias were a lack of clarity regarding how patients were sampled (with no indication of whether a random or consecutive approach was taken) and a lack of endoscopist experience with digital chromoendoscopy before commencement of the study. In 5 patient cohorts, there were abnormally high incidences of conventional adenomas or serrated lesions, reducing the applicability of the findings to usual clinical practice. Furthermore, 4 studies were deemed at high risk of bias when lesions had been assessed with another technology directly before the digital chromoendoscopy. Subgroup analysis of only studies that scored full marks on the QUADAS-2 tool did not significantly change the diagnostic accuracy (data not shown).
Dye chromoendoscopy refers to the application of a dye (instilled via the colonoscope) to the colorectal mucosa, which was applied to 8,336 polyps in 32 cohorts across 30 studies. Most studies used Indigo Carmine (a nonabsorptive dye) to enhance mucosal morphology and allow histologic prediction based on the Kudo criteria (61). Crystal violet (an absorptive dye of eukaryotic nuclei) and acetic acid (not a true dye but causative of color changes through reversible protein denaturing) were used to provide contrast in one study each.
Twenty-nine studies (22,25,69–73,75,78,87–106) assessed the diagnostic accuracy of dye chromoendoscopy in differentiating neoplastic and nonneoplastic polyps (Table 2). Most neoplastic lesions were conventional adenomas or adenocarcinomas, with 7 studies including serrated adenomas as neoplastic at a prevalence of 1.5%.
No difference was evident in the diagnostic accuracy comparing dye and digital chromoendoscopies. When only diminutive polyps were analyzed, dye chromoendoscopy has a trend toward a lower diagnostic accuracy; however, this was not statistically significant. Only one study presented predictions with high confidence and therefore could not be included in a subgroup analysis. There was no change in the diagnostic accuracy for the other subgroups (experienced endoscopists, the use of high magnification, or the use of high-definition endoscopes). Funnel plotting implies a significant publication bias to overestimate the true diagnostic power (Figure 3).
A single study (63) used crystal violet dye for the prediction of adenocarcinoma in adenomatous polyps, showing a sensitivity and specificity of 82.9% and 67.7%, respectively. Quality assessment was conducted using the QUADAS-2 tool, with 29% of patient cohorts scoring full marks (Appendix 3, Supplementary Digital Content, http://links.lww.com/AJG/A67). A lack of sufficient endoscopist training with dye chromoendoscopy was the most common source of potential bias, in addition to a lack of clarity regarding how patients were sampled. It was evident that a different optical technology had been used for histologic prediction immediately before the dye chromoendoscopy in 5 patient cohorts, resulting in a high risk of bias. Subgroup analysis of only patient cohorts that scored full marks on the QUADAS-2 tool did not significantly change the diagnostic accuracy (data not shown).
Histologic prediction based on the analysis of tissue fluorescence was conducted on 558 colorectal polyps across 4 studies. Autofluorescence was most commonly used (n = 525), where fluorescence after excitation with light at 390–470 nm and 540–560 nm can be used either to enhance tissue images with green or magenta (normal and neoplastic, respectively) or for prediction based on the red:green ratio. Three studies (13,19,107) were suitable for the meta-analysis, demonstrating a sensitivity, specificity, positive predictive value (PPV) and NPV of 94.4% (84.0%–99.1% CI), 50.9% (13.2%–88.8% CI), 72.7% (31.8%–95.9% CI), and 85.6% (68.6%–96.6% CI), respectively, in the differentiation of neoplastic and benign polyps. No studies analyzed only diminutive polyps, and there were insufficient studies to conduct any subgroup analyses. Keller et al. (108) used fluorescence analysis to differentiate adenocarcinoma from adenoma using a fluorescein-labeled antibody to the carcinoembryonic antigen, with low diagnostic accuracy in a small cohort. Two of the 3 studies suitable for the meta-analysis were found to have a high risk of bias, as lesions were also assessed with NBI, potentially influencing the prediction made with fluorescence analysis (Appendix 4, Supplementary Digital Content, http://links.lww.com/AJG/A67).
Technologies to endoscopically detail polyps on a cellular level are referred to as microscopic imaging and often incorporate variations in technologies described earlier in this review. The most commonly used technique was confocal laser endomicroscopy (4 studies, n = 519), a technology where probe or scope-based laser light is emitted and images are constructed from the detection of reflected fluorescence. Meta-analysis of these 4 studies (109–112) shows that confocal laser endomicroscopy can differentiate neoplastic and nonneoplastic polyps with a sensitivity, specificity, PPV, and NPV of 93.6% (85.3%–98.3% CI), 92.5% (81.8%–98.1% CI), 92.8% (83.5%–98.1% CI), and 93.1% (82.5%–98.5% CI), respectively. No studies assessed only diminutive polyps, and no predictions were presented with high confidence. It was not possible to conduct a subgroup analysis on the role of high endoscopist experience due to insufficient numbers. Funnel plot of these studies demonstrates a significant publication bias (data not shown). Quality assessment of the included studies with the QUADAS-2 tool demonstrated 2 studies suffering from both unclear endoscopist experience with confocal laser endomicroscopy before start of the study and uncertain patient sampling methods (Appendix 5, Supplementary Digital Content, http://links.lww.com/AJG/A67).
Endocytoscopy was used to assess 149 colorectal polyps across 2 studies (113,114). This technology uses nuclear dyes (toluidine blue or methylene blue) for supra- and subcellular imaging, through the use of ultra-high magnification (×450). The live images give an appearance similar to hemotoxylin- and eosin-stained slides and can be interpreted using criteria based on those for formal histopathologic diagnosis (115). These 2 studies show that this technology has high initial levels of diagnostic accuracy; however, further interpretation is limited by the small sample size.
Three studies used computer-aided recognition of optical inputs to predict adenomatous pathology in colorectal lesions. Laser-induced fluorescence with the wavSTAT4 probe (Spectra Science, San Diego, CA) releases 337 nm light and uses a proprietary algorithm to provide a binary outcome of “suspect” or “nonsuspect” for neoplasia, deployed by 2 studies on 344 colorectal polyps (116,117). Kominami et al. ran machine learning algorithms on 2,247 training images to “teach” a computer the NBI endoscopic features of neoplasia and then performed real-time histologic prediction in vivo on 118 polyps (118). Meta-analysis of these studies demonstrates a sensitivity, specificity, PPV, and NPV of 88.9% (74.2%–96.7% CI), 80.4% (52.6%–95.7% CI), 76.9% (38.8%–96.5% CI), and 88.7% (67.3%–98.1% CI), respectively. There was no high risk of bias or significant concerns with the applicability on quality assessment (Appendix 6, Supplementary Digital Content, http://links.lww.com/AJG/A67).
The NPV for the diagnosis of adenoma in colorectal polyps using both digital and dye chromoendoscopy was subjected to chronologic aggregate meta-analyses, with reference to a clinical implementation threshold of greater than 90%. Given that there was no difference in the diagnostic accuracy of the 3 digital chromoendoscopy technologies, these were pooled to increase the power of the analysis. Figure 4 demonstrates that the estimated NPV of digital chromoendoscopy has never been above 90% and since 2008 has either plateaued or decreased. As the CIs have narrowed with increased power, the NPV has been statistically significantly below the implementation threshold since 2012, with logarithmic trend analysis implying that it will continue to retreat from the threshold.
The chronologic NPV analysis for dye chromoendoscopy is presented in Figure 4. This technology has also never met the clinical implementation threshold and appears increasingly unlikely to do so given the falling NPV estimate as time progresses. The 95% CI currently spans 90% (84.0%–91.5% CI), with logarithmic trend analysis predicting that the NPV estimate will remain below the threshold and continue to fall in the coming years.
This comprehensive meta-analysis assesses the diagnostic accuracy of optical technologies in the real-time in vivo endoscopic histologic prediction of colorectal polyps. It demonstrates that the diagnostic accuracy in differentiating neoplastic from benign tissue is currently insufficient for most routine clinical applications, and it appears unlikely to meet the clinical implementation thresholds in the near future.
Digital chromoendoscopy continues to be the most tested optical technology in polyp characterization, with NBI alone accounting for more than half of all polyps studied. Identification of adenoma in colorectal polyps is the most prevalent research question, primarily given its direct relevance to clinical practice. This meta-analysis found no difference between the accuracies of the different digital chromoendoscopy technologies, which is expected considering they work in similar manners with the application of the same diagnostic criteria. NBI has been the technology of choice in the largest implementation studies to date (20,23,37,53), likely as a result of this technology being more commonly found in endoscopy suites and support from a greater wealth of published series, rather than because of presumed higher accuracy. The estimated 92.2% (90.6%–93.9% CI) sensitivity, 84.0% (81.5%–86.3% CI) specificity, and 85.7% (82.9%–88.4% CI) NPV of digital chromoendoscopy were similar to those of previous reviews; however, this must be considered within the context of a significant limitation. Publication bias in these studies was stark, inevitably overestimating the diagnostic value. This is evident not only from the inspection of funnel plots but also from the more recent, well-powered implementation studies, where the diagnostic accuracies are below what would be expected from reading many previously reported cohorts.
Dye chromoendoscopy has been historically used for the detection of dysplastic lesions in inflammatory bowel disease rather than for the characterization of polyps in low-risk populations; however, when applied in this setting, it may have greater diagnostic accuracy than digital chromoendoscopy. This could not be distinguished statistically as insufficient power was evident from significantly fewer polyps being analyzed. The reasons for dye chromoendoscopy not being implemented as frequently as digital chromoendoscopy are likely to include the increased time required for the instillation, the risk of allergic reaction to dye agents (albeit low), and a greater learning curve for the application and interpretation. The technology benefits from not requiring any specialist equipment and can be conducted using any colonoscope with an instrument channel, increasing the scope for use in health care settings with limited resources. Heterogeneity exists where different dyes are used, given that they have varying methods of action and therefore may differently reflect underlying pathology. It was not possible to determine an optimum dye here due to the low numbers of studies. This technology was also shown to suffer from a distinct publication bias.
Fluorescence analysis and microscopic imaging are relatively new optical technologies and are limited in their penetration into the endoscopy suite and therefore into clinical studies. This is likely as a result of requiring specialist equipment, the cost of which may not be justified without clinical evidence underpinning the clinical use. Estimates of the diagnostic accuracy were affected by low power and therefore large CIs that make direct comparisons problematic. It was evident that fluorescence imaging suffers from a high level of false positives. Unlike the other optical technologies in this review, which use inputs to make inferences about the underlying tissue structure and therefore histologic subtype, microscopic imaging attempts to visualize cells directly. Given that histologic criteria are the gold standard for the diagnosis of adenoma, this is a promising approach if it can be demonstrated to be accurate and can be reliably incorporated into a clinical setting. The diagnostic accuracy in the initial studies appears high and comparable with digital and dye chromoendoscopy: however, this low number of polyps analyzed limits any further interpretation.
Machine learning techniques can powerfully identify relationships in data with a greater depth than that of the human mind and are increasingly being developed and applied within health care (119). Not only does the computing power now exists to run these algorithms on video inputs in real-time using inexpensive hardware, but the diagnostic accuracy theoretically improves over time as the machine is exposed to new data. The flexibility of the approach is such that a variety of inputs can be subjected for the analysis, with studies in this review interpreting the shape of fluorescence spectra and the morphologic features identified on NBI. A similar approach has been presented by Byrne et al. (120) on 125 diminutive polyps; however, this was not suitable for inclusion in this review as the predictions were not made in vivo. Mori et al. have recently published a notable study, using real-time, computer-aided interpretation of endocytoscope images to predict the histology of diminutive polyps, with high diagnostic accuracy. This is an exciting field in its infancy, and it is likely that further studies will be published using machine learning techniques, but caution must be taken when interpreting the headline diagnostic accuracies. The statistical models are prone to overfitting when created and tested on a single dataset, and there is a high risk of even small selection biases or batch effects being incorporated. These issues can be addressed using rigorous external validation methods once the models have been created.
Subgroup analyses were conducted to identify the clinically relevant technical and polyp features that may influence the diagnostic accuracy. The lack of difference when only diminutive polyps are analyzed implies that the technical challenges of optical technologies analyzing a smaller surface area are not clinically relevant and perhaps are superseded by bias from the endoscopist knowing adenomatous risk based on the size of the polyp. Studies where endoscopists predicted histology with “high confidence” were largely restricted to NBI and did not show an increase in the diagnostic accuracy. This finding may be a type II error, however is consistent with other meta-analyses (6). If such an error does exist, it may demonstrate that the diagnostic accuracy of subconscious histologic prediction at initial white light inspection is sufficiently high that posttest accuracies with high or low confidence are difficult to distinguish. This meta-analysis did not replicate the findings of Rees et al. (53) who showed that more experienced endoscopists may make more accurate predictions; however, this may be related to a lesser definition of experience used here.
Colorectal polyps can have a variety of histologic subtypes, and it was evident that study authors sometimes differed in their definition of neoplastic and nonneoplastic lesions. To address a research question driven by the clinical need, the vast majority of studies compared neoplastic lesions (conventional adenomas or adenocarcinoma) with nonneoplastic lesions (such as hyperplastic, lymphoid aggregates or normal mucosal folds). However, as would be expected in a clinical care, a small number of serrated adenomas were encountered. These were removed from the meta-analysis where the data allowed; however, for the studies where these were included, it was in relatively small numbers, with them almost always defined as neoplastic. Given that sessile and traditional serrated adenomas tend to have a differing appearance to conventional adenomas (121), it may be that the diagnostic accuracy for these lesions is suboptimal in a study optimized to identify conventional adenomas. This is a potential source of heterogeneity in some of the included studies; however, given the small numbers of serrated lesions, this should not significantly affect the interpretation of the presented diagnostic accuracies.
A novel approach in this meta-analysis is a chronologic analysis to detail how diagnostic accuracy has evolved over time, giving an indication as to whether optical technologies are improving at a rate sufficient to meet a threshold for clinical implementation. Defining such a threshold can be challenging as it can feel arbitrary, is difficult to determine from an evidence base, and is problematic when applying it across different polyp types, sizes, and locations. An NPV for adenomatous histology of greater than 90% is the most frequently defined threshold in the literature and was used in this review. Although it is likely that 90% is an inadequately low level to gain widespread acceptance (122), unfortunately it is beyond the scope of this review to suggest changes to such a threshold and this will be a keen area of research interest in the future. This meta-analysis indicates that neither digital nor dye chromoendoscopy is currently meeting the clinical implementation threshold and appears increasingly unlikely to do so over time. As increasingly powerful, preregistered clinical studies are being published, the estimated NPV is either decreasing or plateauing, often diluting the effect of a previous publication bias. To identify niches of clinical practice where optical technologies will have sufficient accuracy for clinical use, there is a move toward defining more restrictive subgroups of interest based on the size and location of the polyp such as in the Preservation and Incorporation of Valuable endoscopic Innovations statements (4). This approach makes recruiting patients for sufficient power increasingly difficult and simultaneously limits the scope for use in clinical practice. Trained endoscopists making high confidence predictions of diminutive recto-sigmoid polyps with digital chromoendoscopy significantly improve the NPV such that it is over 90%, and as such, this could support a “diagnose and leave” strategy as endorsed by the American and European Societies for Gastrointestinal Endoscopy (123,124). This may represent a true accuracy of the technology in this cohort; however, the number of studies was low, many analyses were conducted post hoc, and these data are still limited by a significant publication bias. This review cannot suggest routine clinical implementation given these limitations and prospective studies need to be designed with accuracy in this cohort as the primary outcome.
The QUADAS-2 tool was used to assess the quality of studies that underwent meta-analysis. Due to the restrictive inclusion and exclusion criteria in this review, the studies were generally of high quality and it was rare that a high source of bias or lack of applicability was encountered. The greatest risk of bias was caused when an optical technology was used to assess a lesion directly after another technology had been used on the same lesion. It is likely that this approach would bias the second interpretation. Furthermore, many studies did not adequately describe how patients were sampled and assigned to the intervention, raising the possibility of a selection bias. Studies in this field are well suited to a randomized approach with one technology used in each arm, which should be applied to future studies to resolve many of the quality issues identified in this review. The progressive decline in the diagnostic accuracy as studies of higher quality are published has marked implications for future research. It shows that the current practice of incremental technological improvement with more focused patient cohort definition is unlikely to be sufficient in changing clinical practice in the near future. To disrupt the practice of real-time polyp diagnosis, new clinical implementation studies should be pragmatically designed and conducted once step-change technological innovation has occurred.
A range of additional technologies at an earlier stage of development were identified but did not meet the inclusion criteria. These were largely based on the spectroscopic analysis (Raman and elastic scattering spectroscopies) and novel approaches to digital chromoendoscopy including blue laser imaging. A technology that could finally achieve diagnostic accuracies sufficient for clinical implementation may not necessarily use optical manipulation but may gather biologic data to predict the underlying tissue type. Detailing a polyp's biologic profile has benefits beyond real-time diagnosis and may complement the histopathologic analysis by giving new insights into future risk. Such data may then be integrated into clinical decision-making pathways to improve patient care, such as by manipulating surveillance durations based on new risk profiles. Being able to sample biologic features such as the lipid profile of mucosal cellular membranes can now be conducted in real time using ambient mass spectrometry approaches such as Rapid Evaporative Ionisation Mass Spectrometry. This technique has demonstrated an NPV of 98.6% on ex vivo samples and is due to be prospectively validated in vivo (125). Such technological innovations can also be used to resolve other weaknesses in current optical approaches such as requiring a learning curve for the interpretation by using binary computerized outputs.
This meta-analysis was designed to avoid many of the weaknesses of previous reviews, including being limited by predefining a small number of technologies, including predictions that were not made in real time, methodologic incorporation of heterogeneity, and presenting restricted measures of diagnostic accuracy (4,6,7). Despite this, it is a significant challenge to understand the role of optical technologies when such heterogeneity and publication bias exist. Heterogeneity of constituent study diagnostic accuracies may be due to variations in the underlying prevalence of adenomas, where narrow inclusion criteria could not account for different population ages, the use of previous screening practices, or ethnicity. In addition, inter- and intraobserver variability may result from a differing endoscopist experience, innate skill level, and technological availability. Such factors need to be clearly measured and defined in future studies of implementation to increase the quality and external validity.
This meta-analysis demonstrates that despite publication bias overestimating diagnostic potential, optical technologies are generally insufficient for routine clinical implementation in the prediction of colorectal polyp histology. NBI making predictions of diminutive recto-sigmoid polyps with high confidence appears to be sufficiently accurate to support a “diagnose and leave” strategy; however, concerns over study numbers and methodologies are likely to warrant future prospective trials. Chronologic analysis has identified a falling diagnostic power over time, and step-change technological innovation is likely to be required.
CONFLICTS OF INTEREST
Guarantor of the article: Sam E. Mason, BSc, MBBS, MRCS.
Specific author contributions: S.E.M.: study design, data gathering and synthesis, manuscript generation. L.P.: data gathering and synthesis, manuscript review and contributions, approval of the final draft. Z.T., A.D., and J.M.K.: study design, manuscript review and contributions, approval of the final draft.
Financial support: This research was co-funded by the NIHR Imperial Biomedical Research Centre (BRC) and an NIHR Invention for Innovation grant.
Potential competing interests: S.E.M. is funded under a Cancer Research UK PhD Fellowship. J.M.K. performs consultancy work for Ethicon. The remaining authors have no potential conflicts.
WHAT IS KNOWN
- ✓ Optical technologies attempt to endoscopically predict the histology of colorectal polyps.
- ✓ It is unclear whether diagnostic accuracy is sufficient for routine clinical implementation.
WHAT IS NEW HERE
- ✓ Digital chromoendoscopy making high confidence predictions of diminutive recto-sigmoid polyps is currently the only application of optical technology with an NPV for adenoma of greater than 90%.
- ✓ Chronologic assessment demonstrates that estimates of accuracy are falling over time.
- ✓ Step-change innovation rather than incremental improvement in existing technologies is likely to be necessary.
We thank Dr. Luisa Doria and Mr. Derek K.T. Yeung for their assistance in manuscript translation.
1. Zauber AG, Winawer SJ, O'Brien MJ, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med 2012;366(8):687–96.
2. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2018. CA Cancer J Clin 2018;68(1):7–30.
3. Qumseya BJ, Coe S, Wallace MB. The effect of polyp location and patient gender on the presence of dysplasia in colonic polyps. Clin Transl Gastroenterol 2012;3:e20.
4. Committee AT, Abu Dayyeh BK, Thosani N, et al. ASGE Technology Committee systematic review and meta-analysis assessing the ASGE PIVI thresholds for adopting real-time endoscopic assessment of the histology of diminutive colorectal polyps. Gastrointest Endosc 2015;81(3):502 e501–502.e16.
5. Kessler WR, Imperiale TF, Klein RW, et al. A quantitative assessment of the risks and cost savings of forgoing histologic examination of diminutive polyps. Endoscopy 2011;43(8):683–91.
6. Wanders LK, East JE, Uitentuis SE, et al. Diagnostic performance of narrowed spectrum endoscopy, autofluorescence imaging, and confocal laser endomicroscopy for optical diagnosis of colonic polyps: A meta-analysis. Lancet Oncol 2013;14(13):1337–47.
7. McGill SK, Evangelou E, Ioannidis JP, et al. Narrow band imaging to differentiate neoplastic and non-neoplastic colorectal polyps in real time: A meta-analysis of diagnostic operating characteristics. Gut 2013;62(12):1704–13.
8. McInnes MDF, Moher D, Thombs BD, et al. Preferred reporting items for a systematic review and meta-analysis of diagnostic test accuracy studies: The PRISMA-DTA statement. JAMA 2018;319(4):388–96.
9. Schlemper RJ, Riddell RH, Kato Y, et al. The Vienna classification of gastrointestinal epithelial neoplasia. Gut 2000;47(2):251–5.
10. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: A revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011;155(8):529–36.
11. Guo J, Riebler A. Meta4diag: Bayesian bivariate meta-analysis of diagnostic test studies for routine practice. J Stat Softw 2018;83(1):31.
12. Olympus Medical Systems. Narrow band imaging (NBI). A New Wave of Diagnostic Possibilities. Olympus Europa Co: Hamburg, Germany, 2018.
13. van den Broek FJ, Fockens P, Van Eeden S, et al. Clinical evaluation of endoscopic trimodal imaging for the detection and differentiation of colonic polyps. Clin Gastroenterol Hepatol 2009;7(3):288–95.
14. Buchner AM, Shahid MW, Heckman MG, et al. Comparison of probe-based confocal laser endomicroscopy with virtual chromoendoscopy for classification of colon polyps. Gastroenterology 2010;138(3):834–42.
15. Dai J, Shen YF, Sano Y, et al. Evaluation of narrow-band imaging in the diagnosis of colorectal lesions: Is a learning curve involved? Dig Endosc 2013;25(2):180–8.
16. East JE, Suzuki N, Bassett P, et al. Narrow band imaging with magnification for the characterization of small and diminutive colonic polyps: Pit pattern and vascular pattern intensity. Endoscopy 2008;40(10):811–7.
17. Hewett DG, Huffman ME, Rex DK. Leaving distal colorectal hyperplastic polyps in place can be achieved with high accuracy by using narrow-band imaging: An observational study. Gastrointest Endosc 2012;76(2):374–80.
18. Kuiper T, Marsman WA, Jansen JM, et al. Accuracy for optical diagnosis of small colorectal polyps in nonacademic settings. Clin Gastroenterol Hepatol 2012;10(9):1016–20; quiz e1079.
19. Kuiper T, van den Broek FJ, Naber AH, et al. Endoscopic trimodal imaging detects colonic neoplasia as well as standard video endoscopy. Gastroenterology 2011;140(7):1887–94.
20. Ladabaum U, Fioritto A, Mitani A, et al. Real-time optical biopsy of colon polyps with narrow band imaging in community practice does not yet meet key thresholds for clinical decisions. Gastroenterology 2013;144(1):81–91.
21. Lee CK, Lee SH, 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 2011;74(3):603–9.
22. Machida H, Sano Y, Hamamoto Y, et al. Narrow-band imaging in the diagnosis of colorectal mucosal lesions: A pilot study. Endoscopy 2004;36(12):1094–8.
23. Paggi S, Rondonotti E, Amato A, et al. Resect and discard strategy in clinical practice: A prospective cohort study. Endoscopy 2012;44(10):889–904.
24. Shahid MW, Buchner AM, Heckman MG, et al. Diagnostic accuracy of probe-based confocal laser endomicroscopy and narrow band imaging for small colorectal polyps: A feasibility study. Am J Gastroenterol 2012;107(2):231–9.
25. Sakamoto T, Matsuda T, Aoki T, et al. Time saving with narrow-band imaging for distinguishing between neoplastic and non-neoplastic small colorectal lesions. J Gastroenterol Hepatol 2012;27(2):351–5.
26. Rastogi A, Early DS, Gupta N, et al. Randomized, controlled trial of standard-definition white-light, high-definition white-light, and narrow-band imaging colonoscopy for the detection of colon polyps and prediction of polyp histology. Gastrointest Endosc 2011;74(3):593–602.
27. Rex DK. Narrow-band imaging without optical magnification for histologic analysis of colorectal polyps. Gastroenterology 2009;136(4):1174–81.
28. Rogart JN, Jain D, Siddiqui UD, et al. Narrow-band imaging without high magnification to differentiate polyps during real-time colonoscopy: Improvement with experience. Gastrointest Endosc 2008;68(6):1136–45.
29. Rotondano G, Bianco MA, Sansone S, et al. Trimodal endoscopic imaging for the detection and differentiation of colorectal adenomas: A prospective single-centre clinical evaluation. Int J Colorectal Dis 2012;27(3):331–6.
30. Sano Y, Ikematsu H, Fu KI, et al. Meshed capillary vessels by use of narrow-band imaging for differential diagnosis of small colorectal polyps. Gastrointest Endosc 2009;69(2):278–83.
31. Singh R, Nordeen N, Mei SL, et al. West meets East: Preliminary results of narrow band imaging with optical magnification in the diagnosis of colorectal lesions: A multicenter Australian study using the modified Sano's classification. Dig Endosc 2011;23(Suppl 1):126–30.
32. Yoo HY, Lee MS, Ko BM, et al. Correlation of narrow band imaging with magnifying colonoscopy and histology in colorectal tumors. Clin Endosc 2011;44(1):44–50.
33. Zhou QJ, Yang JM, Fei BY, et al. Narrow-band imaging endoscopy with and without magnification in diagnosis of colorectal neoplasia. World J Gastroenterol 2011;17(5):666–70.
34. Rogart JN, Aslanian HR, Siddiqui UD. Narrow band imaging to detect residual or recurrent neoplastic tissue during surveillance endoscopy. Dig Dis Sci 2011;56(2):472–8.
35. Ikematsu H, Matsuda T, Osera S, et al. Usefulness of narrow-band imaging with dual-focus magnification for differential diagnosis of small colorectal polyps. Surg Endosc 2015;29(4):844–50.
36. Iwatate M, Sano Y, Hattori S, et al. The addition of high magnifying endoscopy improves rates of high confidence optical diagnosis of colorectal polyps. Endosc Int Open 2015;3(2):E140–145.
37. Sano W, Sano Y, Iwatate M, et al. Prospective evaluation of the proportion of sessile serrated adenoma/polyps in endoscopically diagnosed colorectal polyps with hyperplastic features. Endosc Int Open 2015;3(4):E354–358.
38. Sola-Vera J, Cuesta R, Uceda F, et al. Accuracy for optical diagnosis of colorectal polyps in clinical practice. Rev Esp Enferm Dig 2015;107(5):255–61.
39. Chandran S, Parker F, Lontos S, et al. Can we ease the financial burden of colonoscopy? Using real-time endoscopic assessment of polyp histology to predict surveillance intervals. Intern Med J 2015;45(12):1293–9.
40. Paggi S, Rondonotti E, Amato A, et al. Narrow-band imaging in the prediction of surveillance intervals after polypectomy in community practice. Endoscopy 2015;47(9):808–14.
41. Kuruvilla N, Paramsothy R, Gill R, et al. A prospective dual-center proof-of-principle study evaluating the incremental benefit of narrow-band imaging with a fixed zoom function in real-time prediction of polyp histology. Can we resect and discard? Gastrointest Endosc 2015;82(2):362–9.
42. Kaltenbach T, Rastogi A, Rouse RV, et al. Real-time optical diagnosis for diminutive colorectal polyps using narrow-band imaging: The VALID randomised clinical trial. Gut 2015;64(10):1569–77.
43. Seref Koksal A, Yildiz H, Taskiran I, et al. Low magnification narrow band imaging by inexperienced endoscopists has a high accuracy in differentiation of colon polyp histology. Clin Res Hepatol Gastroenterol 2014;38(6):763–9.
44. Wallace MB, Crook JE, Coe S, et al. Accuracy of in vivo colorectal polyp discrimination by using dual-focus high-definition narrow-band imaging colonoscopy. Gastrointest Endosc 2014;80(6):1072–87.
45. Takeuchi Y, Hanafusa M, Kanzaki H, et al. Proposal of a new 'resect and discard' strategy using magnifying narrow band imaging: Pilot study of diagnostic accuracy. Dig Endosc 2014;26(Suppl 2):90–7.
46. Singh R, Jayanna M, Navadgi S, et al. Narrow-band imaging with dual focus magnification in differentiating colorectal neoplasia. Dig Endosc 2013;25(Suppl 2):16–20.
47. Repici A, Hassan C, Radaelli F, et al. Accuracy of narrow-band imaging in predicting colonoscopy surveillance intervals and histology of distal diminutive polyps: Results from a multicenter, prospective trial. Gastrointest Endosc 2013;78(1):106–14.
48. Takeuchi Y, Hanafusa M, Kanzaki H, et al. An alternative option for “resect and discard” strategy, using magnifying narrow-band imaging: A prospective “proof-of-principle” study. J Gastroenterol 2015;50(10):1017–26.
49. Kang HY, Kim YS, Kang SJ, et al. Comparison of narrow band imaging and Fujinon intelligent color enhancement in predicting small colorectal polyp histology. Dig Dis Sci 2015;60(9):2777–84.
50. Pohl H, Bensen SP, Toor A, et al. Quality of optical diagnosis of diminutive polyps and associated factors. Endoscopy 2016;48(9):817–22.
51. Ashktorab H, Etaati F, Rezaeean F, et al. Can optical diagnosis of small colon polyps be accurate? Comparing standard scope without narrow banding to high definition scope with narrow banding. World J Gastroenterol 2016;22(28):6339–456.
52. Klare P, Haller B, Wormbt S, et al. Narrow-band imaging vs. high definition white light for optical diagnosis of small colorectal polyps: A randomized multicenter trial. Endoscopy 2016;48(10):909–15.
53. Rees CJ, Rajasekhar PT, Wilson A, et al. Narrow band imaging optical diagnosis of small colorectal polyps in routine clinical practice: The detect inspect characterise resect and discard 2 (DISCARD 2) study. Gut 2017;66(5):887–95.
54. Szura M, Pasternak A, Bucki K, et al. Two-stage optical system for colorectal polyp assessments. Surg Endosc 2016;30(1):204–14.
55. Ren J, Jiang XL. Narrow-band imaging of meshed capillary vessels for differential diagnosis of colorectal lesions. World Chin J Dig 2012;20(6):473–8.
56. Belderbos TDG, van Oijen MGH, Moons LMG, et al. Implementation of real-time probe-based confocal laser endomicroscopy (pCLE) for differentiation of colorectal polyps during routine colonoscopy. Endosc Int Open 2017;5(11):E1104–E1110.
57. Okamoto Y, Watanabe H, Tominaga K, et al. Evaluation of microvessels in colorectal tumors by narrow band imaging magnification: Including comparison with magnifying chromoendoscopy. Dig Dis Sci 2011;56(2):532–8.
58. Hewett DG, Kaltenbach T, Sano Y, et al. Validation of a simple classification system for endoscopic diagnosis of small colorectal polyps using narrow-band imaging. Gastroenterology 2012;143(3):599–607 e591.
59. Canales Sevilla O, Miyagui Maeda J, Takano Moron J, et al. [NBI utility and optical magnification in the differential diagnosis of neoplastic and non-neoplastic colorectal lesions in Peru]. Rev Gastroenterol Peru 2010;30(4):277–83.
60. Salazar Muente F, Barreda Costa C, Barriga Briceno JA. [Utility of the capilar pattern in the NBI diagnosis of superficial lesions of the colon: Prospective validation in a private endoscopic center Lima, Peru]. Rev Gastroenterol Peru 2012;32(3):281–9.
61. Kudo S, Rubio CA, Teixeira CR, et al. Pit pattern in colorectal neoplasia: Endoscopic magnifying view. Endoscopy 2001;33(4):367–73.
62. Sano Y, Horimatsu T, Fu K, et al. Magnifying observations of microvascular architecture of colorectal lesions using a narrow-band imaging system. Dig Endosc 2006;18(Suppl 1):8.
63. Goto N, Kusaka T, Tomita Y, et al. Magnifying narrow-band imaging with acetic acid to diagnose early colorectal cancer. World J Gastroenterol 2014;20(43):16306–10.
64. Yamashina T, Takeuchi Y, Uedo N, et al. Diagnostic features of sessile serrated adenoma/polyps on magnifying narrow band imaging: A prospective study of diagnostic accuracy. J Gastroenterol Hepatol 2015;30(1):117–23.
65. Katagiri A, Fu KI, Sano Y, et al. Narrow band imaging with magnifying colonoscopy as diagnostic tool for predicting histology of early colorectal neoplasia. Aliment Pharmacol Ther 2008;27(12):1269–74.
66. Tate DJ, Jayanna M, Awadie H, et al. A standardized imaging protocol for the endoscopic prediction of dysplasia within sessile serrated polyps (with video). Gastrointest Endosc 2018;87(1):222–31.e2.
67. Liu H, Wu J, Liu XC, et al. Correlation between microvascular characteristics and the expression of MVD, IGF-1 and STAT3 in the development of colonic polyps carcinogenesis. Exp Ther Med J 2017;13(1):49–54.
68. Fujinon. Fuji Intelligent Chromo Endoscopy. Fujinon Corporation: Saitama, Japan, 2005, pp 6.
69. Liu YX, Huang LY, Bian XP, et al. Fuji Intelligent Chromo Endoscopy and staining technique for the diagnosis of colon tumor. Chin Med J 2008;121(11):977–82.
70. Togashi K, Osawa H, Koinuma K, et al. A comparison of conventional endoscopy, chromoendoscopy, and the optimal-band imaging system for the differentiation of neoplastic and non-neoplastic colonic polyps. Gastrointest Endosc 2009;69(3 Pt 2):734–41.
71. Pohl J, Lotterer E, Balzer C, et al. Computed virtual chromoendoscopy versus standard colonoscopy with targeted indigocarmine chromoscopy: A randomised multicentre trial. Gut 2009;58(1):73–8.
72. dos Santos CE, Lima JC, Lopes CV, et al. Computerized virtual chromoendoscopy versus indigo carmine chromoendoscopy combined with magnification for diagnosis of small colorectal lesions: A randomized and prospective study. Eur J Gastroenterol Hepatol 2010;22(11):1364–71.
73. Longcroft-Wheaton GR, Higgins B, Bhandari P. Flexible spectral imaging color enhancement and indigo carmine in neoplasia diagnosis during colonoscopy: A large prospective UK series. Eur J Gastroenterol Hepatol 2011;23(10):903–11.
74. Kim YS, Kim D, Chung SJ, et al. Differentiating small polyp histologies using real-time screening colonoscopy with Fuji Intelligent Color Enhancement. Clin Gastroenterol Hepatol 2011;9(9):744–9 e741.
75. Dos Santos CE, Malaman D, Lopes CV, et al. Digital chromoendoscopy for diagnosis of diminutive colorectal lesions. Diagn Ther Endosc 2012;2012:279521.
76. Longcroft-Wheaton G, Brown J, Cowlishaw D, Higgins B, et al. High-definition vs. standard-definition colonoscopy in the characterization of small colonic polyps: Results from a randomized trial. Endoscopy 2012;44(10):905–10.
77. Dos Santos CEO, Moreira H, Pereira-Lima JC, et al. Hyoscine butylbromide for colorectal polyp detection: Prospective, randomized, placebo-controlled trial. Clinics (Sao Paulo) 2017;72(7):395–9.
78. Santos CE, Pereira-Lima JC, Lopes CV, et al. [Comparative study between MBI (FICE) and magnification chromoendoscopy with indigo carmine in the differential diagnosis of neoplastic and non-neoplastic lesions of the colorectum]. Arq Gastroenterol 2009;46(2):111–5.
79. Hoffman A, Sar F, Goetz M, et al. High definition colonoscopy combined with i-Scan is superior in the detection of colorectal neoplasias compared with standard video colonoscopy: A prospective randomized controlled trial. Endoscopy 2010;42(10):827–33.
80. Chan J, Lin L, Feiler M, et al. Comparative effectiveness of i-SCANTM and high-definition white light characterizing small colonic polyps. World J Gastroenterol 2012;18(41):5905–11.
81. Hong S, Choe W, Lee J, et al. Prospective, randomized, back-to-back trial evaluating the usefulness of i-SCAN in screening colonoscopy. Gastrointest Endosc 2012;75(5):1011–21.e2.
82. Pigò F, Bertani H, Manno M, et al. i-Scan high-definition white light endoscopy and colorectal polyps: prediction of histology, interobserver and intraobserver agreement. Int J Colorectal Dis 2013;28(3):399–406.
83. Schachschal G, Mayr M, Treszl A, et al. Endoscopic versus histological characterisation of polyps during screening colonoscopy. Gut 2014;63(3):458–65.
84. Basford PJ, Longcroft-Wheaton G, Higgins B, et al. High-definition endoscopy with i-Scan for evaluation of small colon polyps: The HiSCOPE study. Gastrointest Endosc 2014;79(1):111–8.
85. Rath T, Tontini GE, Nägel A, et al. High-definition endoscopy with digital chromoendoscopy for histologic prediction of distal colorectal polyps. BMC Gastroenterol 2015;15:145.
86. Hoffman A, Kagel C, Goetz M, et al. Recognition and characterization of small colonic neoplasia with high-definition colonoscopy using i-Scan is as precise as chromoendoscopy. Dig Liver Dis 2010;42(1):45–50.
87. Kohut M, Liszka L, Boldys H, et al. Pit pattern analysis using acetic-acid magnification chromoendoscopy in predicting histopathology of small colorectal polyps. The diagnostic yield and intra-/inter-observer reproducibility. Przeglad Gastroenterologiczny 2009;4(6):6.
88. Ljubicic N, Kujundzic M, Banic M, et al. The role of standard videochromocolonoscopy in distinguishing adenomatous from nonadenomatous diminutive colorectal polyps. Acta Clin Croat 2001;40(3):197–201.
89. Bianco MA, Rotondano G, Marmo R, et al. Predictive value of magnification chromoendoscopy for diagnosing invasive neoplasia in nonpolypoid colorectal lesions and stratifying patients for endoscopic resection or surgery. Endoscopy 2006;38(5):470–6.
90. De Palma GD, Rega M, Masone S, et al. Conventional colonoscopy and magnified chromoendoscopy for the endoscopic histological prediction of diminutive colorectal polyps: A single operator study. World J Gastroenterol 2006;12(15):2402–5.
91. Apel D, Jakobs R, Schilling D, et al. Accuracy of high-resolution chromoendoscopy in prediction of histologic findings in diminutive lesions of the rectosigmoid. Gastrointest Endosc 2006;63(6):824–8.
92. Togashi K, Hewett DG, Whitaker DA, et al. The use of acetic acid in magnification chromocolonoscopy for pit pattern analysis of small polyps. Endoscopy 2006;38(6):613–6.
93. Kato S, Fu KI, Sano Y, et al. Magnifying colonoscopy as a non-biopsy technique for differential diagnosis of non-neoplastic and neoplastic lesions. World J Gastroenterol 2006;12(9):1416–20.
94. Fu KI, Sano Y, Kato S, et al. Chromoendoscopy using indigo carmine dye spraying with magnifying observation is the most reliable method for differential diagnosis between non-neoplastic and neoplastic colorectal lesions: A prospective study. Endoscopy 2004;36(12):1089–93.
95. Hurlstone DP, Cross SS, Adam I, et al. Efficacy of high magnification chromoscopic colonoscopy for the diagnosis of neoplasia in flat and depressed lesions of the colorectum: A prospective analysis. Gut 2004;53(2):284–90.
96. Liu HH, Kudo SE, Juch JP. Pit pattern analysis by magnifying chromoendoscopy for the diagnosis of colorectal polyps. J Formos Med Assoc 2003;102(3):178–82.
97. Eisen GM, Kim CY, Fleischer DE, et al. High-resolution chromoendoscopy for classifying colonic polyps: A multicenter study. Gastrointest Endosc 2002;55(6):687–94.
98. Kiesslich R, von Bergh M, Hahn M, et al. Chromoendoscopy with indigocarmine improves the detection of adenomatous and nonadenomatous lesions in the colon. Endoscopy 2001;33(12):1001–6.
99. Axelrad AM, Fleischer DE, Geller AJ, et al. High-resolution chromoendoscopy for the diagnosis of diminutive colon polyps: Implications for colon cancer screening. Gastroenterology 1996;110(4):1253–8.
100. Longcroft-Wheaton G, Brown J, Cowlishaw D, et al. High-definition vs. standard-definition endoscopy with indigo carmine for the in vivo diagnosis of colonic polyps. United Eur Gastroenterol J 2013;1(6):425–9.
101. Ince AT, Bolukbas C, Peker O, et al. Pit pattern type analyses of colon polyps with high-resolution colonoscope. Hepatogastroenterology 2007;54(73):67–70.
102. Konishi K, Kaneko K, Kurahashi T, et al. A comparison of magnifying and nonmagnifying colonoscopy for diagnosis of colorectal polyps: A prospective study. Gastrointest Endosc 2003;57(1):48–53.
103. Togashi K, Konishi F, Ishizuka T, et al. Efficacy of magnifying endoscopy in the differential diagnosis of neoplastic and non-neoplastic polyps of the large bowel. Dis Colon Rectum 1999;42(12):1602–8.
104. Tung SY, Wu CS, Su MY. Magnifying colonoscopy in differentiating neoplastic from nonneoplastic colorectal lesions. Am J Gastroenterol 2001;96(9 Suppl):2628–32.
105. Urban O, Fojtík P, Liberda M, et al. Magnifying chromocolonoscopy—Beginner's diagnostic accuracy. Endoskopie 2005;14(1):3.
106. Averbach M, Zanoni EC, Correa PA, et al. [High resolution chromoendoscopy in the differential diagnosis of neoplastic and non-neoplastic polyps]. Arq Gastroenterol 2003;40(2):99–103.
107. Aihara H, Saito S, Inomata H, et al. Computer-aided diagnosis of neoplastic colorectal lesions using “real-time” numerical color analysis during autofluorescence endoscopy. Eur J Gastroenterol Hepatol 2013;25(4):488–94.
108. Keller R, Winde G, Terpe HJ, et al. Fluorescence endoscopy using a fluorescein-labeled monoclonal antibody against carcinoembryonic antigen in patients with colorectal carcinoma and adenoma. Endoscopy 2002;34(10):801–7.
109. Xie XJ, Li CQ, Zuo XL, et al. Differentiation of colonic polyps by confocal laser endomicroscopy. Endoscopy 2011;43(2):87–93.
110. Kiesslich R, Burg J, Vieth M, et al. Confocal laser endoscopy for diagnosing intraepithelial neoplasias and colorectal cancer in vivo. Gastroenterology 2004;127(3):706–13.
111. Sanduleanu S, Driessen A, Gomez-Garcia E, et al. In vivo diagnosis and classification of colorectal neoplasia by chromoendoscopy-guided confocal laser endomicroscopy. Clin Gastroenterol Hepatol 2010;8(4):371–8.
112. Shahid M, Buchner A, Raimondo M, et al. Accuracy of real-time vs. blinded offline diagnosis of neoplastic colorectal polyps using probe-based confocal laser endomicroscopy: A pilot study. Endoscopy 2012;44(4):6.
113. Rotondano G, Bianco MA, Salerno R, et al. Endocytoscopic classification of preneoplastic lesions in the colorectum. Int J Colorectal Dis 2010;25(9):1111–6.
114. Mori Y, Kudo S, Ikehara N, et al. Comprehensive diagnostic ability of endocytoscopy compared with biopsy for colorectal neoplasms: A prospective randomized noninferiority trial. Endoscopy 2013;45(2):98–105.
115. Sasajima K, Kudo SE, Inoue H, et al. Real-time in vivo virtual histology of colorectal lesions when using the endocytoscopy system. Gastrointest Endosc 2006;63(7):1010–7.
116. Rath T, Tontini GE, Vieth M, et al. In vivo real-time assessment of colorectal polyp histology using an optical biopsy forceps system based on laser-induced fluorescence spectroscopy. Endoscopy 2016;48(6):557–62.
117. Kuiper T, Alderlieste YA, Tytgat KM, et al. Automatic optical diagnosis of small colorectal lesions by laser-induced autofluorescence. Endoscopy 2015;47(1):56–62.
118. Kominami Y, Yoshida S, Tanaka S, et al. Computer-aided diagnosis of colorectal polyp histology by using a real-time image recognition system and narrow-band imaging magnifying colonoscopy. Gastrointest Endosc 2016;83(3):643–9.
119. Ravi D, Wong C, Deligianni F, et al. Deep learning for health informatics. IEEE J Biomed Health Inform 2017;21(1):4–21.
120. Byrne MF, Chapados N, Soudan F, et al. Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model. Gut 2019;68:94–100.
121. Rex DK, Ahnen DJ, Baron JA, et al. Serrated lesions of the colorectum: Review and recommendations from an expert panel. Am J Gastroenterol 2012;107(9):1315–30; quiz 1314, 1330.
122. Hogan RB III, Brill JV, Littenberg G, et al. Predict, resect, and discard … really? Gastrointest Endosc 2012;75(3):503–5.
123. Rex DK, Kahi C, O'Brien M, et al. The American Society for Gastrointestinal Endoscopy PIVI (preservation and incorporation of valuable endoscopic innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps. Gastrointest Endosc 2011;73(3):419–22.
124. Kaminski MF, Hassan C, Bisschops R, et al. Advanced imaging for detection and differentiation of colorectal neoplasia: European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Endoscopy 2014;46(5):435–49.
125. Alexander J, Gildea L, Balog J, et al. A novel methodology for in vivo endoscopic phenotyping of colorectal cancer based on real-time analysis of the mucosal lipidome: A prospective observational study of the iKnife. Surg Endosc 2017;31(3):1361–70.