Learning curve and competence for volumetric laser endomicroscopy in Barrett’s esophagus using cumulative sum analysis

• Abstract
• Introduction
• Methods
• Study setting
• Image set development
• Training
• Statistics
• Results
• Discussion
• References

Abstract

Background and study aims Little is known about the learning curve for image interpretation in volumetric laser endomicroscopy (VLE) in Barrett’s esophagus (BE). The goal of this study was to calculate the learning curve, competence of image interpretation, sensitivity, specificity, and accuracy of VLE among novice users.

Methods 31 novice users viewed 96 VLE images electronically at three academic institutions after a brief training session. There were 24 images of each histologic type: normal gastric cardia, normal esophageal squamous epithelium, non-neoplastic BE, and neoplastic BE. The users were asked to identify the correct tissue type and level of confidence. The cumulative summation (CUSUM) technique was used to construct a learning curve.

Results 22 (71 %) of the physicians achieved VLE interpretation competency during their 96-slide review. Half of the physicians achieved competency at 65 images (95 % confidence interval [CI] 45 – 85). There was a statistically significant association between confidence in diagnosis and selecting the correct histologic tissue type (P < 0.001). The median accuracy for esophageal squamous epithelium, normal gastric cardia, non-neoplastic BE, and neoplastic BE was 96 % (95 %CI 95 % – 96 %), 95 % (95 %CI 94 % – 96 %), 90 % (95 %CI 88 % – 91 %), 96 % (95 %CI 95 % – 96 %). The overall accuracy was 95 % (95 %CI 93 % – 95 %).

Conclusion The majority of novice users achieved competence in image interpretation of VLE for BE, using a pre-selected image set, with a favorable learning curve after a brief training session. An electronic review of VLE images, prior to real-time use of VLE, is encouraged.

Introduction

Barrett’s esophagus (BE) is a precursor lesion for esophageal adenocarcinoma. The incidence of esophageal adenocarcinoma continues to rise [1]. However, the vast majority of BE never develops into cancer. Patients with BE and neoplasia have a higher rate of progression to advanced disease than those without neoplasia. Thus, detecting early neoplasia in BE is vital given the evidence that treatment can decrease the risk of progression to advanced disease [2] [3]. Unfortunately, the current standard methods for detection of neoplasia – high definition endoscopy, and targeted plus random biopsies – are limited by sampling error [4] [5]. As a result, advanced imaging techniques have been developed to help target neoplasia in the esophagus for directed biopsies [6] [7].

A relatively new advanced imaging technology, volumetric laser endomicroscopy (VLE), has been developed to help target BE neoplasia [8] [9] [10]. VLE is a second-generation optical coherence tomography (OCT) device capable of real-time, high-resolution (axial resolution of 7 μm), wide-field, cross-sectional imaging (imaging depth of 3 mm). VLE can be used to scan a 6 cm length of the esophagus in approximately 90 seconds. In vivo and ex vivo scoring systems have been developed looking at OCT/VLE features of neoplasia [11] [12] [13]. These features are used in practice today to decide whether an area should be targeted for biopsy or resection. Thus, image interpretation of VLE features is essential in order to correctly target neoplasia. Several recent studies from expert centers suggest that VLE provides highly accurate detection of early neoplasia in BE [12] [13].

The learning curve for novice users to achieve competence in VLE image interpretation in BE is poorly understood. This information is essential to ensure uniform accuracy and to develop training programs. The aim of this study was to estimate the learning curve and competence of image interpretation of VLE among novice users using cumulative sum (CUSUM) analysis.

Methods

Study setting

Novice VLE users from three academic centers in the United States viewed 96 VLE still images electronically during a supervised session. The images were snapshots that consisted of magnified areas of a cross-sectional image. The centers included Lewis Katz School of Medicine at Temple University, Mayo Clinic Jacksonville, and Northwell Health System/Hofstra Northwell School of Medicine. The images were originally obtained from VLE examinations performed at these institutions using a high definition, adult, standard-sized gastroscope. The Institutional Review Board at each site approved the study.

Image set development

A total of 24 images of each histology type were shown: normal esophageal squamous epithelium, normal gastric cardia, BE without neoplasia (12 images of non-dysplastic BE and 12 images of BE with low grade dysplasia [LGD]), and BE with neoplasia (12 images of BE with high grade dysplasia [HGD] and 12 images of intramucosal cancer). Of the 24 images of BE with neoplasia, 8 had visible lesions on endoscopy that were targeted for endoscopic resection. None of the other groups had visible lesions. [Fig. 1] shows normal VLE images of the esophagus and gastric cardia. [Fig. 2] shows endoscopic, VLE, and the histology correlation for patients with BE, non-neoplastic BE (with LGD), and BE with HGD.

We included all four groups described above in the learning curve study, as all four groups need to be identified to successfully understand VLE interpretation in BE and to distinguish normal from abnormal. One may argue that a user only needs to identify BE with neoplasia from BE without neoplasia, and thus more of these images should be included in the dataset; however, there are subtleties that need to be appreciated to differentiate the different subtypes. For example, gastric cardia generally has a high surface reflectance or darker surface on VLE. BE with neoplasia can also have a high surface reflectance. Understanding the different features of gastric cardia and BE can help to distinguish between the two. Nevertheless, it should be noted that the dataset contained a relatively small subset of images in which the assessor had to assess the presence or absence of BE dysplasia.

The same image set had been previously reviewed and rated by high-volume users [14]. The high-volume users reviewed 120 images, 24 of which were discarded for the purpose of the current study owing to low agreement. Only images with high agreement (7 of 8 high-volume users agreed on the interpretation of the slide) were used in the study. This approach was taken because images with low agreement among expert users are not ideal to use in a learning curve study in novice users. Details of how the image sets were derived and the corresponding gold standard have been published previously [14].

VLE images for BE had corresponding histology supporting the morphologic tissue type from the site of the biopsy. This was considered the gold standard for images of BE. The endoscopic biopsies of the corresponding VLE image were obtained using previously described techniques [8] [9] [10]. Esophageal squamous mucosa and gastric cardia images had corresponding endoscopy showing normal mucosa. Images of the normal esophagus were taken well above any esophagus with BE. Biopsies were not taken at these levels to confirm tissue type as, endoscopically, these areas were visibly normal and thus tissue acquisition served no clinical benefit to the patient. Endoscopy was considered the gold standard for normal esophageal and gastric tissue types.

Training

Novice users were shown a brief training session followed by a rating image set. The purpose of the training session was to systematically teach the physicians the established imaging features of each tissue type for correct interpretation. The Poneros criteria [15] were used to distinguish BE from normal esophagus and gastric cardia. The OCT-SI or Evans criteria [11] were used to distinguish non-neoplastic BE (non-dysplastic Barrett’s esophagus or Barrett’s with LGD) from neoplastic BE (Barrett’s with HGD or intramucosal cancer). Each image had a corresponding gold standard as described above. The pathology diagnosis was unknown to the physician. The user was asked to identify the correct tissue type and provide a level of confidence for his/her rating. Images were viewed randomly, with an explanation of the correct answer after each rating. Each novice user independently reviewed each image and made a diagnosis (normal squamous epithelium, normal gastric cardia, non-neoplastic Barrett’s esophagus or neoplastic Barrett’s esophagus). Each image was marked as correct or incorrect based on the confirmed gold standard diagnosis.

Statistics

The sample size of novice users was based on feasibility and availability of resources. To determine a sufficient number of images to include, we used our predefined acceptable (10 %) and unacceptable (20 %) failure rates, and alpha and beta levels (0.1) to calculate the number of images needed for review. Using established methodology [16], the minimum number of images needed for CUSUM analysis was calculated to be 48. Therefore, our total of 96 images was sufficient to show competency.

The CUSUM technique [17] was used to construct a learning curve for each novice enrolled in the study. This technique is the accepted standard for determining learning curves and competence in gastrointestinal endoscopic procedures [18] [19] [20]. Our acceptable and unacceptable failure rates were prespecified as 10 % (p₀ = 0.1) and 20 % (p₁ = 0.2), respectively, and type I (α) and II (β) errors were both set at 0.1. The acceptable failure rate was determined based on the Preservation and Incorporation of Valuable Endoscopic Innovations criteria, which call for a 90 % sensitivity rate for advanced imaging techniques in BE to be accepted. The acceptable failure rate was confirmed as achievable based on the performance of the same slide set when viewed by high-volume VLE users.

The primary outcome variable was competency in reading VLE images. Competency was defined as the learning curve crossing the lower threshold, indicating an acceptable failure rate. A lack of competency was defined as the learning curve crossing the upper threshold, indicating an unacceptable failure rate. Undefined competency occurred when the learning curve did not cross either boundary, and indicated a need for further training.

The acceptable (h₀) and unacceptable (h₁) boundary lines were calculated to be – 2.71 and + 2.71, respectively, based on the formulae presented in the paper by Bolsin and Colson [17]. The image number at which each novice first crossed the lower threshold, indicating competency and an acceptable failure rate, was examined using the Kaplan – Meier estimator, and compared between groups of interest using the log-rank test. Physicians who did not achieve competency were censored at the last slide reviewed.

Generalized linear mixed models (GLMM) with a cumulative logit link for clustered ordinal data was used to model certainty of slide diagnosis (confident in my choice, fairly certain or uncertain) as a function of groups of interest. GLMM was used to account for the hierarchical structure of the data – namely, multiple slides within a physician. Specifically, each physician (n = 31) reviewed 96 slides, resulting in a total of 2976 slides. GLMM for clustered binary data was used to model correct diagnosis on a slide (yes/no) as a function of time spent reviewing each image and the degree of confidence. A linear mixed model was used to model the time spent reviewing each image as a function of slide number. The outcome of time was log-transformed in order to meet the standard assumptions of Gaussian residuals and equality of variance. As the physicians represent a sample from a larger population, physician was included in all models as a random effect and the predictors in each model were included as fixed effects. Random effects were modeled on the G-side to provide subject-specific interpretations and therefore the correlations were modeled indirectly. All analyses were conducted using SAS version 9.4 (Cary, North Carolina, United States).

Results

A total of 31 novice VLE users (25 male) completed the session. The majority of physicians were gastroenterology fellows in training (n = 21; 68 %) and 32 % (n = 10) were attending gastroenterology physicians. The median age of the attending physicians was 52 years. The median years since graduation from medical school for attending physicians was 19 years. The median age of the fellows was 31 years. Five physicians (16 %) had endoscopic ultrasound (EUS) experience and four (13 %) had limited VLE experience (exposure to fewer than 10 real-time VLE cases). Almost half of the users were from Hofstra Northwell School of Medicine (n = 14; 45 %), 32 % (n = 10) were from Temple University, and 23 % (n = 7) were from Mayo Clinic.

The overall average time to complete the total session was 59 minutes (range 33 – 85 minutes). The average time to complete the session for attending physicians was 71 minutes (range 36 – 106 minutes) compared with 53 minutes (range 34 – 72 minutes) for fellows (P = 0.07). The overall learning curves are presented in the [Fig. 3]. A total of 22 physicians (71 %; 95 %CI 52 % – 86 %) achieved competency during their 96-slide review and 9 (29 %; 95 %CI 14 % – 48 %) demonstrated an overall lack of competence. Half of the physicians achieved competency at 65 images (95 %CI 45 – 85 images), while 55 % achieved competence at 65 slides, 60 % at 72 slides, 65 % at 85 slides, and 70 % at 92 slides. These percentages were derived from a Kaplan – Meier curve ([Supplemental Fig. e 4], available online). [Fig. 5] shows the learning curves for fellows and attending physicians.

Among the 2976 images reviewed (96 images were reviewed by 31 physicians), physicians were confident in their diagnosis on 50 % of slides, fairly certain in their diagnosis on 43 %, and uncertain in their diagnosis on 7 %. Images in which the physician was confident or fairly certain in their diagnosis had statistically significantly greater odds of being correct (P < 0.001). Specifically, images with a confident or fairly certain diagnosis were correct 91 % of the time (95 %CI 89 % – 92 %) and images with an uncertain diagnosis were correct 60 % of the time (95 %CI 50 % – 69 %). There was a statistically significant association between confidence in diagnosis and histologic tissue type (P  < 0.001). Non-neoplastic BE slides had statistically significant greater odds of uncertainty compared with BE with neoplasia (odds ratio [OR] 3.17, 95 %CI 2.52 – 3.99), normal gastric cardia (OR 3.87, 95 %CI 3.06 – 4.90), and normal squamous epithelium (OR 2.61, 95 %CI 2.08 – 3.27). The likelihood of correct diagnoses were inversely associated with time spent reviewing the image (P  < 0.001). With each additional second spent reviewing the slide, the likelihood of a correct diagnosis decreased by 5 % (relative risk 0.95, 95 %CI 0.94 – 0.96). For example, slides diagnosed at 10 seconds were 91 % (on average) correct, slides diagnosed at 20 seconds were 86 % correct, and slides diagnosed at 30 seconds were 78 % correct. The time per slide also became faster throughout the image set. With each additional 10 slides reviewed, the time spent decreased by 7 % (P < 0.001).

There were no statistically significant differences with respect to the number of images viewed until competency between: attending physicians and fellows (P = 0.24); EUS experience and no EUS experience (P = 0.84); institution (P = 0.28); or limited VLE experience and no VLE experience (P = 0.94). There were no statistically significant associations in certainty of image diagnosis: between attending physicians and fellows (P = 0.88; OR 0.90, 95 %CI 0.24 – 3.33); EUS experience (P = 0.55; OR 1.67, 95 %CI 0.32 – 8.33); or prior limited VLE experience (P = 0.25; OR 2.86, 95 %CI 0.48 – 17.03). It should be noted though that sample size was not powered for these secondary outcomes.

The overall accuracy, sensitivity, and specificity of a VLE diagnosis for the novice users are presented in [Table 1]. The median accuracy for esophageal squamous epithelium, gastric epithelium, non-neoplastic BE, and neoplastic BE was 96 % (95 %CI 95 % – 96 %), 95 % (95 %CI 94 % – 96 %), 90 % (95 %CI 88 % – 91 %), and 96 % (95 %CI 95 % – 96 %). The overall accuracy was 95 % (95 %CI 93 % – 95 %). [Fig. 6] shows the receiver operating characteristic curve analyses. All curves for all tissue types fall above the no-discrimination line and therefore represent good classification of results.

Discussion

In this study, we found that the majority of novice users achieved competence similar to high-frequency users in interpretation of VLE images from a pre-selected image set after a brief online training session. Although the majority of users achieved competence, there was a high degree of variability with regard to number of images reviewed to reach competence, although all were within the range of a short 1 – 2 hour training and testing session.

This result is consistent with other advanced imaging technologies showing that competence and a rapid learning curve can be achieved in less-experienced or novice users after a computer training simulation. Studies with narrow-band imaging (NBI) have shown a rapid learning curve for imaging in BE, neoplastic gastric lesions, and colon polyps [21] [22] [23] [24] [25] [26] [27] [28]. These studies are similar to ours in that less-experienced endoscopists were given multiple NBI images to characterize after a brief training program. Less-experienced endoscopists approached the accuracy of experts in these NBI studies. Similar results have also been achieved for learning curve studies of confocal laser endomicroscopy in BE, esophageal squamous cancer, and classification of colorectal polyps [29] [30] [31] [32].

The degree of confidence was statistically associated with classifying an image correctly. This is an expected finding for a learning curve study [24]. When analyzed by tissue type, novice users were most uncertain about interpretation of BE without neoplasia. Similarly, this histology subgroup also had the lowest level of agreement among high-frequency users [14]. This study was not designed to evaluate the specifics of why this might be the case, but further research is needed to better understand this outcome. Importantly, novice users, like high-frequency users, are able to correctly classify BE with neoplasia. In fact the VLE accuracy for each tissue type was similar to high-volume users [14], although the novice user image set excluded 24 images that had lower agreement among the experts. These accuracies are high even when including images from early in the learning curve study.

It should be noted that the skill set to obtain VLE images endoscopically in the tubular esophagus is separate from image interpretation. This study did not assess the learning curve to obtain images endoscopically. In our experience, obtaining the images involves a short learning process, as it mainly entails inflating a balloon that contains the imaging probe, and centering it. The most challenging component is image interpretation. Based on the current study, it would be reasonable for training programs teaching VLE to first require novice users to undergo an online training program similar to that used in the current study. Thus, competence in VLE image interpretation could be defined prior to performing endoscopic VLE procedures in real time on patients.

Our study does have limitations. The images from the training program were chosen and were not from consecutive patients. However, there were limited images to choose from for BE and BE with neoplasia from the database of images, and thus we do not feel that this impacts the generalizability of our study. The three sites had a combined total of 35 images of BE with neoplasia and 40 without neoplasia, with corresponding histology (24 slides chosen for each group). It should be noted that when the slide set was shown to high-volume users, the image set was 120 images. The image set was reduced to 96 to exclude images with low agreement, as it did not seem appropriate to use these to study the learning curve in novice users. By excluding these images, we could be underestimating the number of images needed to achieve competency in image interpretation, and overestimating agreement in image interpretation among novice users.

Another limitation is that this study provides a snapshot of VLE interpretation at one point in time. It does not evaluate whether this training session has a durable or sustained effect. It is also unclear whether VLE image interpretation from a pre-selected image set during a training session correlates with real-time image interpretation. In real-time imaging, a user reviews 1200 cross-sectional images for one scan in real time, and has to interpret neoplasia from these images. The dataset from this study is different from real-time imaging in that the subjects are interpreting the pre-selected snapshots (magnified areas of a cross-sectional image). This needs to be evaluated in future studies.

Finally, we used an arbitrary failure rate of 10 % as the cutoff for what was acceptable in this study. Although we deemed this to be stringent, there are no defined standards for an acceptable miss rate for VLE interpretation. Although interpretation of 90 % of images may be acceptable and stringent for BE with neoplasia, it may not be acceptable for determining normal esophagus. However, using VLE accurately in BE requires correct interpretation of all four tissue types, and thus all four tissues types are built into the learning curve model for VLE image interpretation of BE. Finally, the results should be interpreted with caution in terms of distinguishing neoplasia from non-neoplasia, as only 25 % of images were of neoplastic BE. This number is too small to apply a separate CUSUM analysis learning curve to assess for competency in this subgroup

In conclusion, this study shows that the majority of novice users of VLE can achieve competence similar to high-frequency users in VLE image interpretation in BE, for a pre-selected data set of images, after an online training session. This study supports the use of an online tool for training of VLE image interpretation prior to real-time use for novice users.

Competing interests

K.J.C., C.J.L., C.L.L., M.S.S., and M.B.W. have received research funding from Ninepoint Medical. D.K.P. and M.S.S. are consultants for Ninepoint Medical. G.J.T. (Massachusetts General Hospital) has a licensing arrangement with NinePoint Medical, and has the right to receive royalties from this licensing arrangement.G.J.T. also receives grants from the Bill and Melinda Gates Foundation and iLumen Medical.

Acknowledgment

The authors would like to thank the following people at Ninepoint Medical who helped develop the database of volumetric laser endomicroscopy images in conjunction with the authors: Amna Soomro, Anny Fonseca, Brian Sundell, and Simon Schlachter.

• References

• 1 Pohl H, Welch HG. The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence. J Natl Cancer Inst 2005; 97: 142-146
  CrossrefPubMedSearch in Google Scholar
• 2 Shaheen NJ, Sharma P, Overholt BF. et al. Radiofrequency ablation in Barrett’s esophagus with dysplasia. N Engl J Med 2009; 360: 2277-2288
  CrossrefPubMedSearch in Google Scholar
• 3 Phoa KN, Vilsteren FGIvan, Weusten BLAM. et al. Radiofrequency ablation vs endoscopic surveillance for patients with Barrett esophagus and low-grade dysplasia: a randomized clinical trial. JAMA 2014; 311: 1209-1217
  CrossrefPubMedSearch in Google Scholar
• 4 Sharma P, Brill J, Canto M. et al. White Paper AGA: Advanced Imaging in Barrett’s Esophagus. Clin Gastroenterol Hepatol 2015; 13: 2209-2218
  CrossrefPubMedSearch in Google Scholar
• 5 Sharma P, Falk GW, Weston AP. et al. Dysplasia and cancer in a large multicenter cohort of patients with Barrett’s esophagus. Clin Gastroenterol Hepatol 2006; 4: 566-572
  CrossrefPubMedSearch in Google Scholar
• 6 Thosani N, Abu DayyehBK. ASGE Technology Committee. 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 2016; 83: 684-698
  CrossrefPubMedSearch in Google Scholar
• 7 Qumseya BJ, Wang H, Badie N. et al. Advanced imaging technologies increase detection of dysplasia and neoplasia in patients with Barrett’s esophagus: a meta-analysis and systematic review. Clin Gastroenterol Hepatol 2013; 11: 1562-1570
  CrossrefPubMedSearch in Google Scholar
• 8 Trindade AJ, George BJ, Berkowitz J. et al. Volumetric laser endomicroscopy can target neoplasia not detected by conventional endoscopic measures in long segment Barrett’s esophagus. Endosc Int Open 2016; 4: E318-322
  Thieme ConnectPubMedSearch in Google Scholar
• 9 Trindade AJ, Smith MS, Pleskow DK. The new kid on the block for advanced imaging in Barrett’s esophagus: a review of volumetric laser endomicroscopy. Therap Adv Gastroenterol 2016; 9: 408-416
  CrossrefPubMedSearch in Google Scholar
• 10 Wolfsen HC, Sharma P, Wallace MB. et al. Safety and feasibility of volumetric laser endomicroscopy in patients with Barrett’s esophagus (with videos). Gastrointest Endosc 2015; 82: 631-640
  CrossrefPubMedSearch in Google Scholar
• 11 Evans JA, Poneros JM, Bouma BE. et al. Optical coherence tomography to identify intramucosal carcinoma and high-grade dysplasia in Barrett’s esophagus. Clin Gastroenterol Hepatol 2006; 4: 38-43
  CrossrefPubMedSearch in Google Scholar
• 12 Swager A-F, Tearney GJ, Leggett CL. et al. Identification of volumetric laser endomicroscopy features predictive for early neoplasia in Barrett’s esophagus using high-quality histological correlation. Gastrointest Endosc 2017; 85: 918-926
  CrossrefPubMedSearch in Google Scholar
• 13 Leggett CL, Gorospe EC, Chan DK. 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 2016; 83: 880-888
  CrossrefPubMedSearch in Google Scholar
• 14 Trindade AJ, Inamdar S, Smith MS. et al. Volumetric laser endomicroscopy in Barrett’s esophagus: interobserver agreement for interpretation of Barrett’s esophagus and associated neoplasia among high-frequency users. Gastrointest Endosc 2017; 86: 133-139
  CrossrefPubMedSearch in Google Scholar
• 15 Poneros JM, Brand S, Bouma BE. et al. Diagnosis of specialized intestinal metaplasia by optical coherence tomography. Gastroenterology 2001; 120: 7-12
  CrossrefPubMedSearch in Google Scholar
• 16 Correa JB, Dellazzana JE, Sturm A. et al. Using the Cusum curve to evaluate the training of orotracheal intubation with the Truview EVO2 laryngoscope. Rev Bras Anestesiol 2009; 59: 321-331
  CrossrefPubMedSearch in Google Scholar
• 17 Bolsin S, Colson M. The use of the Cusum technique in the assessment of trainee competence in new procedures. Int J Qual Heal Care J 2000; 12: 433-438
  PubMedSearch in Google Scholar
• 18 Wani S, Coté GA, Keswani R. et al. Learning curves for EUS by using cumulative sum analysis: implications for American Society for Gastrointestinal Endoscopy recommendations for training. Gastrointest Endosc 2013; 77: 558-565
  CrossrefPubMedSearch in Google Scholar
• 19 Wani S, Hall M, Wang AY. et al. Variation in learning curves and competence for ERCP among advanced endoscopy trainees by using cumulative sum analysis. Gastrointest Endosc 2016; 83: 711-719
  CrossrefPubMedSearch in Google Scholar
• 20 Wani S, Hall M, Keswani RN. et al. Variation in aptitude of trainees in endoscopic ultrasonography, based on cumulative sum analysis. Clin Gastroenterol Hepatol 2015; 13: 1318-1325
  CrossrefPubMedSearch in Google Scholar
• 21 Higashi R, Uraoka T, Kato J. et al. Diagnostic accuracy of narrow-band imaging and pit pattern analysis significantly improved for less-experienced endoscopists after an expanded training program. Gastrointest Endosc 2010; 72: 127-135
  CrossrefPubMedSearch in Google Scholar
• 22 Raghavendra M, Hewett DG, Rex DK. Differentiating adenomas from hyperplastic colorectal polyps: narrow-band imaging can be learned in 20 minutes. Gastrointest Endosc 2010; 72: 572-576
  CrossrefPubMedSearch in Google Scholar
• 23 Dias-Silva D, Pimentel-Nunes P, Magalhães J. et al. The learning curve for narrow-band imaging in the diagnosis of precancerous gastric lesions by using Web-based video. Gastrointest Endosc 2014; 79: 910-920
  CrossrefPubMedSearch in Google Scholar
• 24 Patel SG, Rastogi A, Austin G. et al. Gastroenterology trainees can easily learn histologic characterization of diminutive colorectal polyps with narrow band imaging. Clin Gastroenterol Hepatol 2013; 11: 997-1003
  CrossrefPubMedSearch in Google Scholar
• 25 Rastogi A, Rao DS, Gupta N. et al. Impact of a computer-based teaching module on characterization of diminutive colon polyps by using narrow-band imaging by non-experts in academic and community practice: a video-based study. Gastrointest Endosc 2014; 79: 390-398
  CrossrefPubMedSearch in Google Scholar
• 26 Rastogi A, Pondugula K, Bansal A. et al. Recognition of surface mucosal and vascular patterns of colon polyps by using narrow-band imaging: interobserver and intraobserver agreement and prediction of polyp histology. Gastrointest Endosc 2009; 69: 716-722
  CrossrefPubMedSearch in Google Scholar
• 27 Daly C, Vennalaganti P, Soudagar S. et al. Randomized controlled trial of self-directed versus in-classroom teaching of narrow-band imaging for diagnosis of Barrett’s esophagus-associated neoplasia. Gastrointest Endosc 2016; 83: 101-106
  CrossrefPubMedSearch in Google Scholar
• 28 Alvarez HerreroL, Curvers WL, Bansal A. et al. Zooming in on Barrett oesophagus using narrow-band imaging: an international observer agreement study. Eur J Gastroenterol Hepatol 2009; 21: 1068-1075
  CrossrefPubMedSearch in Google Scholar
• 29 Rzouq F, Vennalaganti P, Pakseresht K. et al. In-class didactic versus self-directed teaching of the probe-based confocal laser endomicroscopy (pCLE) criteria for Barrett’s esophagus. Endoscopy 2016; 48: 123-127
  Thieme ConnectPubMedSearch in Google Scholar
• 30 Buchner AM, Gomez V, Heckman MG. et al. The learning curve of in vivo probe-based confocal laser endomicroscopy for prediction of colorectal neoplasia. Gastrointest Endosc 2011; 73: 556-560
  CrossrefPubMedSearch in Google Scholar
• 31 Liu J, Li M, Li Z. et al. Learning curve and interobserver agreement of confocal laser endomicroscopy for detecting precancerous or early-stage esophageal squamous cancer. PLoS One 2014; 9: e99089
  CrossrefPubMedSearch in Google Scholar
• 32 Kuiper T, Kiesslich R, Ponsioen C. et al. The learning curve, accuracy, and interobserver agreement of endoscope-based confocal laser endomicroscopy for the differentiation of colorectal lesions. Gastrointest Endosc 2012; 75: 1211-1217
  CrossrefPubMedSearch in Google Scholar

Corresponding author

• References

• 1 Pohl H, Welch HG. The role of overdiagnosis and reclassification in the marked increase of esophageal adenocarcinoma incidence. J Natl Cancer Inst 2005; 97: 142-146
  CrossrefPubMedSearch in Google Scholar
• 2 Shaheen NJ, Sharma P, Overholt BF. et al. Radiofrequency ablation in Barrett’s esophagus with dysplasia. N Engl J Med 2009; 360: 2277-2288
  CrossrefPubMedSearch in Google Scholar
• 3 Phoa KN, Vilsteren FGIvan, Weusten BLAM. et al. Radiofrequency ablation vs endoscopic surveillance for patients with Barrett esophagus and low-grade dysplasia: a randomized clinical trial. JAMA 2014; 311: 1209-1217
  CrossrefPubMedSearch in Google Scholar
• 4 Sharma P, Brill J, Canto M. et al. White Paper AGA: Advanced Imaging in Barrett’s Esophagus. Clin Gastroenterol Hepatol 2015; 13: 2209-2218
  CrossrefPubMedSearch in Google Scholar
• 5 Sharma P, Falk GW, Weston AP. et al. Dysplasia and cancer in a large multicenter cohort of patients with Barrett’s esophagus. Clin Gastroenterol Hepatol 2006; 4: 566-572
  CrossrefPubMedSearch in Google Scholar
• 6 Thosani N, Abu DayyehBK. ASGE Technology Committee. 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 2016; 83: 684-698
  CrossrefPubMedSearch in Google Scholar
• 7 Qumseya BJ, Wang H, Badie N. et al. Advanced imaging technologies increase detection of dysplasia and neoplasia in patients with Barrett’s esophagus: a meta-analysis and systematic review. Clin Gastroenterol Hepatol 2013; 11: 1562-1570
  CrossrefPubMedSearch in Google Scholar
• 8 Trindade AJ, George BJ, Berkowitz J. et al. Volumetric laser endomicroscopy can target neoplasia not detected by conventional endoscopic measures in long segment Barrett’s esophagus. Endosc Int Open 2016; 4: E318-322
  Thieme ConnectPubMedSearch in Google Scholar
• 9 Trindade AJ, Smith MS, Pleskow DK. The new kid on the block for advanced imaging in Barrett’s esophagus: a review of volumetric laser endomicroscopy. Therap Adv Gastroenterol 2016; 9: 408-416
  CrossrefPubMedSearch in Google Scholar
• 10 Wolfsen HC, Sharma P, Wallace MB. et al. Safety and feasibility of volumetric laser endomicroscopy in patients with Barrett’s esophagus (with videos). Gastrointest Endosc 2015; 82: 631-640
  CrossrefPubMedSearch in Google Scholar
• 11 Evans JA, Poneros JM, Bouma BE. et al. Optical coherence tomography to identify intramucosal carcinoma and high-grade dysplasia in Barrett’s esophagus. Clin Gastroenterol Hepatol 2006; 4: 38-43
  CrossrefPubMedSearch in Google Scholar
• 12 Swager A-F, Tearney GJ, Leggett CL. et al. Identification of volumetric laser endomicroscopy features predictive for early neoplasia in Barrett’s esophagus using high-quality histological correlation. Gastrointest Endosc 2017; 85: 918-926
  CrossrefPubMedSearch in Google Scholar
• 13 Leggett CL, Gorospe EC, Chan DK. 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 2016; 83: 880-888
  CrossrefPubMedSearch in Google Scholar
• 14 Trindade AJ, Inamdar S, Smith MS. et al. Volumetric laser endomicroscopy in Barrett’s esophagus: interobserver agreement for interpretation of Barrett’s esophagus and associated neoplasia among high-frequency users. Gastrointest Endosc 2017; 86: 133-139
  CrossrefPubMedSearch in Google Scholar
• 15 Poneros JM, Brand S, Bouma BE. et al. Diagnosis of specialized intestinal metaplasia by optical coherence tomography. Gastroenterology 2001; 120: 7-12
  CrossrefPubMedSearch in Google Scholar
• 16 Correa JB, Dellazzana JE, Sturm A. et al. Using the Cusum curve to evaluate the training of orotracheal intubation with the Truview EVO2 laryngoscope. Rev Bras Anestesiol 2009; 59: 321-331
  CrossrefPubMedSearch in Google Scholar
• 17 Bolsin S, Colson M. The use of the Cusum technique in the assessment of trainee competence in new procedures. Int J Qual Heal Care J 2000; 12: 433-438
  PubMedSearch in Google Scholar
• 18 Wani S, Coté GA, Keswani R. et al. Learning curves for EUS by using cumulative sum analysis: implications for American Society for Gastrointestinal Endoscopy recommendations for training. Gastrointest Endosc 2013; 77: 558-565
  CrossrefPubMedSearch in Google Scholar
• 19 Wani S, Hall M, Wang AY. et al. Variation in learning curves and competence for ERCP among advanced endoscopy trainees by using cumulative sum analysis. Gastrointest Endosc 2016; 83: 711-719
  CrossrefPubMedSearch in Google Scholar
• 20 Wani S, Hall M, Keswani RN. et al. Variation in aptitude of trainees in endoscopic ultrasonography, based on cumulative sum analysis. Clin Gastroenterol Hepatol 2015; 13: 1318-1325
  CrossrefPubMedSearch in Google Scholar
• 21 Higashi R, Uraoka T, Kato J. et al. Diagnostic accuracy of narrow-band imaging and pit pattern analysis significantly improved for less-experienced endoscopists after an expanded training program. Gastrointest Endosc 2010; 72: 127-135
  CrossrefPubMedSearch in Google Scholar
• 22 Raghavendra M, Hewett DG, Rex DK. Differentiating adenomas from hyperplastic colorectal polyps: narrow-band imaging can be learned in 20 minutes. Gastrointest Endosc 2010; 72: 572-576
  CrossrefPubMedSearch in Google Scholar
• 23 Dias-Silva D, Pimentel-Nunes P, Magalhães J. et al. The learning curve for narrow-band imaging in the diagnosis of precancerous gastric lesions by using Web-based video. Gastrointest Endosc 2014; 79: 910-920
  CrossrefPubMedSearch in Google Scholar
• 24 Patel SG, Rastogi A, Austin G. et al. Gastroenterology trainees can easily learn histologic characterization of diminutive colorectal polyps with narrow band imaging. Clin Gastroenterol Hepatol 2013; 11: 997-1003
  CrossrefPubMedSearch in Google Scholar
• 25 Rastogi A, Rao DS, Gupta N. et al. Impact of a computer-based teaching module on characterization of diminutive colon polyps by using narrow-band imaging by non-experts in academic and community practice: a video-based study. Gastrointest Endosc 2014; 79: 390-398
  CrossrefPubMedSearch in Google Scholar
• 26 Rastogi A, Pondugula K, Bansal A. et al. Recognition of surface mucosal and vascular patterns of colon polyps by using narrow-band imaging: interobserver and intraobserver agreement and prediction of polyp histology. Gastrointest Endosc 2009; 69: 716-722
  CrossrefPubMedSearch in Google Scholar
• 27 Daly C, Vennalaganti P, Soudagar S. et al. Randomized controlled trial of self-directed versus in-classroom teaching of narrow-band imaging for diagnosis of Barrett’s esophagus-associated neoplasia. Gastrointest Endosc 2016; 83: 101-106
  CrossrefPubMedSearch in Google Scholar
• 28 Alvarez HerreroL, Curvers WL, Bansal A. et al. Zooming in on Barrett oesophagus using narrow-band imaging: an international observer agreement study. Eur J Gastroenterol Hepatol 2009; 21: 1068-1075
  CrossrefPubMedSearch in Google Scholar
• 29 Rzouq F, Vennalaganti P, Pakseresht K. et al. In-class didactic versus self-directed teaching of the probe-based confocal laser endomicroscopy (pCLE) criteria for Barrett’s esophagus. Endoscopy 2016; 48: 123-127
  Thieme ConnectPubMedSearch in Google Scholar
• 30 Buchner AM, Gomez V, Heckman MG. et al. The learning curve of in vivo probe-based confocal laser endomicroscopy for prediction of colorectal neoplasia. Gastrointest Endosc 2011; 73: 556-560
  CrossrefPubMedSearch in Google Scholar
• 31 Liu J, Li M, Li Z. et al. Learning curve and interobserver agreement of confocal laser endomicroscopy for detecting precancerous or early-stage esophageal squamous cancer. PLoS One 2014; 9: e99089
  CrossrefPubMedSearch in Google Scholar
• 32 Kuiper T, Kiesslich R, Ponsioen C. et al. The learning curve, accuracy, and interobserver agreement of endoscope-based confocal laser endomicroscopy for the differentiation of colorectal lesions. Gastrointest Endosc 2012; 75: 1211-1217
  CrossrefPubMedSearch in Google Scholar

• Supplementary Material