Narrow-band imaging (NBI) magnifying endoscopic classification of colorectal tumors proposed by the Japan NBI Expert Team

Many clinical studies on narrow-band imaging (NBI) magnifying endoscopy classifications advocated so far in Japan (Sano, Hiroshima, Showa, and Jikei classifications) have reported the usefulness of NBI magnifying endoscopy for qualitative and quantitative diagnosis of colorectal lesions. However, discussions at professional meetings have raised issues such as: (i) the presence of multiple terms for the same or similar findings; (ii) the necessity of including surface patterns in magnifying endoscopic classifications; and (iii) differences in the NBI findings in elevated and superficial lesions. To resolve these problems, the Japan NBI Expert Team (JNET) was constituted with the aim of establishing a universal NBI magnifying endoscopic classification for colorectal tumors (JNET classification) in 2011. Consensus was reached on this classification using the modified Delphi method, and this classification was proposed in June 2014. The JNET classification consists of four categories of vessel and surface pattern (i.e. Types 1, 2A, 2B, and 3). Types 1, 2A, 2B, and 3 are correlated with the histopathological findings of hyperplastic polyp/sessile serrated polyp (SSP), low-grade intramucosal neoplasia, high-grade intramucosal neoplasia/shallow submucosal invasive cancer, and deep submucosal invasive cancer, respectively.

Real-Time Use of Artificial Intelligence in Identification of Diminutive Polyps During Colonoscopy: A Prospective Study

BACKGROUND

Computer-aided diagnosis (CAD) for colonoscopy may help endoscopists distinguish neoplastic polyps (adenomas) requiring resection from nonneoplastic polyps not requiring resection, potentially reducing cost.

OBJECTIVE

To evaluate the performance of real-time CAD with endocytoscopes (×520 ultramagnifying colonoscopes providing microvascular and cellular visualization of colorectal polyps after application of the narrow-band imaging [NBI] and methylene blue staining modes, respectively).

DESIGN

Single-group, open-label, prospective study. (UMIN [University hospital Medical Information Network] Clinical Trial Registry: UMIN000027360).

SETTING

University hospital.

PARTICIPANTS

791 consecutive patients undergoing colonoscopy and 23 endoscopists.

INTERVENTION

Real-time use of CAD during colonoscopy.

MEASUREMENTS

CAD-predicted pathology (neoplastic or nonneoplastic) of detected diminutive polyps (≤5 mm) on the basis of real-time outputs compared with pathologic diagnosis of the resected specimen (gold standard). The primary end point was whether CAD with the stained mode produced a negative predictive value (NPV) of 90% or greater for identifying diminutive rectosigmoid adenomas, the threshold required to "diagnose-and-leave" nonneoplastic polyps. Best- and worst-case scenarios assumed that polyps lacking either CAD diagnosis or pathology were true- or false-positive or true- or false-negative, respectively.

RESULTS

Overall, 466 diminutive (including 250 rectosigmoid) polyps from 325 patients were assessed by CAD, with a pathologic prediction rate of 98.1% (457 of 466). The NPVs of CAD for diminutive rectosigmoid adenomas were 96.4% (95% CI, 91.8% to 98.8%) (best-case scenario) and 93.7% (CI, 88.3% to 97.1%) (worst-case scenario) with stained mode and 96.5% (CI, 92.1% to 98.9%) (best-case scenario) and 95.2% (CI, 90.3% to 98.0%) (worst-case scenario) with NBI.

LIMITATION

Two thirds of the colonoscopies were conducted by experts who had each experienced more than 200 endocytoscopies; 186 polyps not assessed by CAD were excluded.

CONCLUSION

Real-time CAD can achieve the performance level required for a diagnose-and-leave strategy for diminutive, nonneoplastic rectosigmoid polyps.

PRIMARY FUNDING SOURCE

Japan Society for the Promotion of Science.

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

In vivo real-time assessment of the histology of diminutive (≤5 mm) colorectal polyps detected at colonoscopy can be achieved by means of an "optical biopsy" by using currently available endoscopic technologies. This systematic review and meta-analysis was performed by the American Society for Gastrointestinal Endoscopy (ASGE) Technology Committee to specifically assess whether acceptable performance thresholds outlined by an ASGE Preservation and Incorporation of Valuable endoscopic Innovations (PIVI) document for clinical adoption of these technologies have been met. We conducted direct meta-analyses calculating the pooled negative predictive value (NPV) for narrow-band imaging (NBI), i-SCAN, and Fujinon Intelligent Color Enhancement (FICE)-assisted optical biopsy for predicting adenomatous polyp histology of small/diminutive colorectal polyps. We also calculated the pooled percentage agreement with histopathology when assigning postpolypectomy surveillance intervals based on combining real-time optical biopsy of colorectal polyps 5 mm or smaller with histopathologic assessment of polyps larger than 5 mm. Random-effects meta-analysis models were used. Statistical heterogeneity was evaluated by means of I(2) statistics. Our meta-analyses indicate that optical biopsy with NBI, exceeds the NPV threshold for adenomatous polyp histology, supporting a "diagnose-and-leave" strategy for diminutive predicted nonneoplastic polyps in the rectosigmoid colon. The pooled NPV of NBI for adenomatous polyp histology by using the random-effects model was 91% (95% confidence interval [CI], 88-94). This finding was associated with a high degree of heterogeneity (I(2) = 89%). Subgroup analysis indicated that the pooled NPV was greater than 90% for academic medical centers (91.8%; 95% CI, 89-94), for experts (93%; 95% CI, 91-96), and when the optical biopsy assessment was made with high confidence (93%; 95% CI, 90-96). Our meta-analyses also indicate that the agreement in assignment of postpolypectomy surveillance intervals based on optical biopsy with NBI of diminutive colorectal polyps is 90% or greater in academic settings (91%; 95% CI, 86-95), with experienced endoscopists (92%; 95% CI, 88-96) and when optical biopsy assessments are made with high confidence (91%; 95% CI, 88-95). Our systematic review and meta-analysis confirms that the thresholds established by the ASGE PIVI for real-time endoscopic assessment of the histology of diminutive polyps have been met, at least with NBI optical biopsy, with endoscopists who are expert in using this advanced imaging technology and when assessments are made with high confidence.

A multicenter prospective study of the real-time use of narrow-band imaging in the diagnosis of premalignant gastric conditions and lesions

BACKGROUND AND AIM

Some studies suggest that narrow-band imaging (NBI) can be more accurate at diagnosing gastric intestinal metaplasia and dysplasia than white-light endoscopy (WLE) alone. We aimed to assess the real-time diagnostic validity of high resolution endoscopy with and without NBI in the diagnosis of gastric premalignant conditions and to derive a classification for endoscopic grading of gastric intestinal metaplasia (EGGIM).

METHODS

A multicenter prospective study (five centers: Portugal, Italy, Romania, UK, USA) was performed involving the systematic use of high resolution gastroscopes with image registry with and without NBI in a centralized informatics platform (available online). All users used the same NBI classification. Histologic result was considered the diagnostic gold standard.

RESULTS

A total of 238 patients and 1123 endoscopic biopsies were included. NBI globally increased diagnostic accuracy by 11 percentage points (NBI 94 % vs. WLE 83 %; P < 0.001) with no difference in the identification of Helicobacter pylori gastritis (73 % vs. 74 %). NBI increased sensitivity for the diagnosis of intestinal metaplasia significantly (87 % vs. 53 %; P < 0.001) and for the diagnosis of dysplasia (92 % vs. 74 %). The added benefit of NBI in terms of diagnostic accuracy was greater in OLGIM III/IV than in OLGIM I/II (25 percentage points vs. 15 percentage points, respectively; P < 0.001). The area under the curve (AUC) of the receiver operating characteristic (ROC) curve for EGGIM in the identification of extensive metaplasia was 0.98.

CONCLUSIONS

In a real-time scenario, NBI demonstrates a high concordance with gastric histology, superior to WLE. Diagnostic accuracy higher than 90 % suggests that routine use of NBI allows targeted instead of random biopsy samples. EGGIM also permits immediate grading of intestinal metaplasia without biopsies and merits further investigation.

Real-time automated diagnosis of precancerous lesions and early esophageal squamous cell carcinoma using a deep learning model (with videos)

BACKGROUND AND AIMS

We developed a system for computer-assisted diagnosis (CAD) for real-time automated diagnosis of precancerous lesions and early esophageal squamous cell carcinomas (ESCCs) to assist the diagnosis of esophageal cancer.

METHODS

A total of 6473 narrow-band imaging (NBI) images, including precancerous lesions, early ESCCs, and noncancerous lesions, were used to train the CAD system. We validated the CAD system using both endoscopic images and video datasets. The receiver operating characteristic curve of the CAD system was generated based on image datasets. An artificial intelligence probability heat map was generated for each input of endoscopic images. The yellow color indicated high possibility of cancerous lesion, and the blue color indicated noncancerous lesions on the probability heat map. When the CAD system detected any precancerous lesion or early ESCCs, the lesion of interest was masked with color.

RESULTS

The image datasets contained 1480 malignant NBI images from 59 consecutive cancerous cases (sensitivity, 98.04%) and 5191 noncancerous NBI images from 2004 cases (specificity, 95.03%). The area under curve was 0.989. The video datasets of precancerous lesions or early ESCCs included 27 nonmagnifying videos (per-frame sensitivity 60.8%, per-lesion sensitivity, 100%) and 20 magnifying videos (per-frame sensitivity 96.1%, per-lesion sensitivity, 100%). Unaltered full-range normal esophagus videos included 33 videos (per-frame specificity 99.9%, per-case specificity, 90.9%).

CONCLUSIONS

A deep learning model demonstrated high sensitivity and specificity for both endoscopic images and video datasets. The real-time CAD system has a promising potential in the near future to assist endoscopists in diagnosing precancerous lesions and ESCCs.

Artificial Intelligence-assisted System Improves Endoscopic Identification of Colorectal Neoplasms

BACKGROUND & AIMS

Precise optical diagnosis of colorectal polyps could improve the cost-effectiveness of colonoscopy and reduce polypectomy-related complications. However, it is difficult for community-based non-experts to obtain sufficient diagnostic performance. Artificial intelligence-based systems have been developed to analyze endoscopic images; they identify neoplasms with high accuracy and low interobserver variation. We performed a multi-center study to determine the diagnostic accuracy of EndoBRAIN, an artificial intelligence-based system that analyzes cell nuclei, crypt structure, and microvessels in endoscopic images, in identification of colon neoplasms.

METHODS

The EndoBRAIN system was initially trained using 69,142 endocytoscopic images, taken at 520-fold magnification, from patients with colorectal polyps who underwent endoscopy at 5 academic centers in Japan from October 2017 through March 2018. We performed a retrospective comparative analysis of the diagnostic performance of EndoBRAIN vs that of 30 endoscopists (20 trainees and 10 experts); the endoscopists assessed images from 100 cases produced via white-light microscopy, endocytoscopy with methylene blue staining, and endocytoscopy with narrow-band imaging. EndoBRAIN was used to assess endocytoscopic, but not white-light, images. The primary outcome was the accuracy of EndoBrain in distinguishing neoplasms from non-neoplasms, compared with that of endoscopists, using findings from pathology analysis as the reference standard.

RESULTS

In analysis of stained endocytoscopic images, EndoBRAIN identified colon lesions with 96.9% sensitivity (95% CI, 95.8%-97.8%), 100% specificity (95% CI, 99.6%-100%), 98% accuracy (95% CI, 97.3%-98.6%), a 100% positive-predictive value (95% CI, 99.8%-100%), and a 94.6% negative-predictive (95% CI, 92.7%-96.1%); these values were all significantly greater than those of the endoscopy trainees and experts. In analysis of narrow-band images, EndoBRAIN distinguished neoplastic from non-neoplastic lesions with 96.9% sensitivity (95% CI, 95.8-97.8), 94.3% specificity (95% CI, 92.3-95.9), 96.0% accuracy (95% CI, 95.1-96.8), a 96.9% positive-predictive value, (95% CI, 95.8-97.8), and a 94.3% negative-predictive value (95% CI, 92.3-95.9); these values were all significantly higher than those of the endoscopy trainees, sensitivity and negative-predictive value were significantly higher but the other values are comparable to those of the experts.

CONCLUSIONS

EndoBRAIN accurately differentiated neoplastic from non-neoplastic lesions in stained endocytoscopic images and endocytoscopic narrow-band images, when pathology findings were used as the standard. This technology has been authorized for clinical use by the Japanese regulatory agency and should be used in endoscopic evaluation of small polyps more widespread clinical settings. UMIN clinical trial no: UMIN000028843.

The ability of a novel blue laser imaging system for the diagnosis of invasion depth of colorectal neoplasms

BACKGROUND

Fujifilm has developed a novel endoscope system with two kinds of lasers that enables us to allow narrow-band light observation with blue laser imaging (BLI). The aim of this study was to evaluate BLI magnification in comparison with narrow-band imaging (NBI) magnification for the diagnosis of colorectal neoplasms.

METHODS

This was a multicenter open study. A total of 104 colorectal neoplasms were examined with BLI and NBI magnifications in Kyoto Prefectural University of Medicine and Fukuoka University Chikushi Hospital. Vascular and surface patterns of tumors under BLI magnification were compared with those under NBI magnification, using a published NBI classification. The main outcome was the correlation between the NBI classification diagnosed by BLI or NBI magnification and the histopathological analyses.

RESULTS

Sixty-two cases of adenoma, 34 cases of intramucosal cancer and shallowly invaded submucosal cancer, and eight cases of deeply invaded submucosal cancer were diagnosed. The diagnostic accuracy of BLI magnification in the NBI classification was 74.0 % (77/104), similar to that of NBI magnification (77.8 %). The consistency rate between BLI and NBI magnification in the NBI classification was 74.0 %. Concerning image evaluation, the interobserver variability of two expert endoscopists (N.Y. and T.H.) in BLI magnification was κ = 0.863. On the other hand, the intraobserver variability of the two endoscopists was κ = 0.893 (N.Y.) and 0.851 (T.H.).

CONCLUSIONS

BLI magnification by laser source could predict histopathological diagnosis and invasion depth of colorectal neoplasms. The diagnostic effectiveness of this method was similar to that of NBI magnification.

Highly accurate artificial intelligence systems to predict the invasion depth of gastric cancer: efficacy of conventional white-light imaging, nonmagnifying narrow-band imaging, and indigo-carmine dye contrast imaging

BACKGROUND AND AIMS

Diagnosing the invasion depth of gastric cancer (GC) is necessary to determine the optimal method of treatment. Although the efficacy of evaluating macroscopic features and EUS has been reported, there is a need for more accurate and objective methods. The primary aim of this study was to test the efficacy of novel artificial intelligence (AI) systems in predicting the invasion depth of GC.

METHODS

A total of 16,557 images from 1084 cases of GC for which endoscopic resection or surgery was performed between January 2013 and June 2019 were extracted. Cases were randomly assigned to training and test datasets at a ratio of 4:1. Through transfer learning leveraging a convolutional neural network architecture, ResNet50, 3 independent AI systems were developed. Each system was trained to predict the invasion depth of GC using conventional white-light imaging (WLI), nonmagnifying narrow-band imaging (NBI), and indigo-carmine dye contrast imaging (Indigo).

RESULTS

The area under the curve of the WLI AI system was .9590. The lesion-based sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of the WLI AI system were 84.4%, 99.4%, 94.5%, 98.5%, and 92.9%, respectively. The lesion-based accuracies of the WLI, NBI, and Indigo AI systems were 94.5%, 94.3%, and 95.5%, respectively, with no significant difference.

CONCLUSIONS

These new AI systems trained with multiple images from different angles and distances could predict the invasion depth of GC with high accuracy. The lesion-based accuracy of the WLI, NBI, and Indigo AI systems was not significantly different.

Evaluation of an e-learning system for diagnosis of gastric lesions using magnifying narrow-band imaging: a multicenter randomized controlled study

Background and study aim Magnifying narrow-band imaging (M-NBI) is useful for the accurate diagnosis of early gastric cancer (EGC). However, acquiring skill at M-NBI diagnosis takes substantial effort. An Internet-based e-learning system to teach endoscopic diagnosis of EGC using M-NBI has been developed. This study evaluated its effectiveness. Participants and methods This study was designed as a multicenter randomized controlled trial. We recruited endoscopists as participants from all over Japan. After completing Test 1, which consisted of M-NBI images of 40 gastric lesions, participants were randomly assigned to the e-learning or non-e-learning groups. Only the e-learning group was allowed to access the e-learning system. After the e-learning period, both groups received Test 2. The analysis set was participants who scored < 80 % accuracy on Test 1. The primary end point was the difference in accuracy between Test 1 and Test 2 for the two groups. Results A total of 395 participants from 77 institutions completed Test 1 (198 in the e-learning group and 197 in the non-e-learning group). After the e-learning period, all 395 completed Test 2. The analysis sets were e-learning group: n = 184; and non-e-learning group: n = 184. The mean Test 1 score was 59.9 % for the e-learning group and 61.7 % for the non-e-learning group. The change in accuracy in Test 2 was significantly higher in the e-learning group than in the non-e-learning group (7.4 points vs. 0.14 points, respectively; P < 0.001). Conclusion This study clearly demonstrated the efficacy of the e-learning system in improving practitioners' capabilities to diagnose EGC using M-NBI.Trial registered at University Hospital Medical Information Network Clinical Trials Registry (UMIN000008569).

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

BACKGROUND

Experts can accurately characterize the histology of diminutive polyps with narrow-band imaging (NBI). There are limited data on the performance of non-experts.

OBJECTIVE

To assess the impact of a computer-based teaching module on the accuracy of predicting polyp histology with NBI by non-experts (in academics and community practice) by using video clips.

DESIGN

Prospective, observational study.

SETTING

Academic and community practice.

PARTICIPANTS

A total of 15 gastroenterologists participated-5 experts in NBI, 5 non-experts in academic practice, and 5 non-experts in community practice.

INTERVENTION

Participants reviewed a 20-minute, computer-based teaching module outlining the different NBI features for hyperplastic and adenomatous polyps.

MAIN OUTCOME MEASUREMENTS

Performance characteristics in characterizing the histology of diminutive polyps with NBI by using short video clips before (pretest) and after (posttest) reviewing the teaching module.

RESULTS

Non-experts in academic practice showed a significant improvement in the sensitivity (54% vs 79%; P < .001), accuracy (64% vs 81%; P < .001), and proportion of high-confidence diagnoses (49% vs 69%; P < .001) in the posttest. Non-experts in community practice had significantly higher sensitivity (58% vs 75%; P = .004), specificity (76% vs 90%; P = .04), accuracy (64% vs 81%; P < .001), and proportion of high-confidence diagnoses (49% vs 72%; P < .001) in the posttest. Performance of experts in NBI was significantly better than non-experts in both academic and community practice.

LIMITATIONS

Selection bias in selecting good quality videos. Performance not assessed during live colonoscopy.

CONCLUSION

Academic and community gastroenterologists without prior experience in NBI can achieve significant improvements in characterizing diminutive polyp histology after a brief computer-based training. The durability of these results and applicability in everyday practice are uncertain.

Prospective cohort study assessing outcomes of patients from families fulfilling criteria for hereditary diffuse gastric cancer undergoing endoscopic surveillance

BACKGROUND

Prophylactic total gastrectomy is performed in hereditary diffuse gastric cancer (HDGC) patients carrying the CDH1 mutation because endoscopic surveillance often fails to detect microscopic disease.

OBJECTIVE

The aim of this study was to determine the natural history and outcomes of patients with HDGC undergoing endoscopy.

DESIGN

Prospective, cohort observational study.

SETTINGS

Tertiary referral center.

PATIENTS

Patients fulfilling criteria for HDGC who opted to undergo endoscopy.

INTERVENTION

Research surveillance program using high-resolution white-light endoscopy with autofluorescence and narrow-band imaging combined with targeted and multiple random biopsies assessed by an expert histopathologist for the presence of signet ring cell carcinoma.

MAIN OUTCOME MEASUREMENTS

The primary endpoint was the endoscopic yield of microscopic signet ring cell carcinoma according to patient mutation status and subsequent decision to undergo surgery. The secondary endpoint was the additional yield of targeted biopsies compared with random biopsies.

RESULTS

Between September 2007 and March 2013, 29 patients from 17 families underwent 70 surveillance endoscopies. Signet ring cell carcinoma foci were identified in 14 of 22 (63.6%) patients with confirmed CDH1 germline mutations and 2 of 7 (28.6%) with no pathogenic mutation identified. Eleven of 16 (9 CDH1-positive) patients proceeded to gastrectomy in a median 5.7 months. Five patients delayed surgery. In 1 patient, advanced gastric cancer developed 40.2 months after the first endoscopic findings.

LIMITATIONS

No control group.

CONCLUSIONS

Careful white-light examination with targeted and random biopsies combined with detailed histopathology can identify early lesions and help to inform decision making with regard to gastrectomy. Autofluorescence and narrow-band imaging are of limited utility. Delaying gastrectomy in individuals with signet ring cell carcinoma foci carries a high risk and has to be weighed carefully.

A deep learning-based system for identifying differentiation status and delineating the margins of early gastric cancer in magnifying narrow-band imaging endoscopy

BACKGROUND : Accurate identification of the differentiation status and margins for early gastric cancer (EGC) is critical for determining the surgical strategy and achieving curative resection in EGC patients. The aim of this study was to develop a real-time system to accurately identify differentiation status and delineate the margins of EGC on magnifying narrow-band imaging (ME-NBI) endoscopy. METHODS : 2217 images from 145 EGC patients and 1870 images from 139 EGC patients were retrospectively collected to train and test the first convolutional neural network (CNN1) to identify EGC differentiation status. The performance of CNN1 was then compared with that of experts using 882 images from 58 EGC patients. Finally, 928 images from 132 EGC patients and 742 images from 87 EGC patients were used to train and test CNN2 to delineate the EGC margins. RESULTS : The system correctly predicted the differentiation status of EGCs with an accuracy of 83.3 % (95 % confidence interval [CI] 81.5 % - 84.9 %) in the testing dataset. In the man - machine contest, CNN1 performed significantly better than the five experts (86.2 %, 95 %CI 75.1 % - 92.8 % vs. 69.7 %, 95 %CI 64.1 % - 74.7 %). For delineating EGC margins, the system achieved an accuracy of 82.7 % (95 %CI 78.6 % - 86.1 %) in differentiated EGC and 88.1 % (95 %CI 84.2 % - 91.1 %) in undifferentiated EGC under an overlap ratio of 0.80. In unprocessed EGC videos, the system achieved real-time diagnosis of EGC differentiation status and EGC margin delineation in ME-NBI endoscopy. CONCLUSION : We developed a deep learning-based system to accurately identify differentiation status and delineate the margins of EGC in ME-NBI endoscopy. This system achieved superior performance when compared with experts and was successfully tested in real EGC videos.

A computer-assisted algorithm for narrow-band imaging-based tissue characterization in Barrett's esophagus

BACKGROUND AND AIMS

The endoscopic evaluation of narrow-band imaging (NBI) zoom imagery in Barrett's esophagus (BE) is associated with suboptimal diagnostic accuracy and poor interobserver agreement. Computer-aided diagnosis (CAD) systems may assist endoscopists in the characterization of Barrett's mucosa. Our aim was to demonstrate the feasibility of a deep-learning CAD system for tissue characterization of NBI zoom imagery in BE.

METHODS

The CAD system was first trained using 494,364 endoscopic images of general endoscopic imagery. Next, 690 neoplastic BE and 557 nondysplastic BE (NDBE) white-light endoscopy overview images were used for refinement training. Subsequently, a third dataset of 112 neoplastic and 71 NDBE NBI zoom images with histologic correlation was used for training and internal validation. Finally, the CAD system was further trained and validated with a fourth, histologically confirmed dataset of 59 neoplastic and 98 NDBE NBI zoom videos. Performance was evaluated using fourfold cross-validation. The primary outcome was the diagnostic performance of the CAD system for classification of neoplasia in NBI zoom videos.

RESULTS

The CAD system demonstrated accuracy, sensitivity, and specificity for detection of BE neoplasia using NBI zoom images of 84%, 88%, and 78%, respectively. In total, 30,021 individual video frames were analyzed by the CAD system. Accuracy, sensitivity, and specificity of the video-based CAD system were 83% (95% confidence interval [CI], 78%-89%), 85% (95% CI, 76%-94%), and 83% (95% CI, 76%-90%), respectively. The mean assessment speed was 38 frames per second.

CONCLUSION

We have demonstrated promising diagnostic accuracy of predicting the presence/absence of Barrett's neoplasia on histologically confirmed unaltered NBI zoom videos with fast corresponding assessment time.

Comparison of performances of artificial intelligence versus expert endoscopists for real-time assisted diagnosis of esophageal squamous cell carcinoma (with video)

BACKGROUND AND AIMS

Narrow-band imaging (NBI) is currently regarded as the standard modality for diagnosing esophageal squamous cell carcinoma (SCC). We developed a computerized image-analysis system for diagnosing esophageal SCC by NBI and estimated its performance with video images.

METHODS

Altogether, 23,746 images from 1544 pathologically proven superficial esophageal SCCs and 4587 images from 458 noncancerous and normal tissue were used to construct an artificial intelligence (AI) system. Five- to 9-second video clips from 144 patients captured by NBI or blue-light imaging were used as the validation dataset. These video images were diagnosed by the AI system and 13 board-certified specialists (experts).

RESULTS

The diagnostic process was divided into 2 parts: detection (identify suspicious lesions) and characterization (differentiate cancer from noncancer). The sensitivities, specificities, and accuracies for the detection of SCC were, respectively, 91%, 51%, and 63% for the AI system and 79%, 72%, and 75% for the experts. The sensitivity of the AI system was significantly higher than that of the experts, but its specificity was significantly lower. Sensitivities, specificities, and accuracy for the characterization of SCC were, respectively, 86%, 89%, and 88% for the AI system and 74%, 76%, and 75% for the experts. The receiver operating characteristic curve showed that the AI system had significantly better diagnostic performance than the experts.

CONCLUSIONS

Our AI system showed significantly higher sensitivity for detecting SCC and higher accuracy for characterizing SCC from noncancerous tissue than endoscopic experts.

Application of magnifying endoscopy with narrow-band imaging in diagnosing gastric lesions: a prospective study

BACKGROUND

Magnifying endoscopy with narrow-band imaging (ME-NBI) can more clearly assess the surface pattern and microvascular architecture of gastric lesions.

OBJECTIVE

To evaluate the diagnostic efficacy of ME-NBI in patients with early gastric cancer.

DESIGN

Prospective study.

SETTING

Single academic center.

PATIENTS

This study involved 164 suspected gastric lesions in 146 consecutive patients who underwent ME-NBI for additional differential diagnosis before treatment.

INTERVENTION

ME-NBI findings were classified into 3 groups based on irregularities, absence of surface pattern, and microvascular architecture. All lesions were treated endoscopically or surgically, and ME-NBI diagnosis was compared with histopathological findings.

MAIN OUTCOME MEASUREMENTS

Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of real-time ME-NBI diagnosis were determined.

RESULTS

The sensitivity, specificity, and accuracy of ME-NBI were 97.3%, 84.4%, and 90.2%, respectively, in distinguishing between cancerous and noncancerous lesions and were 92.3%, 89.7%, and 90.4%, respectively, in distinguishing undifferentiated from differentiated adenocarcinoma. ME-NBI accurately predicted depth of invasion in 37 of 39 differentiated adenocarcinomas (95%).

LIMITATIONS

The sample size was relatively small.

CONCLUSIONS

ME-NBI can successfully distinguish between cancerous and noncancerous lesions and between undifferentiated and differentiated adenocarcinomas. Of the 3 patterns on ME-NBI, type A is mainly characteristic of noncancerous lesions, type B is a good indicator of differentiated adenocarcinoma and intramucosal/superficially invasive cancers, and type C is indicative of undifferentiated adenocarcinoma or differentiated cancer with deep submucosal invasion.

Recurrent intestinal metaplasia after radiofrequency ablation for Barrett's esophagus: endoscopic findings and anatomic location

BACKGROUND

Radiofrequency ablation (RFA) is a safe and effective treatment for Barrett's esophagus (BE) that results in high rates of complete eradication of intestinal metaplasia (CEIM). However, recurrence is common after CEIM, and surveillance endoscopy is recommended. Neither the anatomic location nor the endoscopic appearance of these recurrences is well-described.

OBJECTIVE

Describe the location of histologic specimens positive for recurrence after CEIM and the testing performance of endoscopic findings for the histopathologic detection of recurrence.

DESIGN

Retrospective cohort.

SETTING

Single referral center.

PATIENTS

A total of 198 patients with BE with at least 2 surveillance endoscopies after CEIM.

INTERVENTIONS

RFA, EMR, surveillance endoscopy.

MAIN OUTCOME MEASUREMENTS

The anatomic location and histologic grade of recurrence.

RESULTS

In a mean 3.0 years of follow-up, 32 (16.2%; 95% confidence interval [CI], 11.0%-22.0%) patients had recurrence of disease, 5 (2.5%; 95% CI, 0.3%-4.7%) of whom progressed beyond their worst before-treatment histology. Recurrence was most common at or near the gastroesophageal junction (GEJ). Recurrence>1 cm proximal to the GEJ always was accompanied by endoscopic findings, and random biopsies in these areas detected no additional cases. The sensitivity of any esophageal sign under high-definition white light or narrow-band imaging for recurrence was 59.4% (42.4%, 76.4%), and the specificity was 80.6% (77.2%, 84.0%).

LIMITATIONS

Single-center study.

CONCLUSION

Recurrent intestinal metaplasia often is not visible to the endoscopist and is most common near the GEJ. Random biopsies>1 cm above the GEJ had no yield for recurrence. In addition to biopsy of prior EMR sites and of suspicious lesions, random biopsies oversampling the GEJ are recommended.

Real-time assessment of video images for esophageal squamous cell carcinoma invasion depth using artificial intelligence

BACKGROUND

Although optimal treatment of superficial esophageal squamous cell carcinoma (SCC) requires accurate evaluation of cancer invasion depth, the current process is rather subjective and may vary by observer. We, therefore, aimed to develop an AI system to calculate cancer invasion depth.

METHODS

We gathered and selected 23,977 images (6857 WLI and 17,120 NBI/BLI images) of pathologically proven superficial esophageal SCC from endoscopic videos and still images of superficial esophageal SCC taken in our facility, to use as a learning dataset. We annotated the images with information [such as magnified endoscopy (ME) or non-ME, pEP-LPM, pMM, pSM1, and pSM2-3 cancers] based on pathologic diagnosis of the resected specimens. We created a model using a convolutional neural network. Performance of the AI system was compared with that of invited experts who used the same validation video set, independent of the learning dataset.

RESULTS

Accuracy, sensitivity, and specificity with non-magnified endoscopy (ME) were 87%, 50%, and 99% for the AI system and 85%, 45%, 97% for the experts. Accuracy, sensitivity, and specificity with ME were 89%, 71%, and 95% for the AI system and 84%, 42%, 97% for the experts.

CONCLUSIONS

Most diagnostic parameters were higher when done by the AI system than by the experts. These results suggest that our AI system could potentially provide useful support during endoscopies.

Accuracy of artificial intelligence-assisted detection of upper GI lesions: a systematic review and meta-analysis

BACKGROUND AND AIMS

Artificial intelligence (AI)-assisted detection is increasingly used in upper endoscopy. We performed a meta-analysis to determine the diagnostic accuracy of AI on detection of gastric and esophageal neoplastic lesions and Helicobacter pylori (HP) status.

METHODS

We searched Embase, PubMed, Medline, Web of Science, and Cochrane databases for studies on AI detection of gastric or esophageal neoplastic lesions and HP status. After assessing study quality using the Quality Assessment of Diagnostic Accuracy Studies tool, a bivariate meta-analysis following a random-effects model was used to summarize the data and plot hierarchical summary receiver-operating characteristic curves. The diagnostic accuracy was determined by the area under the hierarchical summary receiver-operating characteristic curve (AUC).

RESULTS

Twenty-three studies including 969,318 images were included. The AUC of AI detection of neoplastic lesions in the stomach, Barrett's esophagus, and squamous esophagus and HP status were .96 (95% confidence interval [CI], .94-.99), .96 (95% CI, .93-.99), .88 (95% CI, .82-.96), and .92 (95% CI, .88-.97), respectively. AI using narrow-band imaging was superior to white-light imaging on detection of neoplastic lesions in squamous esophagus (.92 vs .83, P < .001). The performance of AI was superior to endoscopists in the detection of neoplastic lesions in the stomach (AUC, .98 vs .87; P < .001), Barrett's esophagus (AUC, .96 vs .82; P < .001), and HP status (AUC, .90 vs .82; P < .001).

CONCLUSIONS

AI is accurate in the detection of upper GI neoplastic lesions and HP infection status. However, most studies were based on retrospective reviews of selected images, which requires further validation in prospective trials.

Preoperative endoscopic diagnosis of superficial non-ampullary duodenal epithelial tumors, including magnifying endoscopy

Superficial non-ampullary duodenal epithelial tumor (SNADET) is defined as a sporadic tumor that is confined to the mucosa or submucosa that does not arise from Vater's papilla, and it includes adenoma and adenocarcinoma. Recent developments in endoscopic technology, such as high-resolution endoscopy and image-enhanced endoscopy, may increase the chances of detecting SNADET lesions. However, because SNADET is rare, little is known about its preoperative endoscopic diagnosis. The use of endoscopic resection for SNADET, which has no risk of metastasis, is increasing, but the incidence of complications, such as perforation, is significantly higher than in any other part of the digestive tract. A preoperative diagnosis is required to distinguish between lesions that should be followed up and those that require treatment. Retrospective studies have revealed certain endoscopic findings that suggest malignancy. In recent years, several new imaging modalities have been developed and explored for real-time diagnosis of these lesion types. Establishing an endoscopic diagnostic tool to differentiate between adenoma and adenocarcinoma in SNADET lesions is required to select the most appropriate treatment. This review describes the current state of knowledge about preoperative endoscopic diagnosis of SNADETs, such as duodenal adenoma and duodenal adenocarcinoma. Newer endoscopic techniques, including magnifying endoscopy, may help to guide these diagnostics, but their additional advantages remain unclear, and further studies are required to clarify these issues.

Deep learning system compared with expert endoscopists in predicting early gastric cancer and its invasion depth and differentiation status (with videos)

BACKGROUND AND AIMS

We aimed to develop and validate a deep learning-based system that covers various aspects of early gastric cancer (EGC) diagnosis, including detecting gastric neoplasm, identifying EGC, and predicting EGC invasion depth and differentiation status. Herein, we provide a state-of-the-art comparison of the system with endoscopists using real-time videos in a nationwide human-machine competition.

METHODS

This multicenter, prospective, real-time, competitive comparative, diagnostic study enrolled consecutive patients who received magnifying narrow-band imaging endoscopy at the Peking University Cancer Hospital from June 9, 2020 to November 17, 2020. The offline competition was conducted in Wuhan, China, and the endoscopists and the system simultaneously read patients' videos and made diagnoses. The primary outcomes were sensitivity in detecting neoplasms and diagnosing EGCs.

RESULTS

One hundred videos, including 37 EGCs and 63 noncancerous lesions, were enrolled; 46 endoscopists from 44 hospitals in 19 provinces in China participated in the competition. The sensitivity rates of the system for detecting neoplasms and diagnosing EGCs were 87.81% and 100%, respectively, significantly higher than those of endoscopists (83.51% [95% confidence interval [CI], 81.23-85.79] and 87.13% [95% CI, 83.75-90.51], respectively). Accuracy rates of the system for predicting EGC invasion depth and differentiation status were 78.57% and 71.43%, respectively, slightly higher than those of endoscopists (63.75% [95% CI, 61.12-66.39] and 64.41% [95% CI, 60.65-68.16], respectively).

CONCLUSIONS

The system outperformed endoscopists in identifying EGCs and was comparable with endoscopists in predicting EGC invasion depth and differentiation status in videos. This deep learning-based system could be a powerful tool to assist endoscopists in EGC diagnosis in clinical practice.

Pit pattern analysis with high-definition chromoendoscopy and narrow-band imaging for optical diagnosis of dysplasia in patients with ulcerative colitis

BACKGROUND AND AIMS

Patients with longstanding ulcerative colitis (UC) are at increased risk of developing colorectal neoplasia. Chromoendoscopy (CE) increases detection of lesions, and Kudo pit pattern classification I and II have been suggested to be predictive of benign polyps in UC. Little is known on the use of this classification in nonmagnified high-definition (HD) (virtual) CE and narrow-band Imaging (NBI) or on the interobserver agreement. The aim of this pilot study was to assess the diagnostic accuracy and the interobserver agreement of the Kudo pit pattern classification in UC patients undergoing surveillance with methylene blue CE or NBI in a multicenter study.

METHODS

Fifty images of lesions identified in 27 UC patients (13 neoplastic) either with classical CE (methylene blue .1%; n = 24) or NBI (n = 26) were selected by an independent investigator. Images were selected from a randomized controlled trial to compare CE and NBI. All nonmagnified images were obtained with a processor and mounted in a PowerPoint file in a standardized way (same size; black background). Ten endoscopists with extensive experience in NBI/CE were asked to assess the lesions for the predominant Kudo pit pattern (I, II, IIIL, IIIS, IV, and V) to indicate if they believed the lesion was neoplastic and how confident they were about the diagnosis. Histology was used as the criterion standard.

RESULTS

Median sensitivity, specificity, negative predictive value, and positive predictive value for diagnosing neoplasia based on the presence of pit pattern other than I or II was 77%, 68%, 88%, and 46%, respectively. Diagnostic accuracy was significantly higher when a diagnosis was made with a high level of confidence (77% vs 21%, P < .001). The overall interobserver agreement for any pit pattern was only fair (κ = .282), with CE being significantly better than NBI (.322 vs .224, P < .001). From a clinical viewpoint the difference between neoplastic and non-neoplastic lesions is important. The agreement for differentiation between non-neoplastic patterns (I, II) and neoplastic patterns (IIIL, IIIS, IV, or V) was moderate (κ = .587) and even significantly better for NBI in comparison with CE (κ = .653 vs .495, P < .001).

CONCLUSIONS

Differentiation between non-neoplastic and neoplastic pit patterns in UC lesions shows a moderate to substantial agreement among expert endoscopists. The agreement for differentiating neoplastic from non-neoplastic lesions is significantly better for NBI in comparison with HD CE. The assessment of pit pattern I or II with nonmagnified HD CE or NBI has a high negative predictive value to rule out neoplasia. (Clinical trial registration number: NCT01882205.).

Magnifying narrow-band imaging endoscopy is superior in diagnosis of early gastric cancer

AIM: To evaluate the diagnostic effectiveness of white light endoscopy, magnifying endoscopy (ME), and magnifying narrow-band imaging endoscopy (ME-NBI) in detecting early gastric cancer (EGC).

METHODS: From March 2010 to June 2012, a total of 3616 patients received screening for gastric cancer by magnifying endoscopy. There were 3675 focal gastric lesions detected using conventional high definition white light endoscopy (HD-WLE) in four different referential hospitals that were recruited for further investigation using ME and ME-NBI. The images obtained from HD-WLE, ME, and ME-NBI were reviewed by four experienced endoscopists to evaluate their diagnostic effectiveness for EGC. The diagnosis of cancerous and non-cancerous lesions was conducted by evaluating the microvascular and microsurface patterns using the VS classification system. The final endoscopic diagnosis of each lesion was determined by consultation when a disagreement occurred. We used histopathological results as the gold standard for the diagnosis of EGC.

RESULTS: Among the 3675 lesions found, 1508 were validated by pathological findings as chronic gastritis, 1279 as chronic gastritis with intestinal metaplasia, 631 as low-grade neoplasia, and 257 as EGC. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of HD-WLE for the diagnosis of EGC were 71.2%, 99.1%, 85.5%, 97.9% and 97.1%, respectively. The results of ME for diagnosing EGC were 81.3%, 98.8%, 83.3%, 98.6% and 97.6%, respectively. The results of ME-NBI for the diagnosis of EGC were 87.2%, 98.6%, 82.1%, 99.0% and 97.8%, respectively. The diagnostic sensitivity and accuracy of paired ME and ME-NBI were significantly better than those of HD-WLE (P < 0.05).

CONCLUSION: HD-WLE has a relatively high accuracy for diagnosing EGC and is an effective screening tool. Further investigations of ME and ME-NBI are required to achieve superior accuracy.

Diagnostic performance of magnifying narrow-band imaging for early gastric cancer: A meta-analysis

AIM: To investigate the performance of magnifying endoscopy with narrow-band imaging (ME-NBI) in the diagnosis of early gastric cancer (EGC).

METHODS: Systematic literature searches were conducted until February 2014 in PubMed, EMBASE, Web of Science, Ovid, Scopus and the Cochrane Library databases by two independent reviewers. Meta-analysis was performed to calculate the pooled sensitivity, specificity and diagnostic odds ratio and to construct a summary receiver operating characteristic (ROC) curve. Subgroup analyses were performed based on the morphology type of lesions, diagnostic standard, the size of lesions, type of assessment, country and sample size to explore possible sources of heterogeneity. A Deeks’ asymmetry test was used to evaluate the publication bias.

RESULTS: Fourteen studies enrolling 2171 patients were included. The pooled sensitivity, specificity and diagnostic odds ratio for ME-NBI diagnosis of EGC were 0.86 (95%CI: 0.83-0.89), 0.96 (95%CI: 0.95-0.97) and 102.75 (95%CI: 48.14-219.32), respectively, with the area under ROC curve being 0.9623. Among the 14 studies, six also evaluated the diagnostic value of conventional white-light imaging, with a sensitivity of 0.57 (95%CI: 0.50-0.64) and a specificity of 0.79 (95%CI: 0.76-0.81). When using “VS” (vessel plus surface) ME-NBI diagnostic systems in gastric lesions of depressed macroscopic type, the pooled sensitivity and specificity were 0.64 (95%CI: 0.52-0.75) and 0.96 (95%CI: 0.95-0.98). For the lesions with a diameter less than 10 mm, the sensitivity and specificity were 0.74 (95%CI: 0.65-0.82) and 0.98 (95%CI: 0.97-0.98).

CONCLUSION: ME-NBI is a promising endoscopic tool in the diagnosis of early gastric cancer and might be helpful in further target biopsy.

Image-enhanced endoscopy for gastric preneoplastic conditions and neoplastic lesions: a systematic review and meta-analysis

BACKGROUND : Image-enhanced endoscopy (IEE) improves the accuracy of endoscopic diagnosis. We aimed to assess the value of IEE for gastric preneoplastic conditions and neoplastic lesions. METHODS : Medline and Embase were searched until December 2018. Studies allowing calculation of diagnostic measures were included. Risk of bias and applicability were assessed using QUADAS-2. Subgroup analysis was performed to explore heterogeneity. RESULTS : 44 studies met the inclusion criteria. For gastric intestinal metaplasia (GIM), narrow-band imaging (NBI) obtained a pooled sensitivity and specificity of 0.79 (95 %CI 0.72-0.85) and 0.91 (95 %CI 0.88-0.94) on per-patient basis; on per-biopsy basis, it was 0.84 (95 %CI 0.81-0.86) and 0.95 (95 %CI 0.94-0.96), respectively. Tubulovillous pattern was the most accurate marker to detect GIM and it was effectively assessed without high magnification. For dysplasia, NBI showed a pooled sensitivity and specificity of 0.87 (95 %CI 0.84-0.89) and 0.97 (95 %CI 0.97-0.98) on per-biopsy basis. The use of magnification improved the performance of NBI to characterize early gastric cancer (EGC), especially when the vessel plus surface (VS) classification was applied. Regarding other technologies, trimodal imaging also obtained a high accuracy for dysplasia (sensitivity 0.93 [95 %CI 0.85-0.98], specificity 0.98 [95 %CI 0.92-1.00]). For atrophic gastritis, no specific pattern was noted and none of the technologies reached good diagnostic yield. CONCLUSION : NBI is highly accurate for GIM and dysplasia. The presence of tubulovillous pattern and the VS classification seem to be useful to detect GIM and characterize EGC, respectively. These features should be used in current practice and to standardize endoscopic criteria for other technologies.

Comparison of Endoscopic Dysplasia Detection Techniques in Patients With Ulcerative Colitis: A Systematic Review and Network Meta-analysis

Background

We assessed the comparative efficacy of different dysplasia detection techniques in patients with ulcerative colitis (UC) through a network meta-analysis and rated the quality of evidence using GRADE approach.

Methods

Through a systematic literature review of multiple databases through June 30, 2017, we identified parallel-group randomized controlled trials (RCTs) in adults with long-standing UC undergoing surveillance colonoscopy with standard definition-white light endoscopy (SD-WLE), high-definition WLE (HD-WLE), narrow band imaging (NBI), or dye-based chromoendoscopy. The primary outcome was the total number of dysplastic lesions. Pairwise and network meta-analysis was performed; ranking was assessed using surface under the cumulative ranking (SUCRA) probabilities.

Results

Based on 8 parallel-group RCTs (924 patients), low-quality evidence supports chromoendoscopy over SD-WLE (odds ratio [OR], 2.37; 95% credible interval [CrI], 0.81-6.94) for any dysplasia detection, whereas very low-quality evidence supports using HD-WLE or NBI over SD-WLE (HD-WLE [vs SD-WLE]: OR, 1.21; 95% CrI, 0.30-4.85; NBI: OR, 1.68; 95% CrI, 0.54-5.22). Very low-quality evidence from indirect comparative analysis supports the use of chromoendoscopy over HD-WLE (OR, 1.96; 95% CrI, 0.72-5.34) or NBI (OR, 1.41; 95% CrI, 0.70-2.84) for any dysplasia detection. The number of patients with advanced neoplasia was very small, precluding meaningful analysis.

Conclusions

Although we did not find any single technique to be superior, chromoendoscopy is probably more effective than SD-WLE for detecting any dysplasia, and there is low confidence in estimates supporting its use over HD-WLE or NBI. There is very low-quality evidence to inform the comparative efficacy of these interventions in detecting advanced neoplasia or preventing future colorectal cancer. Pragmatic, parallel-group RCTs with longitudinal follow-up are warranted to inform optimal dysplasia surveillance techniques. 10.1093/ibd/izy188_video1izy188.video15789702674001.