Rate of missed oesophageal cancer at routine endoscopy and survival outcomes: A multicentric cohort study

Esophageal cancers missed at upper endoscopy in Central Norway 2004 to 2021 - A population-based study

INTRODUCTION

The incidence of esophageal cancers is increasing in many Western countries and the rate of missed esophageal cancers (MEC) at upper endoscopy is of concern. We aimed to calculate the MEC rate and identify factors associated with MEC.

METHODS

This was a retrospective population-based cohort study including 613 patients diagnosed with esophageal cancer in Central Norway 2004-2021. MEC was defined as esophageal cancer diagnosed 6-36 months after a non-diagnostic upper endoscopy. Patient characteristics, tumor localization, histological type and cTNM stage were recorded. Symptoms, endoscopic findings, use of sedation and endoscopists experience at the endoscopy prior to esophageal cancer diagnosis and at the time of diagnosis were recorded. The association between these factors and MEC was assessed.

RESULTS

Forty-nine (8.0%) of 613 cancers were MEC. There was a significant increase in annual numbers of esophageal cancer (p < 0.001) as well as of MEC (p = 0.009), but MEC rate did not change significantly (p = 0.382). The median time from prior upper endoscopy to MEC diagnosis was 22.9 (12.1-28.6) months. MEC patients were older and were diagnosed with disease with a lower cTNM stage and cT category than non-missed cancers, whereas tumor localization and histological type were similar between the groups. The use of sedation or endoscopist experience did not differ between the endoscopy prior to esophageal cancer diagnosis and at the time of diagnosis. High proportions of MEC patients had Barrett's esophagus (n = 25, 51.0%), hiatus hernia (n = 26, 53.1%), esophagitis (n = 10, 20.4%) or ulceration (n = 4, 8.2%). Significant proportions of MECs were diagnosed after inappropriate follow-up of endoscopic Barrett's esophagus, histological dysplasia or ulcerations.

CONCLUSIONS

The annual number of MEC increased during the study period, while the MEC rate remained unchanged. Endoscopic findings related to gastroesophageal reflux disease such as esophagitis and Barrett's esophagus were identified in a high proportion of patients with subsequent MECs. Cautious follow-up of these patients could potentially reduce MEC-rate.

Will artificial intelligence reach any limit in gastroenterology?

Endoscopy is the cornerstone in the management of digestive diseases. Over the last few decades, technology has played an important role in the development of this field, helping endoscopists in better detecting and characterizing luminal lesions. However, despite ongoing advancements in endoscopic technology, the incidence of missed pre-neoplastic and neoplastic lesions remains high due to the operator-dependent nature of endoscopy and the challenging learning curve associated with new technologies. Artificial intelligence (AI), an operator-independent field, could be an invaluable solution. AI can serve as a “second observer”, enhancing the performance of endoscopists in detecting and characterizing luminal lesions. By utilizing deep learning (DL), an innovation within machine learning, AI automatically extracts input features from targeted endoscopic images. DL encompasses both computer-aided detection and computer-aided diagnosis, assisting endoscopists in reducing missed detection rates and predicting the histology of luminal digestive lesions. AI applications in clinical gastrointestinal diseases are continuously expanding and evolving the entire digestive tract. In all published studies, real-time AI assists endoscopists in improving the performance of non-expert gastroenterologists, bringing it to a level comparable to that of experts. The development of DL may be affected by selection biases. Studies have utilized different AI-assisted models, which are heterogeneous. In the future, algorithms need validation through large, randomized trials. Theoretically, AI has no limit to assist endoscopists in increasing the accuracy and the quality of endoscopic exams. However, practically, we still have a long way to go before standardizing our AI models to be accepted and applied by all gastroenterologists.

Surveillance esophagogastroduodenoscopy using linked color imaging and narrow-band imaging: A multicenter randomized controlled trial

BACKGROUND AND AIM

There has been no report on a direct comparison between linked color imaging (LCI) and second-generation narrow-band imaging (2G-NBI) for surveillance of epithelial neoplasms in the upper gastrointestinal tract (UGIT). The aim of this study was to verify the superiority of LCI to 2G-NBI for surveillance esophagogastroduodenoscopy and to clarify how each endoscopic system should be used.

METHODS

This study was conducted as an open-label, two-arm-parallel (1:1), multicenter, randomized controlled trial at six institutions. Patients aged 20-85 years with a treatment history of epithelial neoplasms in the UGIT were recruited. Patients were assigned to a 2G-NBI group and an LCI group, and esophagogastroduodenoscopy was performed with primary image-enhanced endoscopy followed by white light imaging (WLI). The primary endpoint was the detection rate of one or more epithelial neoplasms in the primary image-enhanced endoscopy. A WLI-detected epithelial neoplasm was defined as a lesion that was detected in only WLI.

RESULTS

A total of 372 patients in the 2G-NBI group and 378 patients in the LCI group were analyzed. Epithelial neoplasms in the UGIT were detected by 2G-NBI in 18 patients (4.6%) and were detected by LCI in 20 patients (5.3%) (P = 0.87). WLI-detected epithelial neoplasms were in 11 patients in the 2G-NBI group (3.0%) and in 1 patient in the LCI group (0.27%) (P = 0.003).

CONCLUSIONS

Linked color imaging did not show superiority to 2G-NBI for the detection of epithelial neoplasms. Also, the percentage of WLI-detected epithelial neoplasms in primary NBI was significantly higher than that in primary LCI.

Revolutionizing healthcare by use of artificial intelligence in esophageal carcinoma - a narrative review

Esophageal cancer is a major cause of cancer-related mortality worldwide, with significant regional disparities. Early detection of precursor lesions is essential to improve patient outcomes. Artificial intelligence (AI) techniques, including deep learning and machine learning, have proved to be of assistance to both gastroenterologists and pathologists in the diagnosis and characterization of upper gastrointestinal malignancies by correlating with the histopathology. The primary diagnostic method in gastroenterology is white light endoscopic evaluation, but conventional endoscopy is partially inefficient in detecting esophageal cancer. However, other endoscopic modalities, such as narrow-band imaging, endocytoscopy, and endomicroscopy, have shown improved visualization of mucosal structures and vasculature, which provides a set of baseline data to develop efficient AI-assisted predictive models for quick interpretation. The main challenges in managing esophageal cancer are identifying high-risk patients and the disease's poor prognosis. Thus, AI techniques can play a vital role in improving the early detection and diagnosis of precursor lesions, assisting gastroenterologists in performing targeted biopsies and real-time decisions of endoscopic mucosal resection or endoscopic submucosal dissection. Combining AI techniques and endoscopic modalities can enhance the diagnosis and management of esophageal cancer, improving patient outcomes and reducing cancer-related mortality rates. The aim of this review is to grasp a better understanding of the application of AI in the diagnosis, treatment, and prognosis of esophageal cancer and how computer-aided diagnosis and computer-aided detection can act as vital tools for clinicians in the long run.

Optimal timing of simethicone administration prior to upper endoscopy: A multicenter, single-blind, randomized controlled trial

Background and study aims Simethicone is useful as premedication for upper endoscopy because of its antifoaming effects. We aimed to evaluate the effect of timing of simethicone administration on mucosal visibility. Patients and methods In this multicenter, randomized, endoscopist-blinded study, patients scheduled for upper endoscopy were randomized to receive 40 mg simethicone at the following time points prior to the procedure: 20 to 30 minutes (early group), 0 to 10 minutes (late group) or 20 mg simethicone at both time points (split-dose group). Images were taken from nine predefined locations in the esophagus, stomach, and duodenum before endoscopic flushing. Each image was scored on mucosal visibility by three independent endoscopists on a 4-point scale (lower scores indicating better visibility), with adequate mucosal visibility defined as a score ≤ 2. Primary outcome was the percentage of patients with adequate total mucosal visibility (TMV), reached if all median subscores for each location were ≤ 2. Results A total of 386 patients were included (early group: 132; late group: 128; split-dose group: 126). Percentages of adequate TMV were 55%, 42%, and 61% in the early, late, and split-dose group, respectively ( P < 0.01). Adequate TMV was significantly higher in the split-dose group compared to the late group ( P < 0.01), but not compared to the early group ( P = 0.29). Differences between groups were largest in the stomach, where percentages of adequate mucosal visibility were higher in the early (68% vs 53%, P = 0.03) and split-dose group (69% vs 53%, P = 0.02) compared to the late group. Conclusions Mucosal visibility can be optimized with early simethicone administration, either as a single administration or in a split-dose regimen.

Artificial intelligence in endoscopy: Overview, applications, and future directions

Since the emergence of artificial intelligence (AI) in medicine, endoscopy applications in gastroenterology have been at the forefront of innovations. The ever-increasing number of studies necessitates the need to organize and classify applications in a useful way. Separating AI capabilities by computer aided detection (CADe), diagnosis (CADx), and quality assessment (CADq) allows for a systematic evaluation of each application. CADe studies have shown promise in accurate detection of esophageal, gastric and colonic neoplasia as well as identifying sources of bleeding and Crohn's disease in the small bowel. While more advanced CADx applications employ optical biopsies to give further information to characterize neoplasia and grade inflammatory disease, diverse CADq applications ensure quality and increase the efficiency of procedures. Future applications show promise in advanced therapeutic modalities and integrated systems that provide multimodal capabilities. AI is set to revolutionize clinical decision making and performance of endoscopy.

Artificial intelligence in esophageal cancer diagnosis and treatment: where are we now?-a narrative review

Background and Objective

Artificial intelligence (AI) use is becoming increasingly prevalent directly or indirectly in daily clinical practice, including esophageal cancer (EC) diagnosis and treatment. Although the limits of its adoption and their clinical benefits are still unknown, any physician related to EC patients' management should be aware of the status and future perspectives of AI use in their field. The purpose of this review is to summarize the existing literature regarding the role of AI in diagnosis and treatment of EC. We have focused on the aids AI entails in the management of this pathology and we have tried to offer an updated perspective to maximize current applications and to identify potential future uses of it.

Methods

Data concerning AI applied to EC diagnosis and treatment is not limited, including direct (those specifically related to them) and indirect (those referring to other specialties as radiology or pathology), applications. However, the clinical relevance of the discussed and presented models is still unknown. We performed a research in PubMed of English and Spanish written studies from January 1970 to June 2022.

Key Content and Findings

Information regarding the role of AI in EC diagnosis and treatment has increased exponentially in recent years. Several models, including different variables and features have been investigated and some of them internally and externally validated. However, the main challenge remains to apply and introduce all these data into clinical practice, and, as some of the discussed studies argue, if the models are able to enhance experienced endoscopists' judgement. Although AI use is increasing steadily in different medical specialties, the truth is, most of the time, the gap between model development and clinical implementation is not closed. Learning to understand the routinely application of AI, as well as future improvements, would lead to a broadened adoption.

Conclusions

Physicians should be aware of the multiple current clinical uses of AI in EC diagnosis and treatment and should take part in their clinical application and future developments to enhance patient care.

Analysis of white-light imaging-based features predictive for determination of lesion depths of superficial flat esophageal squamous cell carcinoma: a retrospective multicenter study from China

OBJECTIVES

Endoscopic diagnosis of invasion depth of superficial esophageal squamous cell carcinoma (SESCC) by white-light imaging (WLI) modality remains difficult. This study aims to clarify WLI-based features which are predictive for invasion depth of SESCC.

METHODS

A two-phase study was performed by enrolling 1288 patients with 1396 SESCC lesions. Endoscopic appearances, clinical characteristics and post-operative pathological outcomes were collected and reviewed. The association between lesion features and invasion depth were analyzed. A predictive nomogram was constructed for prediction of invasion depth.

RESULTS

Among 1396 lesions in derivation and validation cohort, 1139 (81.6%), 194 (13.9%) and 63 (4.5%) lesions were diagnosed as lesions confined into the intraepithelium or the lamina propria mucosa (T1a-EP/LPM), lesions invading the muscularis mucosa (T1a-MM) or superficial submucosa (T1b-SM1) and tumor with moderate invasion into the submucosa or deeper submucosal invasion (≥ T1b-SM2), respectively. Lesion length > 2 cm (p < 0.001), wider circumferential extension (p < 0.001, 0.002 and 0.048 for > 3/4, 1/2-3/4 and 1/4-1/2 circumferential extension, respectively), surface unevenness (p < 0.001 for both type 0-IIa/0-IIc lesions and mixed type lesions), spontaneous bleeding (p < 0.001), granularity (p < 0.001) and nodules (p < 0.001) were identified as significant factors predictive for lesion depth. A nomogram based on these factors was constructed and the values of area under the Receiver Operating Characteristics curve were 0.89 and 0.90 in the internal and external patient cohort.

CONCLUSIONS

Our study provides six WLI-based morphological features predicting for lesion depth of SESCC. Our findings will make endoscopic evaluation of invasion depth for SESCC more convenient by assessing these profiles.

The current state of artificial intelligence in endoscopic diagnosis of early esophageal squamous cell carcinoma

Esophageal squamous cell carcinoma (ESCC) is a common malignant tumor of the digestive tract. The most effective method of reducing the disease burden in areas with a high incidence of esophageal cancer is to prevent the disease from developing into invasive cancer through screening. Endoscopic screening is key for the early diagnosis and treatment of ESCC. However, due to the uneven professional level of endoscopists, there are still many missed cases because of failure to recognize lesions. In recent years, along with remarkable progress in medical imaging and video evaluation technology based on deep machine learning, the development of artificial intelligence (AI) is expected to provide new auxiliary methods of endoscopic diagnosis and the treatment of early ESCC. The convolution neural network (CNN) in the deep learning model extracts the key features of the input image data using continuous convolution layers and then classifies images through full-layer connections. The CNN is widely used in medical image classification, and greatly improves the accuracy of endoscopic image classification. This review focuses on the AI-assisted diagnosis of early ESCC and prediction of early ESCC invasion depth under multiple imaging modalities. The excellent image recognition ability of AI is suitable for the detection and diagnosis of ESCC and can reduce missed diagnoses and help endoscopists better complete endoscopic examinations. However, the selective bias used in the training dataset of the AI system affects its general utility.

Computer-aided characterization of early cancer in Barrett's esophagus on i-scan magnification imaging: a multicenter international study

BACKGROUND AND AIMS

We aimed to develop a computer-aided characterization system that could support the diagnosis of dysplasia in Barrett's esophagus (BE) on magnification endoscopy.

METHODS

Videos were collected in high-definition magnification white-light and virtual chromoendoscopy with i-scan (Pentax Hoya, Japan) imaging in patients with dysplastic and nondysplastic BE (NDBE) from 4 centers. We trained a neural network with a Resnet101 architecture to classify frames as dysplastic or nondysplastic. The network was tested on 3 different scenarios: high-quality still images, all available video frames, and a selected sequence within each video.

RESULTS

Fifty-seven patients, each with videos of magnification areas of BE (34 dysplasia, 23 NDBE), were included. Performance was evaluated by a leave-1-patient-out cross-validation method. In all, 60,174 (39,347 dysplasia, 20,827 NDBE) magnification video frames were used to train the network. The testing set included 49,726 i-scan-3/optical enhancement magnification frames. On 350 high-quality still images, the network achieved a sensitivity of 94%, specificity of 86%, and area under the receiver operator curve (AUROC) of 96%. On all 49,726 available video frames, the network achieved a sensitivity of 92%, specificity of 82%, and AUROC of 95%. On a selected sequence of frames per case (total of 11,471 frames), we used an exponentially weighted moving average of classifications on consecutive frames to characterize dysplasia. The network achieved a sensitivity of 92%, specificity of 84%, and AUROC of 96%. The mean assessment speed per frame was 0.0135 seconds (SD ± 0.006).

CONCLUSION

Our network can characterize BE dysplasia with high accuracy and speed on high-quality magnification images and sequence of video frames, moving it toward real-time automated diagnosis.

Linked color imaging provides enhanced visibility with a high color difference in upper gastrointestinal neoplasms

BACKGROUND AND AIM

The aim of this post-hoc analysis in a randomized, controlled, multicenter trial was to evaluate the visibility of upper gastrointestinal (UGI) neoplasms detected using linked color imaging (LCI) compared with those detected using white light imaging (WLI).

METHODS

The visibility of the detected UGI neoplasm images obtained using both WLI and LCI was subjectively reviewed, and the median color difference (ΔE) between each lesion and the surrounding mucosa according to the CIE L*a*b* color space was evaluated objectively. Multivariate logistic regression analysis was performed to identify factors associated with neoplasms that were missed under WLI and detected under LCI.

RESULTS

A total of 120 neoplasms, including 10, 32, and 78 neoplasms in the pharynx, esophagus, and stomach, respectively, were analyzed in this study. LCI enhanced the visibility 80.9% and 93.6% of neoplasms in pharynx/esophagus and stomach compared with WLI, respectively. LCI also achieved a higher ΔE of enhanced neoplasms compared with WLI in the pharynx/esophagus and stomach. The median WLI ΔE values for gastric neoplasms missed under WLI and later detected under LCI were significantly lower than those for gastric neoplasms detected under WLI (8.2 vs 9.6, respectively). Furthermore, low levels of WLI ΔE (odds ratio [OR], 7.215) and high levels of LCI ΔE (OR, 22.202) were significantly associated with gastric neoplasms missed under WLI and later detected under LCI.

CONCLUSION

Color differences were independently associated with missing gastric neoplasms under WLI, suggesting that LCI has an obvious advantage over WLI in enhancing neoplastic visibility.

Expected value of artificial intelligence in gastrointestinal endoscopy: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement

This ESGE Position Statement defines the expected value of artificial intelligence (AI) for the diagnosis and management of gastrointestinal neoplasia within the framework of the performance measures already defined by ESGE. This is based on the clinical relevance of the expected task and the preliminary evidence regarding artificial intelligence in artificial or clinical settings. MAIN RECOMMENDATIONS:: (1) For acceptance of AI in assessment of completeness of upper GI endoscopy, the adequate level of mucosal inspection with AI should be comparable to that assessed by experienced endoscopists. (2) For acceptance of AI in assessment of completeness of upper GI endoscopy, automated recognition and photodocumentation of relevant anatomical landmarks should be obtained in ≥90% of the procedures. (3) For acceptance of AI in the detection of Barrett's high grade intraepithelial neoplasia or cancer, the AI-assisted detection rate for suspicious lesions for targeted biopsies should be comparable to that of experienced endoscopists with or without advanced imaging techniques. (4) For acceptance of AI in the management of Barrett's neoplasia, AI-assisted selection of lesions amenable to endoscopic resection should be comparable to that of experienced endoscopists. (5) For acceptance of AI in the diagnosis of gastric precancerous conditions, AI-assisted diagnosis of atrophy and intestinal metaplasia should be comparable to that provided by the established biopsy protocol, including the estimation of extent, and consequent allocation to the correct endoscopic surveillance interval. (6) For acceptance of artificial intelligence for automated lesion detection in small-bowel capsule endoscopy (SBCE), the performance of AI-assisted reading should be comparable to that of experienced endoscopists for lesion detection, without increasing but possibly reducing the reading time of the operator. (7) For acceptance of AI in the detection of colorectal polyps, the AI-assisted adenoma detection rate should be comparable to that of experienced endoscopists. (8) For acceptance of AI optical diagnosis (computer-aided diagnosis [CADx]) of diminutive polyps (≤5 mm), AI-assisted characterization should match performance standards for implementing resect-and-discard and diagnose-and-leave strategies. (9) For acceptance of AI in the management of polyps ≥ 6 mm, AI-assisted characterization should be comparable to that of experienced endoscopists in selecting lesions amenable to endoscopic resection.

Application of an artificial intelligence system for endoscopic diagnosis of superficial esophageal squamous cell carcinoma

BACKGROUND

Upper gastrointestinal endoscopy is critical for esophageal squamous cell carcinoma (ESCC) detection; however, endoscopists require long-term training to avoid missing superficial lesions.

AIM

To develop a deep learning computer-assisted diagnosis (CAD) system for endoscopic detection of superficial ESCC and investigate its application value.

METHODS

We configured the CAD system for white-light and narrow-band imaging modes based on the YOLO v5 algorithm. A total of 4447 images from 837 patients and 1695 images from 323 patients were included in the training and testing datasets, respectively. Two experts and two non-expert endoscopists reviewed the testing dataset independently and with computer assistance. The diagnostic performance was evaluated in terms of the area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity.

RESULTS

The area under the receiver operating characteristics curve, accuracy, sensitivity, and specificity of the CAD system were 0.982 [95% confidence interval (CI): 0.969-0.994], 92.9% (95%CI: 89.5%-95.2%), 91.9% (95%CI: 87.4%-94.9%), and 94.7% (95%CI: 89.0%-97.6%), respectively. The accuracy of CAD was significantly higher than that of non-expert endoscopists (78.3%, P < 0.001 compared with CAD) and comparable to that of expert endoscopists (91.0%, P = 0.129 compared with CAD). After referring to the CAD results, the accuracy of the non-expert endoscopists significantly improved (88.2% vs 78.3%, P < 0.001). Lesions with Paris classification type 0-IIb were more likely to be inaccurately identified by the CAD system.

CONCLUSION

The diagnostic performance of the CAD system is promising and may assist in improving detectability, particularly for inexperienced endoscopists.

Prevalence and risk factors of upper gastrointestinal cancers missed during endoscopy: a nationwide registry-based study

BACKGROUND

A significant proportion of upper gastrointestinal cancers (UGICs) remain undetected during esophagogastroduodenoscopy (EGD). We investigated the characteristics and risk factors of UGICs missed during endoscopy.

METHODS

In this nationwide registry-based study, we analyzed two large Polish datasets (National Health Fund and National Cancer Registry) to identify individuals who underwent EGD and were subsequently diagnosed with UGIC. Cancers diagnosed < 6 months after EGD were defined as "prevalent" and those within ≥ 6- < 36 months as "missed." We compared the characteristics of missed and prevalent cancers, and analyzed the risk factors for missed UGICs in a multivariable regression model.

RESULTS

We included 4 105 399 patients (mean age 56.0 years [SD 17.4]; 57.5 % female) who underwent 5 877 674 EGDs in 2012-2018. Within this cohort, 33 241 UGICs were diagnosed, of which 1993 (6.0 %) were missed. Within esophageal neoplasms, adenocarcinomas were more frequently missed than squamous cell cancers (6.1 % vs. 4.2 %), with a relative risk of 1.4 (95 % confidence interval [CI] 1.1-1.8, P = 0.01). Most gastric cancers were adenocarcinomas, of which 5.7 % were classified as missed. Overall, a higher proportion of missed UGICs than prevalent cancers presented at an advanced stage (42.2 % vs. 36.2 %, P < 0.001). Risk factors for missed UGICs included initial EGD performed within primary (vs. secondary) care (odds ratio [OR] 1.3, 95 %CI 1.2-1.5), female sex (OR 1.3, 95 %CI 1.2-1.4), and higher comorbidity (Charlson comorbidity index ≥ 5 vs. 0; OR 6.0, 95 %CI 4.7-7.5).

CONCLUSIONS

Among UGICs, esophageal adenocarcinomas were missed most frequently. Missed cancers occur more frequently within the primary care sector and are found more often in women and individuals with multiple comorbidities.

The relation of esophagogastroduodenoscopy time and novel upper gastrointestinal quality measures

OBJECTIVES

Various measures of esophagogastroduodenoscopy (EGD) quality have been proposed so far and the examination time was one of the first. The aim of the study was to compare the procedure time with novel novel quality measures - composite detection rate (CDR) and endoscopist biopsy rate (EBR).

METHODS

It was prospective observational study. A total of 880 diagnostic EGDs conducted from 01.2019 to 07.2019 have been enrolled in the study.

RESULTS

Median EGD time was 4.2 min. Procedures of longer duration were marked with higher CDR (26.3% vs. 11.8%; P < 0.0001), higher EBR (44.9% vs. 12.3%; P < 0.0001), and better upper gastrointestinal neoplasm (UGN) detection (1.8% vs. 0%; P = 0.004) in comparison with procedures of shorter duration. The procedures were divided into 4 groups based on the time quartiles (group 1 <3.3 min; group 2 3.3-4.2 min; group 3 4.2-5.3 min; group 4 >5.3 min). The odds ratios of groups 2, 3, and 4 for biopsy rate were 2.42 (95% CI, 1.33-4.55), 4.33 (95% CI, 2.46-7.94), and 5.51 (95% CI, 3.18-10.03), respectively, in comparison with group 1. The odds ratios of groups 2, 3, and 4 for CDR were 3.18 (95% CI, 2.03-4.97), 5.46 (95% CI, 3.51-8.50), and 23.44 (95% CI, 14.3-38.4), respectively, in comparison with group 1.

CONCLUSIONS

The procedure time is related to novel metrics - CDR and EBR. It is also related to UGN. Based on our findings it could be concluded that EGD should not last less than 4.2 min.

Advances and horizons for artificial intelligence of endoscopic screening and surveillance of gastric and esophageal disease

The development of artificial intelligence in endoscopic assessment of the gastrointestinal tract has shown progressive enhancement in diagnostic acuity. This review discusses the expanding applications for gastric and esophageal diseases. The gastric section covers the utility of AI in detecting and characterizing gastric polyps and further explores prevention, detection, and classification of gastric cancer. The esophageal discussion highlights applications for use in screening and surveillance in Barrett's esophagus and in high-risk conditions for esophageal squamous cell carcinoma. Additionally, these discussions highlight applications for use in assessing eosinophilic esophagitis and future potential in assessing esophageal microbiome changes.

Artificial Intelligence in the Management of Barrett’s Esophagus and Early Esophageal Adenocarcinoma

Simple Summary

Esophageal adenocarcinoma is increasing in incidence and is the most common subtype of esophageal cancer in Western societies. AI systems are currently under development and validation in many fields of gas-troenterology.

Abstract

Esophageal adenocarcinoma is increasing in incidence and is the most common subtype of esophageal cancer in Western societies. The stepwise progression of Barrett´s metaplasia to high-grade dysplasia and invasive adenocarcinoma provides an opportunity for screening and surveillance. There are important unresolved issues, which include (i) refining the definition of the screening population in order to avoid unnecessary invasive diagnostics, (ii) a more precise prediction of the (very heterogeneous) individual progression risk from metaplasia to invasive cancer in order to better tailor surveillance recommendations, (iii) improvement of the quality of endoscopy in order to reduce the high miss rate for early neoplastic lesions, and (iv) support for the diagnosis of tumor infiltration depth in order to guide treatment decisions. Artificial intelligence (AI) systems might be useful as a support to better solve the above-mentioned issues.

Endoscopists' diagnostic accuracy in detecting upper gastrointestinal neoplasia in the framework of artificial intelligence studies

BACKGROUND

Estimates on miss rates for upper gastrointestinal neoplasia (UGIN) rely on registry data or old studies. Quality assurance programs for upper GI endoscopy are not fully established owing to the lack of infrastructure to measure endoscopists' competence. We aimed to assess endoscopists' accuracy for the recognition of UGIN exploiting the framework of artificial intelligence (AI) validation studies.

METHODS

Literature searches of databases (PubMed/MEDLINE, EMBASE, Scopus) up to August 2020 were performed to identify articles evaluating the accuracy of individual endoscopists for the recognition of UGIN within studies validating AI against a histologically verified expert-annotated ground-truth. The main outcomes were endoscopists' pooled sensitivity, specificity, positive and negative predictive value (PPV/NPV), and area under the curve (AUC) for all UGIN, for esophageal squamous cell neoplasia (ESCN), Barrett esophagus-related neoplasia (BERN), and gastric adenocarcinoma (GAC).

RESULTS

Seven studies (2 ESCN, 3 BERN, 1 GAC, 1 UGIN overall) with 122 endoscopists were included. The pooled endoscopists' sensitivity and specificity for UGIN were 82 % (95 % confidence interval [CI] 80 %-84 %) and 79 % (95 %CI 76 %-81 %), respectively. Endoscopists' accuracy was higher for GAC detection (AUC 0.95 [95 %CI 0.93-0.98]) than for ESCN (AUC 0.90 [95 %CI 0.88-0.92]) and BERN detection (AUC 0.86 [95 %CI 0.84-0.88]). Sensitivity was higher for Eastern vs. Western endoscopists (87 % [95 %CI 84 %-89 %] vs. 75 % [95 %CI 72 %-78 %]), and for expert vs. non-expert endoscopists (85 % [95 %CI 83 %-87 %] vs. 71 % [95 %CI 67 %-75 %]).

CONCLUSION

We show suboptimal accuracy of endoscopists for the recognition of UGIN even within a framework that included a higher prevalence and disease awareness. Future AI validation studies represent a framework to assess endoscopist competence.

Clinical and Endoscopic Characteristics Associated With Post-Endoscopy Upper Gastrointestinal Cancers: A Systematic Review and Meta-analysis

BACKGROUND & AIMS

Ten percent of patients with an upper gastrointestinal cancer will have received an esophagogastroduodenoscopy (EGD) within 3 years before diagnosis, termed post-endoscopy upper gastrointestinal cancers (PEUGIC). We aimed to determine the characteristics of PEUGIC, and compare these with detected cancers.

METHODS

We searched MEDLINE and Embase from inception for studies comparing the characteristics of PEUGIC and detected upper gastrointestinal cancers, and reported findings at the initial "cancer-negative" endoscopy. We synthesized results using random effects meta-analysis. This review is registered on PROSPERO, CRD42019125780.

RESULTS

A total of 2696 citations were screened and 25 studies were included, comprising 81,184 UGI cancers, of which 7926 were considered PEUGIC. For PEUGIC assessed within 6 to 36 months of a "cancer-negative" EGD, the mean interval was approximately 17 months. Patients with PEUGIC were less likely to present with dysphagia (odds ratio [OR] 0.37) and weight loss (OR 0.58) and were more likely to present with gastroesophageal reflux (OR 2.64) than detected cancers. PEUGICs were more common in women in Western populations (OR 1.30). PEUGICs were typically smaller at diagnosis and associated with less advanced disease staging compared with detected cancers (OR 2.87 for stage 1 vs 2-4). Most EGDs (>75%) were abnormal preceding diagnosis of PEUGIC.

CONCLUSIONS

There is a substantial delay in the diagnosis of PEUGIC. They are less likely to present with alarm symptoms than detected cancers. PEUGICs are overall less advanced at diagnosis. Most patients with PEUGIC have abnormalities reported at the preceding "cancer-negative" EGD. The epidemiology of PEUGIC may inform preventive strategy.

'Artificial intelligence in Barrett's Esophagus'

Despite advances in endoscopic imaging modalities, there are still significant miss rates of dysplasia and cancer in Barrett's esophagus. Artificial intelligence (AI) is a promising tool that may potentially be a useful adjunct to the endoscopist in detecting subtle dysplasia and cancer. Studies have shown AI systems have a sensitivity of more than 90% and specificity of more than 80% in detecting Barrett's related dysplasia and cancer. Beyond visual detection and diagnosis, AI may also prove to be useful in quality control, streamlining clinical work, documentation, and lessening the administrative load on physicians. Research in this area is advancing at a rapid rate, and as the field expands, regulations and guidelines will need to be put into place to better regulate the growth and use of AI. This review provides an overview of the present and future role of AI in Barrett's esophagus.

Recent advances in early esophageal cancer: diagnosis and treatment based on endoscopy

Esophageal cancer (EC) often cannot be discovered in time because of its asymptomatic or symptom-atypical characteristics in early stage. The risk and probability of lymph node metastasis and distant metastasis increase correspondingly as the cancer aggressively invades deeper layers. Treatment regimens may be shifted to surgery and chemoradiotherapy (CRT) from endoscopic eradication therapy (EET) with poor quality of life and prognosis. It is imperative to identify dysplasia and EC early and enable early curative endoscopic treatments. Newer methods have been attempted in the clinical setting to achieve early detection at a more microscopic and precise level. Newer imaging techniques and artificial intelligence (AI) technology have been involved in targeted biopsies and will gradually unveil the visualization of pathology in the future. Early detection and diagnosis are the prerequisite to choose personal and precise treatment regimens. EET has also been undergoing development and improvement to benefit more patients as the first option or the firstly chosen alternative therapy, when compared with esophagectomy. More clinical studies are needed to provide more possibilities for EET.

Artificial intelligence-assisted endoscopic detection of esophageal neoplasia in early stage: The next step?

Esophageal cancer (EC) is a common malignant tumor of the digestive tract and originates from the epithelium of the esophageal mucosa. It has been confirmed that early EC lesions can be cured by endoscopic therapy, and the curative effect is equivalent to that of surgical operation. Upper gastrointestinal endoscopy is still the gold standard for EC diagnosis. The accuracy of endoscopic examination results largely depends on the professional level of the examiner. Artificial intelligence (AI) has been applied in the screening of early EC and has shown advantages; notably, it is more accurate than less-experienced endoscopists. This paper reviews the application of AI in the field of endoscopic detection of early EC, including squamous cell carcinoma and adenocarcinoma, and describes the relevant progress. Although up to now most of the studies evaluating the clinical application of AI in early EC endoscopic detection are focused on still images, AI-assisted real-time detection based on live-stream video may be the next step.

Current Trends in Endoscopic Diagnosis and Treatment of Early Esophageal Cancer

Simple Summary

Early esophageal cancer is diagnosed in the context of reflux disease, surveillance of Barrett’s metaplasia, or during upper gastrointestinal endoscopy for other indications. High definition and virtual or dye chromoendoscopy are mandatory for the screening and evaluation of neoplasia. Endoscopic treatment options include endoscopic mucosal resection (EMR) or endoscopic submucosal dissection (ESD). Resection is considered curative if histopathology confirms low or absent risk of lymph node metastasis. Barrett’s high-grade dysplasia or early adenocarcinoma is treated by EMR or ESD, followed by ablation of Barrett’s epithelium to avoid metachronous cancer. ESD is the treatment of choice for squamous cell neoplasia. Excellent outcomes have been reported if the ESD of squamous cell cancer with slight submucosal infiltration and thus substantial risk for lymph node metastasis was combined with adjuvant chemo-radiotherapy. In contrast, infiltration of squamous cell cancer exceeding the lamina propria mucosae is not curative. However, despite a substantial risk of lymph node metastasis, excellent outcomes have recently been reported if endoscopic resection of tumors with up to 200 µm submucosal infiltration was combined with adjuvant chemo-radiotherapy.

Abstract

Diagnosis of esophageal adenocarcinoma mostly occurs in the context of reflux disease or surveillance of Barrett’s metaplasia. Optimal detection rates are obtained with high definition and virtual or dye chromoendoscopy. Smaller lesions can be treated with endoscopic mucosal resection. Endoscopic submucosal dissection (ESD) is an option for larger lesions. Endoscopic resection is considered curative (i.e., without significant risk of lymph node metastasis) if histopathology confirms en bloc and R0 resection of a well-differentiated (G1/2) tumor without infiltration of lymphatic or blood vessels and the maximal submucosal infiltration depth is 500µm. Ablation of remaining Barrett’s metaplasia is important, to reduce the risk of metachronous cancer. Esophageal squamous cell cancer is associated with different risk factors, and most of the detected lesions are diagnosed during upper gastrointestinal endoscopy for other indications. Virtual high definition and dye chromoendoscopy with Lugol’s solution are used for screening and evaluation. ESD is the preferred resection technique. The criteria for curative resection are similar to Barrett’s cancer, but the maximum infiltration depth must not exceed lamina propria mucosae. Although a submucosal infiltration depth of up to 200 µm carries a substantial risk of lymph node metastasis, ESD combined with adjuvant chemo-radiotherapy gives excellent results. The complication rates of endoscopic resection are low, and the functional outcomes are favorable compared to surgery.

Application of Artificial Intelligence in Gastrointestinal Endoscopy

Artificial intelligence (AI), also known as computer-aided diagnosis, is a technology that enables machines to process information and functions at or above human level and has great potential in gastrointestinal endoscopy applications. At present, the research on medical image recognition usually adopts the deep-learning algorithm based on the convolutional neural network. AI has been used in gastrointestinal endoscopy including esophagogastroduodenoscopy, capsule endoscopy, colonoscopy, etc. AI can help endoscopic physicians improve the diagnosis rate of various lesions, reduce the rate of missed diagnosis, improve the quality of endoscopy, assess the severity of the disease, and improve the efficiency of endoscopy. The diversity, susceptibility, and imaging specificity of gastrointestinal endoscopic images are all difficulties and challenges on the road to intelligence. We need more large-scale, high-quality, multicenter prospective studies to explore the clinical applicability of AI, and ethical issues need to be taken into account.

Comparative study on artificial intelligence systems for detecting early esophageal squamous cell carcinoma between narrow-band and white-light imaging

BACKGROUND

Non-magnifying endoscopy with narrow-band imaging (NM-NBI) has been frequently used in routine screening of esophagus squamous cell carcinoma (ESCC). The performance of NBI for screening of early ESCC is, however, significantly affected by operator experience. Artificial intelligence may be a unique approach to compensate for the lack of operator experience.

AIM

To construct a computer-aided detection (CAD) system for application in NM-NBI to identify early ESCC and to compare it with our previously reported CAD system with endoscopic white-light imaging (WLI).

METHODS

A total of 2167 abnormal NM-NBI images of early ESCC and 2568 normal images were collected from three institutions (Zhongshan Hospital of Fudan University, Xuhui Hospital, and Kiang Wu Hospital) as the training dataset, and 316 pairs of images, each pair including images obtained by WLI and NBI (same part), were collected for validation. Twenty endoscopists participated in this study to review the validation images with or without the assistance of the CAD systems. The diagnostic results of the two CAD systems and improvement in diagnostic efficacy of endoscopists were compared in terms of sensitivity, specificity, accuracy, positive predictive value, and negative predictive value.

RESULTS

The area under receiver operating characteristic curve for CAD-NBI was 0.9761. For the validation dataset, the sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of CAD-NBI were 91.0%, 96.7%, 94.3%, 95.3%, and 93.6%, respectively, while those of CAD-WLI were 98.5%, 83.1%, 89.5%, 80.8%, and 98.7%, respectively. CAD-NBI showed superior accuracy and specificity than CAD-WLI (P = 0.028 and P ≤ 0.001, respectively), while CAD-WLI had higher sensitivity than CAD-NBI (P = 0.006). By using both CAD-WLI and CAD-NBI, the endoscopists could improve their diagnostic efficacy to the highest level, with accuracy, sensitivity, and specificity of 94.9%, 92.4%, and 96.7%, respectively.

CONCLUSION

The CAD-NBI system for screening early ESCC has higher accuracy and specificity than CAD-WLI. Endoscopists can achieve the best diagnostic efficacy using both CAD-WLI and CAD-NBI.

Artificial intelligence for non-polypoid colorectal neoplasms

The miss rate of flat advanced colorectal neoplasia is still unacceptably high, especially in the Western setting, notwithstanding the widespread implementation of quality improvement programs and training. It is well known that flat morphology is associated with miss rate of colorectal neoplasia, and that this subset of lesions often shows a more aggressive biological behaviour. Artificial intelligence (AI) applied to the detection of colorectal neoplasia has been shown to increase adenoma detection rate, consistently across all lesion sizes and locations in the colon. However, there is still uncertainty whether AI can reduce the miss rate of flat advanced neoplasia, mainly because all published trials report a low number of flat colorectal lesions in their training sets, and this could reduce AI accuracy for this subset of lesions. In addition, flat lesions have different morphologies with variable prevalence and potentially different accuracy in their detection. For example, the subtle appearance and rarer frequency of a non-granular laterally spreading tumor (LST) could be much harder to identify than a granular mixed LST. In this review, we present a summary of the evidence on the role of AI in the identification of colorectal flat neoplasia, with a focus on data regarding presence of LSTs in the training/validation sets of the AI systems currently available on the market.

Artificial intelligence in upper GI endoscopy - current status, challenges and future promise

White-light endoscopy with biopsy is the current gold standard modality for detecting and diagnosing upper gastrointestinal (GI) pathology. However, missed lesions remain a challenge. To overcome interobserver variability and learning curve issues, artificial intelligence (AI) has recently been introduced to assist endoscopists in the detection and diagnosis of upper GI neoplasia. In contrast to AI in colonoscopy, current AI studies for upper GI endoscopy are smaller pilot studies. Researchers currently lack large volume, well-annotated, high-quality datasets in gastric cancer, dysplasia in Barrett's esophagus and early esophageal squamous cell cancer. This review will look at the latest studies of AI in upper GI endoscopy, discuss some of the challenges facing researchers, and predict what the future may hold in this rapidly changing field.

Composite detection rate as an upper gastrointestinal endoscopy quality measure correlating with detection of neoplasia

BACKGROUND

Esophagogastroduodenoscopy (EGD) is commonly used diagnostic method with no widely accepted quality measure. We assessed quality indicator-composite detection rate (CDR)-consisting of detection of at least one of the following: cervical inlet patch, gastric polyp and post-ulcer duodenal bulb deformation. The aim of the study was to validate CDR according to detection rate of upper gastrointestinal neoplasms (UGN).

METHODS

It was a multicenter, prospective, observational study conducted from January 2019 to October 2019. The endoscopic reports from 2896 symptomatic patients who underwent diagnostic EGD were analyzed. The EGDs were performed in three endoscopy units located in tertiary university hospital, private outpatient clinic and local hospital.

RESULTS

64 UGNs were detected. The mean CDR was 21.9%. The CDR correlated with UGN detection rate (R = 0.49, p = 0.045). Based on CDR quartiles, operators were divided into group 1 with CDR < 10%, group 2 with CDR 10-17%, group 3 with CDR 17.1-26%, and group 4 with CDR > 26%. Detection rate of UGN was significantly higher in the group 4 in comparison to group 1 (OR 4.4; 95% CI 2.2 - 9.0). In the multivariate regression model, patient age, male gender and operator's CDR > 26% were independent risk factors of UGN detection (OR 1.03; 95% CI 1.01 - 1.05, OR 2; 95% CI 1.2 - 3.5, and OR 5.7 95% CI 1.5 - 22.3, respectively).

CONCLUSIONS

The CDR is associated with the detection of upper gastrointestinal neoplasms. This parameter may be a useful quality measure of EGD to be applied in general setting.

Esophageal squamous dysplasia and cancer: Is artificial intelligence our best weapon?

Esophageal cancer is the eight most common cancer in the world and is associated with a poor prognosis. Significant efforts are necessary to improve the detection of early squamous cell cancer such that curative endoscopic therapy can be offered. Studies have shown an overall miss rate of esophageal cancer of up to 6.4%. Human factors including fatigue and lack of attention may be a contributory factor. Computer aided detection and characterisation of early squamous cell cancer can be a second reader which potentially offsets these factors. Recent studies developing artificial intelligence systems show real promise in the detection of early squamous cell cancer and predicting depth of invasion to aid in the management of patients in the same endoscopic session. This has the potential to revolutionise this area of endoscopy.

Systematic review with meta-analysis: age-related malignancy detection rates at upper gastrointestinal endoscopy

BACKGROUND

Age is an important and objective risk factor for upper gastrointestinal (GI) malignancy. The accuracy of various age limits to detect upper GI malignancy is unclear. Determination of this accuracy may aid in the decision to refer symptomatic patients for upper GI endoscopy. The aim of this analysis was to synthesize data on upper GI malignancy detection rates for various age limits worldwide through meta-analysis.

METHODS

We searched MEDLINE, EMBASE, and Web of Science in November 2018. Selection criteria included studies addressing malignant findings at upper GI endoscopy in a symptomatic population reporting age at time of diagnosis. Meta-analyses were conducted to derive continent-specific cancer detection rates.

RESULTS

A total of 33 studies including 346,641 patients across 21 countries fulfilled the inclusion criteria. To detect >80% of malignant cases all symptomatic patients over 40 years of age should be investigated in Africa, over 50 years of age in South America and Asia, and over 55 years of age in North America and Europe.

CONCLUSION

This systematic review and meta-analysis provides data on intercontinental variation in age at time of upper GI malignancy diagnosis in symptomatic patients referred for upper GI endoscopy. Guideline recommendations for age-based selection should be tailored to local age-related detection rates.

Standalone performance of artificial intelligence for upper GI neoplasia: a meta-analysis

OBJECTIVE

Artificial intelligence (AI) may reduce underdiagnosed or overlooked upper GI (UGI) neoplastic and preneoplastic conditions, due to subtle appearance and low disease prevalence. Only disease-specific AI performances have been reported, generating uncertainty on its clinical value.

DESIGN

We searched PubMed, Embase and Scopus until July 2020, for studies on the diagnostic performance of AI in detection and characterisation of UGI lesions. Primary outcomes were pooled diagnostic accuracy, sensitivity and specificity of AI. Secondary outcomes were pooled positive (PPV) and negative (NPV) predictive values. We calculated pooled proportion rates (%), designed summary receiving operating characteristic curves with respective area under the curves (AUCs) and performed metaregression and sensitivity analysis.

RESULTS

Overall, 19 studies on detection of oesophageal squamous cell neoplasia (ESCN) or Barrett's esophagus-related neoplasia (BERN) or gastric adenocarcinoma (GCA) were included with 218, 445, 453 patients and 7976, 2340, 13 562 images, respectively. AI-sensitivity/specificity/PPV/NPV/positive likelihood ratio/negative likelihood ratio for UGI neoplasia detection were 90% (CI 85% to 94%)/89% (CI 85% to 92%)/87% (CI 83% to 91%)/91% (CI 87% to 94%)/8.2 (CI 5.7 to 11.7)/0.111 (CI 0.071 to 0.175), respectively, with an overall AUC of 0.95 (CI 0.93 to 0.97). No difference in AI performance across ESCN, BERN and GCA was found, AUC being 0.94 (CI 0.52 to 0.99), 0.96 (CI 0.95 to 0.98), 0.93 (CI 0.83 to 0.99), respectively. Overall, study quality was low, with high risk of selection bias. No significant publication bias was found.

CONCLUSION

We found a high overall AI accuracy for the diagnosis of any neoplastic lesion of the UGI tract that was independent of the underlying condition. This may be expected to substantially reduce the miss rate of precancerous lesions and early cancer when implemented in clinical practice.

A Gratifying Step forward for the Application of Artificial Intelligence in the Field of Endoscopy: A Narrative Review

Endoscopy is the optimal choice of diagnosis of gastrointestinal (GI) diseases. Following the advancements made in medical technology, different kinds of novel endoscopy-methods have emerged. Although the significant progress in the penetration of endoscopic tools that have markedly improved the diagnostic rate of GI diseases, there are still some limitations, including instability of human diagnostic performance caused by intensive labor burden and high missed diagnosis rate of subtle lesions. Recently, artificial intelligence (AI) has been applied gradually to assist endoscopists in addressing these issues.

Role of artificial intelligence in the diagnosis of oesophageal neoplasia: 2020 an endoscopic odyssey

The past decade has seen significant advances in endoscopic imaging and optical enhancements to aid early diagnosis. There is still a treatment gap due to the underdiagnosis of lesions of the oesophagus. Computer aided diagnosis may play an important role in the coming years in providing an adjunct to endoscopists in the early detection and diagnosis of early oesophageal cancers, therefore curative endoscopic therapy can be offered. Research in this area of artificial intelligence is expanding and the future looks promising. In this review article we will review current advances in artificial intelligence in the oesophagus and future directions for development.

Survival outcomes and rate of missed upper gastrointestinal cancers at routine endoscopy: a single centre retrospective cohort study

OBJECTIVES

Data on the long-term survival outcome of patients with missed upper gastrointestinal cancers (MUGC) is lacking. Retrospective studies have found no difference in 1- and 2-year survival among patients with missed gastric and oesophageal cancers; we thus aimed to assess 3-year survival of patients with MUGC at oesophagogastroduodenoscopy.

METHODS

This was a retrospective cohort study conducted at a single tertiary endoscopy centre. All oesophagogastroduodenoscopies performed between January 2007 and December 2015 were included in the study. The endoscopy database was cross-matched with the Slovenian Cancer Registry database. Missed cancers were defined as those diagnosed within 36 months after a negative oesophagogastroduodenoscopy.

RESULTS

During the study period, 29 617 oesophagogastroduodenoscopies were performed. In total, 422 upper gastrointestinal cancers were diagnosed and the rate of missed gastric cancers was 7.3% (95% CI, 4.9-10.6%) (26/354), and 4.4% (95% CI, 0.9-12.4%) for oesophageal cancers (3/68). Three-year survival of patients with MUGC was shorter than that of those with non-MUGC, being 12% (95% CI, 1-25%) vs. 31% (95% CI, 26-36%) (P = 0.043) for gastric and 0 vs. 9% (95% CI, 1-17%) (P = 0.121) for oesophageal cancer.

CONCLUSION

Missed gastric cancer during oesophagogastroduodenoscopy may be associated with shorter 3-year survival compared to patients whose gastric cancer was diagnosed at index oesophagogastroduodenoscopy.

Automatic classification of esophageal lesions in endoscopic images using a convolutional neural network

Background

Using deep learning techniques in image analysis is a dynamically emerging field. This study aims to use a convolutional neural network (CNN), a deep learning approach, to automatically classify esophageal cancer (EC) and distinguish it from premalignant lesions.

Methods

A total of 1,272 white-light images were adopted from 748 subjects, including normal cases, premalignant lesions, and cancerous lesions; 1,017 images were used to train the CNN, and another 255 images were examined to evaluate the CNN architecture. Our proposed CNN structure consists of two subnetworks (O-stream and P-stream). The original images were used as the inputs of the O-stream to extract the color and global features, and the pre-processed esophageal images were used as the inputs of the P-stream to extract the texture and detail features.

Results

The CNN system we developed achieved an accuracy of 85.83%, a sensitivity of 94.23%, and a specificity of 94.67% after the fusion of the 2 streams was accomplished. The classification accuracy of normal esophagus, premalignant lesion, and EC were 94.23%, 82.5%, and 77.14%, respectively, which shows a better performance than the Local Binary Patterns (LBP) + Support Vector Machine (SVM) and Histogram of Gradient (HOG) + SVM methods. A total of 8 of the 35 (22.85%) EC lesions were categorized as premalignant lesions because of their slightly reddish and flat lesions.

Conclusions

The CNN system, with 2 streams, demonstrated high sensitivity and specificity with the endoscopic images. It obtained better detection performance than the currently used methods based on the same datasets and has great application prospects in assisting endoscopists to distinguish esophageal lesion subclasses.

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.