Diagnosis of gastric cancer up to three years after negative upper gastrointestinal endoscopy

The Significance of Screening Endoscopy at 2-Year Interval for Detecting Pharyngeal and Esophageal Cancer in the Japanese General Population

PURPOSE

We previously identified examination time and 2-year interval as effective quality indicators for detecting upper gastrointestinal cancer during screening esophagogastroduodenoscopy (EGD). We evaluated the significance of biennial endoscopy in detecting pharyngeal and esophageal cancer.

METHODS

Individuals who underwent multiple EGD examinations as health checkups at our center between August 2012 and January 2022 were included. Endoscopists were categorized into two groups based on their mean examination time for normal EGD findings, using a threshold of 5 min. The associations between cancer detection rate and these indicators were analyzed using multivariable analyses.

RESULTS

Twelve endoscopists performed more than 1000 examinations each, for a total of 28,842 examinations, detecting all (6 pharyngeal and 26 esophageal) cancers. Intervals between endoscopies were classified into three groups: ≤ 1 year (Group 1), 1 to ≤ 2 years (Group 2), and > 2 years (Group 3). This cancer detection rate did not differ significantly between Groups 1 and 2 (p = 0.761) but was significantly higher in Group 3 than in Group 2 (OR = 4.356, 95% CI = 1.161-16.333, p = 0.029). When Groups 1 and 2 were combined (≤ 2 years) and compared with Group 3, an interval of > 2 years was significantly associated with a higher detection rate, including cases not curable by endoscopic resection (OR, 4.710; 95% CI, 1.806-12.282; p = 0.002). Examination time was not associated with the cancer detection rate (p = 0.944).

CONCLUSIONS

These findings support the significance of biennial endoscopy in detecting pharyngeal and esophageal cancer as well.

Endoscopists and endoscopic assistants' qualifications, but not their biopsy rates, improve gastric precancerous lesions detection rate

BACKGROUND

Detecting gastric precancerous lesions (GPLs) is critical for the early diagnosis and treatment of gastric cancer. Endoscopy combined with tissue examination is an important method for detecting GPLs. However, negative biopsy results often increase patients' risks, economic burdens, and lead to additional healthcare costs. Improving the detection rate of GPLs and reducing the rate of negative biopsies is currently a key focus in endoscopic quality control.

AIM

To explore the relationships between the endoscopist biopsy rate (EBR), qualifications of endoscopists and endoscopic assistants, and detection rate of GPLs.

METHODS

EBR, endoscopists, and endoscopic assistants were divided into four groups: Low, moderate, high, and very high levels. Multivariable logistic regression analysis was used to analyze the relationships between EBR and the qualifications of endoscopists with respect to the detection rate of positive lesions. Pearson and Spearman correlation analyses were used to evaluate the correlation between EBR, endoscopist or endoscopic assistant qualifications, and the detection rate of positive lesions.

RESULTS

Compared with those in the low EBR group, the odds ratio (OR) values for detecting positive lesions in the moderate, high, and very high EBR groups were 1.12 [95% confidence interval (CI): 1.06-1.19, P < 0.001], 1.22 (95%CI: 1.14-1.31, P < 0.001), and 1.38 (95%CI: 1.29-1.47, P < 0.001), respectively. EBR was positively correlated with the detection rate of gastric precancerous conditions (atrophic gastritis/intestinal metaplasia) (ρ = 0.465, P = 0.004). In contrast, the qualifications of the endoscopists were positively correlated with GPLs detection (ρ = 0.448, P = 0.005). Compared to endoscopists with low qualification levels, those with moderate, high, and very high qualification levels endoscopists demonstrated increased detection rates of GPLs by 13% (OR = 1.13, 95%CI: 0.98-1.31), 20% (OR = 1.20, 95%CI: 1.03-1.39), and 32% (OR = 1.32, 95%CI: 1.15-1.52), respectively. Further analysis revealed that the qualifications of endoscopists were positively correlated with the detection rates of GPLs in the cardia (ρ = 0.350, P = 0.034), angularis (ρ = 0.396, P = 0.015) and gastric body (ρ = 0.453, P = 0.005) but not in the antrum (ρ = 0.292, P = 0.079). Moreover, the experience of endoscopic assistants was positively correlated with the detection rate of precancerous lesions by endoscopists with low or moderate qualifications (ρ = 0.427, P = 0.015).

CONCLUSION

Endoscopists and endoscopic assistants with high/very high qualifications, but not EBR, can improve the detection rate of GPLs. These results provide reliable evidence for the development of gastroscopic quality control indicators.

Artificial intelligence-assisted esophagogastroduodenoscopy improves procedure quality for endoscopists in early stages of training

Background and study aims

Completeness of esophagagogastroduodenoscopy (EGD) varies among endoscopists, leading to a high miss rate for gastric neoplasms. This study aimed to determine the effect of the Cerebro real-time artificial intelligence (AI) system on completeness of EGD for endoscopists in early stages of training.

Patients and methods

The AI system was built with CNN and Motion Adaptive Temporal Feature Aggregation (MA-TFA). A prospective sequential cohort study was conducted. Endoscopists were taught about the standardized EGD protocol to examine 27 sites. Then, each subject performed diagnostic EGDs per protocol (control arm). After completion of the required sample size, subjects performed diagnostic EGDs with assistance of the AI (study arm). The primary outcome was the rate of completeness of EGD. Secondary outcomes included overall inspection time, individual site inspection time, completeness of photodocumentation, and rate of positive pathologies.

Results

A total of 466 EGDs were performed with 233 in each group. Use of AI significantly improved completeness of EGD [mean (SD) (92.6% (6.2%) vs 71.2% (16.8%)]; P <0.001 (95% confidence interval 19.2%-23.8%, SD 0.012). There was no difference in overall mean (SD) inspection time [765.5 (338.4) seconds vs 740.4 (266.2); P =0.374]. Mean (SD) number of photos for photo-documentation significantly increased in the AI group [26.9 (0.4) vs 10.3 (4.4); P <0.001]. There was no difference in detection rates for pathologies in the two groups [8/233 (3.43%) vs 5/233 (2.16%), P =0.399].

Conclusions

Completeness of EGD examination and photodocumentation by endoscopists in early stages of are improved by the AI-assisted software Cerebro.

Role of Artificial Intelligence in the Detection and Management of Premalignant and Malignant Lesions of the Esophagus and Stomach

The advent of artificial intelligence (AI) and deep learning algorithms, particularly convolutional neural networks, promises to address pitfalls, bridging the care for patients at high risk with improved detection (computer-aided detection [CADe]) and characterization (computer-aided diagnosis [CADx]) of lesions. This review describes the available artificial intelligence (AI) technology and the current data on AI tools for screening esophageal squamous cell cancer, Barret's esophagus-related neoplasia, and gastric cancer. These tools outperformed endoscopists in many situations. Recent randomized controlled trials have demonstrated the successful application of AI tools in clinical practice with improved outcomes.

Effectiveness of Artificial Intelligence in Screening Esophagogastroduodenoscopy

BACKGROUND

This retrospective study investigated the usefulness of the EW10-EG01 artificial intelligence (AI) application for screening esophagogastroduodenoscopy (EGD).

METHODOLOGY

A total of 7,655 subjects (4,863 men, 2,792 women; mean age 54.9±10.1 years) who underwent EGD during a medical checkup were enrolled in the study. The number of diagnosed upper gastrointestinal tumors was compared between EGD examinations performed with and without the AI system.

RESULTS

EGD examinations with and without the AI system were performed on 3,841 and 3,814 subjects, respectively. Biopsy procedures were more frequently performed, and examination time was longer in EGD with AI applications than in those without. Upper gastrointestinal tumors diagnosed by EGD with and without AI were 39 (1.02%) and 24 (0.63%), respectively (P = 0.062). There was a significant difference in the detection rate of esophageal and gastric tumors between EGD with (30, 0.78%) and without (14, 0.37%) the AI system (P = 0.017). When endoscopists were divided into three groups based on their experience with the EW10-EG01 application, higher detection rates of esophageal and gastric tumors were observed in each group when using EGD with AI.

CONCLUSIONS

Usage of the EW10-EG01 AI system may be useful for screening EGD due to the increased esophageal and gastric tumor detection rate.

GastroHUN an Endoscopy Dataset of Complete Systematic Screening Protocol for the Stomach

Endoscopy is vital for detecting and diagnosing gastrointestinal diseases. Systematic examination protocols are key to enhancing detection, particularly for the early identification of premalignant conditions. Publicly available endoscopy image databases are crucial for machine learning research, yet challenges persist, particularly in identifying upper gastrointestinal anatomical landmarks to ensure effective and precise endoscopic procedures. However, many existing datasets have inconsistent labeling and limited accessibility, leading to biased models and reduced generalizability. This paper introduces GastroHUN, an open dataset documenting stomach screening procedures based on a systematic protocol. GastroHUN includes 8,834 images from 387 patients and 4,729 labeled video sequences, all annotated by four experts. The dataset covers 22 anatomical landmarks in the stomach and includes an additional category for unqualified images, making it a valuable resource for AI model development. By providing a robust public dataset and baseline deep learning models for image and sequence classification, GastroHUN serves as a benchmark for future research and aids in the development of more effective algorithms.

Diagnostic performance of blue laser imaging for early detection of gastric cancer: A systematic review and meta-analysis

BACKGROUND

Gastric cancer (GC) is associated with a significant global health burden and high mortality rates when diagnosed at later stages. The diagnosis often occurs at advanced stages when treatment options are limited and less effective. Early detection strategies are crucial to improving survival rates and outcomes for patients. Blue laser imaging (BLI) is an image-enhanced endoscopy technique that utilizes white light and narrow-band light to detect pathological changes in the mucosal architecture. This study aims at investigating the diagnostic performance of BLI for the detection of GC.

METHODS

A comprehensive search was conducted across multiple databases from inception until March 2023. Studies assessing the diagnostic efficacy of BLI for GC detection were included. The sensitivity, specificity and accuracy of BLI were calculated using pooled proportions and 95% confidence intervals (CI) with a random-effects model. Heterogeneity among the included studies was assessed using the I2 statistic.

RESULTS

Six studies were included in the pooled analysis. There were 708 patients with 380 GC lesions. Most of the lesions involved the lower two-thirds of the stomach. The pooled performance metrics of BLI for GC detection were as follows: sensitivity of 91.9% (95% CI 83.3-96.3%; I2 = 82.3%), specificity of 93.4% (95% CI 82.0-97.8%; I2 = 87.9%) and accuracy of 95.4% (95% CI 72.6-99.8%; I2 = 73.6%).

CONCLUSION

BLI demonstrates high diagnostic efficacy for the detection of GC. BLI can be a valuable tool in clinical practice. However, large-scale, randomized controlled studies are needed to further establish the role of BLI in routine clinical practice for GC detection.

Effect of texture and color enhancement imaging on the visibility of gastric tumors

Texture and color enhancement imaging (TXI) may improve the visibility of gastric tumors and allow their early detection. However, few reports have examined the utility of TXI. Between June 2021 and October 2022, 56 gastric tumors in 51 patients undergoing endoscopic submucosal dissection at Fukuchiyama City Hospital were evaluated preoperatively using conventional white light imaging (WLI), narrow-band imaging (NBI), and TXI modes 1 and 2. The color differences of the tumors and surrounding mucosae were evaluated using the CIE 1976 L*a*b color space, Additionally, the visibility scores were scaled. Of the 56 gastric tumors, 45 were early gastric cancers, and 11 were adenomas. Overall, the color difference in TXI mode 1 was considerably higher compared to WLI (16.36 ± 7.05 vs. 10.84 ± 4.05; p < 0.01). Moreover, the color difference in early gastric cancers was considerably higher in TXI mode 1 compared to WLI, whereas no significant difference was found in adenomas. The visibility score in TXI mode 1 was the highest, and it was significantly higher compared to WLI. Regarding adenomas, the visibility score in TXI mode 1 was also significantly higher compared to that in WLI. TXI may provide improved gastric tumor visibility.

Optimal number of images and 2-year interval affect cancer detection in screening esophagogastroduodenoscopy: An observational study

We aimed to identify quality indicator for esophagogastroduodenoscopy for detecting upper gastrointestinal (UGI) cancer. Data from 43,526 consecutive health checkups from August 2012 to January 2022 were retrospectively collected. The study ultimately analyzed 42,387 examinations by 12 endoscopists who performed more than 1000 examinations, including all cancers detected. These endoscopists were classified either into fast/slow group based on their mean examination time for a normal finding of esophagogastroduodenoscopy during their first year of the examination, or small/large group based on number of endoscopic images, respectively. The association between UGI cancer detection rate and examination time or the number of images was analyzed, using 5 minutes or 50 images as cutoff values. The detection rate of overall (8 pharyngeal, 39 esophageal, 69 gastric) cancers in the fast, slow, small, and large groups were 0.17%, 0.32%, 0.21%, and 0.31%, respectively. On multivariable analysis, endoscopists in the fast group or the small group were less likely to detect overall UGI cancer (OR: 0.596, 95% CI: 0.373-0.952, P = .030; OR: 0.652, 95% CI: 0.434-0.979, P = .039). Additionally, repeated endoscopy within 2 years had a higher overall cancer detection rate, compared with repeated screening after 2 years. In a sub-analysis, a significant negative relationship was found between the detection rate of gastric cancer and the number of gastric images < 35 (OR: 0.305, 95% CI: 0.189-0.492, P = .000). There was also a negative correlation trend between the detection rate of pharyngeal and esophageal cancers and the number of esophageal images < 11 (OR: 0.395, 95% CI: 0.156-1.001, P = .050). The optimal number of images and screening 2-year interval are considered useful quality indicators for detecting UGI cancer. This study also suggests that a total of 50 images, or 35 images of the stomach are suitable for detecting UGI cancer, or gastric cancer, during screening endoscopy.

CG‐Net: A novel CNN framework for gastrointestinal tract diseases classification

The classification of medical images has had a significant influence on the diagnostic techniques and therapeutic interventions. Conventional disease diagnosis procedures require a substantial amount of time and effort to accurately diagnose. Based on global statistics, gastrointestinal cancer has been recognized as a major contributor to cancer‐related deaths. The complexities involved in resolving gastrointestinal tract (GIT) ailments arise from the need for elaborate methods to precisely identify the exact location of the problem. Therefore, doctors frequently use wireless capsule endoscopy to diagnose and treat GIT problems. This research aims to develop a robust framework using deep learning techniques to effectively classify GIT diseases for therapeutic purposes. A CNN based framework, in conjunction with the feature selection method, has been proposed to improve the classification rate. The proposed framework has been evaluated using various performance measures, including accuracy, recall, precision, F1 measure, mean absolute error, and mean squared error.

Comparison of the Prognosis of Upper-Third Gastric Cancer With That of Middle and Lower-Third Gastric Cancer

PURPOSE

Gastric cancer is one of the most common cancers in Korea, and the proportion of upper-third gastric cancers has been steadily increasing over the last two decades. This study aimed to evaluate the effect of tumor location on gastric cancer prognosis.

MATERIALS AND METHODS

We retrospectively reviewed 2,466 patients who underwent gastrectomy for pathologically proven gastric cancer between January 2011 and December 2016. The patients were divided into an upper-third group (U group; n=419, 17.0%) and a middle- and lower-third group (ML group; n=2,047, 83.0%). Clinicopathological characteristics, overall survival (OS), and recurrence-free survival (RFS) after surgery were compared.

RESULTS

The U group had more advanced disease than the ML group and a higher incidence of N3b disease for T3 (12.0% vs. 4.9%, p=0.023) and T4 tumors (33.3% vs. 17.5%, p=0.001). The 5-year RFS rate for stage III disease was marginally lower in the U group than that in the ML group (47.1% vs. 56.7%, p=0.082). The upper third location was an independent prognostic factor for both OS (hazard ratio [HR], 1.350; 95% confidence interval [CI], 1.065-1.711) and RFS (HR, 1.430; 95% CI, 1.080-1.823).

CONCLUSIONS

Upper-third gastric cancer shows extensive node metastasis compared to those located more distally in ≥T3 tumors. The upper third location is an independent prognostic factor for both OS and RFS and may have an adverse impact on RFS, particularly in patients with stage III gastric cancer.

Diagnostic performance of linked color imaging for gastric cancer by Helicobacter pylori infection status: A subanalysis of the large-scale, multicenter randomized controlled trial LCI-FIND

BACKGROUND

Linked color imaging (LCI) is a new image enhancement technology that facilitates the recognition of subtle differences in mucosal color. In the large-scale, multicenter randomized controlled trial LCI-FIND, LCI demonstrated good diagnostic performance for the detection of tumor lesions in the upper gastrointestinal tract. The aim of the present study was to exploratively evaluate the diagnostic performance of LCI according to H. pylori infection status as a subanalysis of LCI-FIND trial.

METHODS

The patients were randomly allocated to receive white light imaging (WLI) first, followed by LCI (WLI group), or vice versa (LCI group), and the two groups were compared for the detection of tumors. Data from this trial were analyzed by the presence/absence of H. pylori infection and further analyzed by successful or unsuccessful eradication in the H. pylori infection group.

RESULTS

The 752 patients in the WLI group and 750 patients in the LCI group who had participated in the LCI-FIND trial were included. In the successful eradication group, more gastric lesions were detected by primary mode in the LCI group than in the WLI group, indicating that more lesions were missed by WLI. Fisher's exact probability test for the comparison of the WLI and LCI groups yielded a p-value of 0.0068, with missed gastric lesions being detected 0.136 times (95% confidence interval: 0.020-0.923), significantly less with LCI than with WLI.

CONCLUSION

The current study suggests that LCI should be used for gastric cancer screening, particularly in patients with successful H. pylori eradication.

Relationship between observation time and detection rate of focal lesions in Esophagogastroduodenoscopy: a single-center, retrospective study

BACKGROUND

Current studies have shown that longer observation time can improve neoplastic detection rate. This study aimed to clarify whether endoscopists with longer observation times can detect more focal lesions.

METHODS

Based on the mean examination time for Esophagogastroduodenoscopy (EGD) without biopsy, endoscopists were divided into fast and slow groups, and the detection rate of focal lesions was compared between the two groups. Univariate analysis, multivariate analysis and restricted cubic spline were used to explore the factors of focal lesion detection rate.

RESULTS

Mean examination time of EGD without biopsy was 4.5 min. The cut-off times used were 5 min. 17 endoscopists were classified into the fast (4.7 ± 3.6 min), and 16 into the slow (7.11 ± 4.6 min) groups. Compared with fast endoscopists, slow endoscopists had a higher detection rate of focal lesions (47.2% vs. 51.4%, P < 0.001), especially in the detection of gastric lesions (29.7% vs. 35.9%, P < 0.001). In univariate and multivariate analyses, observation time, patient age and gender, expert, biopsy rate, and number of images were factors in FDR. There is a nonlinear relationship between observation time and FDR.

CONCLUSION

Longer examination time improves the detection rate of focal lesions. Observation time is an important quality indicator of the EGD examination.

Comparing artificial intelligence to humans for endoscopic diagnosis of gastric neoplasia: An external validation study

OBJECTIVES

Artificial intelligence (AI) uses deep learning functionalities that may enhance the detection of early gastric cancer during endoscopy. An AI-based endoscopic system for upper endoscopy was recently developed in Japan. We aim to validate this AI-based system in a Singaporean cohort.

METHODS

There were 300 de-identified still images prepared from endoscopy video files obtained from subjects that underwent gastroscopy in National University Hospital (NUH). Five specialists and 6 non-specialists (trainees) from NUH were assigned to read and categorize the images into "neoplastic" or "non-neoplastic." Results were then compared with the readings performed by the endoscopic AI system.

RESULTS

The mean accuracy, sensitivity, and specificity for the 11 endoscopists were 0.847, 0.525, and 0.872, respectively. These values for the AI-based system were 0.777, 0.591, and 0.791, respectively. While AI in general did not perform better than endoscopists on the whole, in the subgroup of high-grade dysplastic lesions, only 29.1% were picked up by the endoscopist rating, but 80% were classified as neoplastic by AI (P = 0.0011). The average diagnostic time was also faster in AI compared with endoscopists (677.1 s vs 42.02 s (P < 0.001).

CONCLUSION

We demonstrated that an AI system developed in another health system was comparable in diagnostic accuracy in the evaluation of static images. AI systems are faster and not fatigable and may have a role in augmenting human diagnosis during endoscopy. With more advances in AI and larger studies to support its efficacy it would likely play a larger role in screening endoscopy in future.

Post-Endoscopy Upper Gastrointestinal Cancer Rate in a Tertiary Referral Centre: An Australian Data Linkage Analysis

OBJECTIVE

High-quality gastroscopy is critical for early diagnosis of upper gastrointestinal cancers (UGCs), and assessment of missed cancers may serve as a key quality metric. Using a prospective gastroscopy database and data linkage with the Queensland Cancer Registry, we assessed the risk of developing UGC within 3 years of a cancer-negative gastroscopy at an Australian tertiary centre. Additional aims were to identify factors predictive of missed cancer, perform root cause analyses for missed cancers and assess overall survival.

DESIGN/METHOD

We identified patients who were diagnosed with UGC within 3 years of undergoing gastroscopy between 2011 and 2016. Non-mucosal cancers, cancers distal to duodenum and patients undergoing surveillance were excluded. Cases diagnosed within 6 months of gastroscopy were defined as detected cancers, while those developing within 6-36 months were defined as missed cancers. Post-endoscopy UGC rate (PEUGIC-3Y) was calculated as ratio of missed over total cancers detected. Demographic, clinical, endoscopic and histologic variables were analysed.

RESULTS

A total of 17,131 gastroscopies were performed for 10,393 patients during the study period. One hundred and twenty-six UGCs were diagnosed, including 120 detected UGCs and 6 missed UGCs. The overall PEUGIC-3Y rate was 4.8% (95% CI 2.1-10.4). The missed UGCs included 3 gastric adenocarcinomas, 2 gastro-oesophageal junction adenocarcinomas and 1 oesophageal squamous cell carcinoma. At the preceding 'cancer-negative gastroscopy', no macroscopic abnormalities were detected at the site of future UGC in 5/6 patients. A UGC developed in 2/6 patients despite an apparent adequate examination at index gastroscopy. Age, sex, indication for endoscopy and cancer location or histology were not predictive of missed cases, and survival was comparable between groups.

CONCLUSION

We demonstrate that the PEUGIC-3Y rate was 4.8% (95% CI 2.1-10.4). The majority of missed cases were adenocarcinomas of the gastro-oesophageal junction or stomach and developed in segments which were found to be normal at index gastroscopy, highlighting the challenges in detecting subtle mucosal lesions in the upper gastrointestinal tract. While overall survival between patients with detected and post-gastroscopy cancers was comparable, these ultimately represent potential missed opportunities for diagnosing an early cancer and underscore the need for quality improvement in gastroscopy.

Variation in the rate of detection of minute and small early gastric cancers at diagnostic endoscopy may reflect the performance of individual endoscopists

OBJECTIVE

The documented variation in gastric cancer (GC) detection among endoscopists has often been dismissed as a coincidental artefact of the low incidence of gastric neoplasms; it is not considered associated with differences in physicians' performance of the esophagogastroduodenoscopy procedure. This study is to confirm whether significant variations among endoscopists in early GC detection suggest the individual performance of the upper endoscopy.

DESIGN

A retrospective observational study at a single centre in Japan assessed the results of 218 early GCs detected during 25 688 routine esophagogastroduodenoscopies by 12 endoscopists. The main outcome was the rate of early GC detection for each endoscopist under the same circumstances. Other measures included the major diameters and locations of the lesions, Helicobacter pylori infection status, and baseline patient characteristics that could affect the prevalence of GC.

RESULTS

The early GC detection rates exhibited wide variation among endoscopists (0.09%-2.87%) despite performing routine esophagogastroduodenoscopies in a population with a similar background. Endoscopists were assigned to a low-detection group (n=6; detection rate: 0.47% (range: 0.09%-0.55%)) and a high-detection group (n=5; detection rate: 0.83% (range: 0.63%-1.12%)), with the single highest detector analysed separately due to his distinct detection rate (2.87%). Endoscopists in the high-detection group had better detection rates for minute (major diameter ≤5 mm) and small (major diameter 6-10 mm) GCs than the low-detection group (0.19%/0.23% vs 0.085%/0.098%). These differences were significant (p<0.01), although there were no significant differences in detection of larger tumours (major diameter ≥11 mm; 0.40% vs 0.28%; p=0.13). The tumour location and H. pylori status were similar in the low-detection group, high-detection group and for the highest detector.

CONCLUSION

Significant variation in the detection of hard-to-find, smaller GCs may reflect individual performance of the examination.

Artificial intelligence in theranostics of gastric cancer, a review

Gastric cancer (GC) is one of the commonest cancers with high morbidity and mortality in the world. How to realize precise diagnosis and therapy of GC owns great clinical requirement. In recent years, artificial intelligence (AI) has been actively explored to apply to early diagnosis and treatment and prognosis of gastric carcinoma. Herein, we review recent advance of AI in early screening, diagnosis, therapy and prognosis of stomach carcinoma. Especially AI combined with breath screening early GC system improved 97.4 % of early GC diagnosis ratio, AI model on stomach cancer diagnosis system of saliva biomarkers obtained an overall accuracy of 97.18 %, specificity of 97.44 %, and sensitivity of 96.88 %. We also discuss concept, issues, approaches and challenges of AI applied in stomach cancer. This review provides a comprehensive view and roadmap for readers working in this field, with the aim of pushing application of AI in theranostics of stomach cancer to increase the early discovery ratio and curative ratio of GC patients.

Image-enhanced endoscopy in the diagnosis of gastric premalignant conditions and gastric cancer

Diagnosis of early gastric cancer and its precancerous lesions remains a challenge for great part of western endoscopists. Changes seen in the mucosal pattern are generally subtle and hence difficult to identify. In this article, we will review the usefulness of conventional and virtual chromoendoscopy and magnification endoscopy in the recognition and classification of these lesions.

Effect of anesthesia assistance on the detection rate of precancerous lesions and early esophageal squamous cell cancer in esophagogastroduodenoscopy screening: A retrospective study based on propensity score matching

Background

Esophagogastroduodenoscopy (EGD) screening is vital for the early diagnosis of esophageal squamous cell cancer (ESCC). However, improvement in the detection rate of precancerous lesions and early ESCC with anesthesia assistance (AA) has not yet been investigated. This retrospective study aimed to evaluate the effect of AA on the detection rate of precancerous lesions and early ESCC in patients undergoing EGD screening and identify risk factors affecting the detection rate.

Methods

We reviewed patients' electronic medical records who underwent EGD screening between May 2019 and August 2020. Patients were divided into two groups based on whether they received AA: those in Group A underwent EGD screening with AA, and patients in Group O underwent EGD screening without AA. Propensity score matching (PSM) was used to account for differences in baseline characteristics. Detection rates of precancerous lesions and early ESCC were compared between the two groups following PSM. Binary logistic regression was used to identify risk factors affecting the detection rate.

Results

The final analysis included 21,835 patients (Group A = 13,319, Group O = 8,516) from 28,985 patients who underwent EGD screening during the study period. Following PSM, 6009 patients remained in each group for analysis. There was no significant difference in the detection rate of precancerous lesions and early ESCC between Groups A and O (1.1% vs. 0.8%, p > 0.05). Binary logistic regression showed that age (50-59 years, 60-69 years and 70-79 years), higher endoscopist seniority, high-definition (HD) endoscopy, narrow-band imaging (NBI), and number of endoscopic images were all independent risk factors that affected the detection rate of precancerous lesions and early ESCC.

Conclusion

There was no statistically significant difference in the detection rate of precancerous lesions and early ESCC between patients who underwent EGD screening with and without AA. All independent risk factors that affected the detection rate of precancerous lesions and early ESCC included the following: age (50-59 years, 60-69 years and 70-79 years), higher endoscopist seniority, HD endoscopy, NBI, and number of endoscopic images. Endoscopists should consider all these factors as much as possible when performing EGD screening.

A system based on deep convolutional neural network improves the detection of early gastric cancer

Background

Early gastric cancer (EGC) has a high survival rate, but it is difficult to diagnosis. Recently, artificial intelligence (AI) based on deep convolutional neural network (DCNN) has made significant progress in the field of gastroenterology. The purpose of this study was to establish a DCNN assist system to improve the detection of EGC.

Methods

3400 EGC and 8600 benign images were collected to train the DCNN to detect EGC. Subsequently, its diagnostic ability was compared to that of endoscopists using an independent internal test set (ITS, including 1289 images) and an external test set (ETS, including 542 images) come from three digestive center.

Results

The diagnostic time of DCNN and endoscopists were 0.028s, 8.05 ± 0.21s, 7.69 ± 0.25s in ITS, and 0.028s, 7.98 ± 0.19s, 7.50 ± 0.23s in ETS, respectively. In ITS, the diagnostic sensitivity and accuracy of DCNN are 88.08%(95% confidence interval,95%CI,85.24%-90.44%), 88.60% (95%CI,86.74%-90.22%), respectively. In ETS, the diagnostic sensitivity and accuracy are 92.08% (95%CI, 87.91%- 94.94%),92.07%(95%CI, 89.46%-94.08%),respectively. DCNN outperformed all endoscopists in ETS, and had a significantly higher sensitivity than the junior endoscopists(JE)(by18.54% (95%CI, 15.64%-21.84%) in ITS, also higher than JE (by21.67%,95%CI, 16.90%-27.32%) and senior endoscopists (SE) (by2.08%, 95%CI, 0.75%-4.92%)in ETS. The accuracy of DCNN model was higher (by10.47%,95%CI, 8.91%-12.27%) than that of JE in ITS, and also higher (by14.58%,95%CI, 11.84%-17.81%; by 1.94%,95%CI,1.25%-2.96%, respectively) than JE and SE in ETS.

Conclusion

The DCNN can detected more EGC images in a shorter time than the endoscopists. It will become an effective tool to assist in the detection of EGC in the near future.

Gastric Cancer Screening in Japan: A Narrative Review

Gastric cancer is the second leading cause of cancer incidence in Japan, although gastric cancer mortality has decreased over the past few decades. This decrease is attributed to a decline in the prevalence of H. pylori infection. Radiographic examination has long been performed as the only method of gastric screening with evidence of reduction in mortality in the past. The revised 2014 Japanese Guidelines for Gastric Cancer Screening approved gastric endoscopy for use in population-based screening, together with radiography. While endoscopic gastric cancer screening has begun, there are some problems associated with its implementation, including endoscopic capacity, equal access, and cost-effectiveness. As H. pylori infection and atrophic gastritis are well-known risk factors for gastric cancer, a different screening method might be considered, depending on its association with the individual's background and gastric cancer risk. In this review, we summarize the current status and problems of gastric cancer screening in Japan. We also introduce and discuss the results of gastric cancer screening using H. pylori infection status in Hoki-cho, Tottori prefecture. Further, we review risk stratification as a system for improving gastric cancer screening in the future.

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.

Emerging texture and color enhancement imaging in early gastric cancer

Screening endoscopy improves detection and prognosis of patients with gastric cancer. However, even expert endoscopists can miss early gastric cancer under standard white light imaging. Texture and color enhancement imaging (TXI) is an image-enhanced endoscopy that enhances brightness, surface irregularities such elevation or depression, and subtle color changes. A few image-oriented studies have compared the gastric color differences between neoplastic and peripheral areas under both white light imaging and TXI. The results not only suggested that the overall color differences to be more pronounced in TXI, but also that TXI mode 1 was superior to white light imaging in the visibility of early gastric cancer. Despite the promising results in these initial studies, it is unclear whether the superiority of the image-enhanced endoscopy will translate into an improvement in early gastric cancer detection in real practice. Therefore, large-scale prospective studies are necessary to investigate the efficacy of this new technology in the evaluation of patients undergoing screening endoscopy.

Development and evaluation of a double-check support system using artificial intelligence in endoscopic screening for gastric cancer

BACKGROUND

This study aimed to prevent missing gastric cancer and point out low-quality images by developing a double-check support system (DCSS) for esophagogastroduodenoscopy (EGD) still images using artificial intelligence.

METHODS

We extracted 12,977 still EGD images from 855 cases with cancer [821 with early gastric carcinoma (EGC) and 34 malignant lymphoma (ML)] and developed a lesion detection system using 10,994 images. The remaining images were used as a test dataset. Additional validation was performed using a new dataset containing 50 EGC and 1,200 non-GC images by comparing the interpretation of ten endoscopists (five trainees and five experts). Furthermore, we developed another system to detect low-quality images, which are not suitable for diagnosis, using 2198 images.

RESULTS

In the validation of 1983 images from the 124 cancer cases, the DCSS diagnosed cancer with a sensitivity of 89.2%, positive predictive value (PPV) of 93.3%, and an accuracy of 83.3%. EGC was detected in 93.2% and ML in 92.5% of cases. Comparing with the endoscopists, sensitivity was significantly higher in the DCSS, and the average diagnostic time was significantly shorter using the DCSS than that by the trainees. The sensitivity, specificity, PPV, and accuracy in detecting low-quality images were 65.8%, 93.1%, 79.6%, and 85.2% for "Blur" and 57.8%, 91.7%, 82.2%, and 78.1% for "Mucus adhesion," respectively.

CONCLUSIONS

The DCSS showed excellent capability in detecting lesions and pointing out low-quality images.

Diagnostic ability of linked color imaging in ultraslim endoscopy to identify neoplastic lesions in the upper gastrointestinal tract

Background and study aims  Linked color imaging (LCI) is a new image-enhancing technique that facilitates the differentiation of slight differences in mucosal color tone. We performed an exploratory analysis to evaluate the diagnostic capability of LCI in ultraslim endoscopy, using data from patients examined in the LCI-Further Improving Neoplasm Detection in upper gastrointestinal (LCI-FIND) trial, a large-scale, multicenter, randomized controlled trial that demonstrated the capability of LCI for detecting neoplastic lesions in the upper gastrointestinal tract. Patients and methods  Data from the LCI-FIND prospective trial were used. In the LCI-FIND trial, 1502 patients with a history of gastrointestinal cancer were randomly assigned to two groups based on examination methods: white light imaging (WLI) followed by LCI (WLI group) and LCI followed by WLI (LCI group). The present exploratory analysis investigated the outcomes of patients who underwent ultraslim and standard endoscopies. Results  Ultraslim endoscopes were used in 223 patients and standard endoscopes in 1279 patients. The primary endpoint of the LCI-FIND trial was the percentage of patients diagnosed with a neoplastic lesion using WLI or LCI. The corresponding percentage tended to be higher with LCI than with WLI among patients who underwent ultraslim endoscopy and among those who underwent standard endoscopy; the crude risk ratio was 2.21 [95 % confidence interval (CI): 1.06-4.67], and the adjusted odds ratio was 2.46 (95 % CI: 1.07-5.63). Conclusions  Our exploratory analysis of data from the LCI-FIND trial showed that LCI is useful in identifying neoplastic lesions, when used in ultraslim endoscopy.

Combination of artificial intelligence-based endoscopy and miR148a methylation for gastric indefinite dysplasia diagnosis

BACKGROUND AND AIM

Gastrointestinal endoscopy and biopsy-based pathological findings are needed to diagnose early gastric cancer. However, the information of biopsy specimen is limited because of the topical procedure; therefore, pathology doctors sometimes diagnose as gastric indefinite for dysplasia (GIN).

METHODS

We compared the accuracy of physician-performed endoscopy (trainee, n = 3; specialists, n = 3), artificial intelligence (AI)-based endoscopy, and/or molecular markers (DNA methylation: BARHL2, MINT31, TET1, miR-148a, miR-124a-3, NKX6-1; mutations: TP53; and microsatellite instability) in diagnosing GIN lesions. We enrolled 24,388 patients who underwent endoscopy, and 71 patients were diagnosed with GIN lesions. Thirty-two cases of endoscopic submucosal dissection (ESD) in 71 GIN lesions and 32 endoscopically resected tissues were assessed by endoscopists, AI, and molecular markers to identify benign or malignant lesions.

RESULTS

The board-certified endoscopic physicians group showed the highest accuracy in the receiver operative characteristic curve (area under the curve [AUC]: 0.931), followed by a combination of AI and miR148a DNA methylation (AUC: 0.825), and finally trainee endoscopists (AUC: 0.588).

CONCLUSION

AI with miR148s DNA methylation-based diagnosis is a potential modality for diagnosing GIN.

A New Dawn for the Use of Artificial Intelligence in Gastroenterology, Hepatology and Pancreatology

Artificial intelligence (AI) is rapidly becoming an essential tool in the medical field as well as in daily life. Recent developments in deep learning, a subfield of AI, have brought remarkable advances in image recognition, which facilitates improvement in the early detection of cancer by endoscopy, ultrasonography, and computed tomography. In addition, AI-assisted big data analysis represents a great step forward for precision medicine. This review provides an overview of AI technology, particularly for gastroenterology, hepatology, and pancreatology, to help clinicians utilize AI in the near future.

Prevalence and characteristics of post-gastroscopy gastric cancer: A retrospective study from an academic medical center

BACKGROUND AND STUDY AIMS

Gastric cancer is diagnosed by endoscopy but false negative rates of up to 10% in the west and 40% in Asia have been reported. In Lebanon, little is known about the rates of post-gastroscopy gastric cancer (PGGC), defined as the proportion of patients diagnosed with gastric cancer with a negative previous examination within 2 years of diagnosis. We aimed to examine the rate of PGGC and its risk factors, clinico-pathologic and endoscopic characteristics at a University medical Center.

PATIENTS AND METHODS

Retrospective analysis of patients with histologically proven gastric malignancy over the last 14 years. Patients with history of upper endoscopy preceding the index diagnostic endoscopy by 6 to 24 months were included.

RESULTS

18,976 patients underwent upper endoscopy and gastric cancer was diagnosed in 323 (1.7%). Of those, only 4 (1.2%) had a preceding endoscopy within 6 to 24 months of diagnosis: 3 adenocarcinoma and one MALT lymphoma. Upon review of the initial endoscopy, a mucosal abnormality had been noted in all 4 patients and biopsies taken in 3 were negative for cancer. The mean time to cancer diagnosis was 8 months (range 6-13 months).

CONCLUSION

A small proportion of gastric carcinomas are missed on endoscopy in this study. Patients with endoscopic evidence of mucosal abnormalities and negative biopsies should undergo repeat examination with multiple biopsies. Proper endoscopic technique, lesion recognition and adoption of performance improvement measures are important to optimize endoscopic practice.

Visibility of early gastric cancer in texture and color enhancement imaging

Objective

There are little data regarding the efficacy of texture and color enhancement imaging (TXI) for early gastric cancer (EGC) diagnosis. This study aimed to compare the color difference and visibility of EGC between white light imaging (WLI) and TXI.

Methods

This study included 20 EGCs of 18 patients undergoing endoscopic submucosal dissection. Still images of EGC in WLI, TXI mode 1 (with color enhancement), and TXI mode 2 (without color enhancement), which were consistent in distance, angle, and air insufflation, were constructed by computer simulation. The center of the lesion, eight equal peripheral points 5 mm outside the lesion, and eight inner points two‐thirds of the distance from peripheral points to the EGC lesion center were annotated. Mean color differences (ΔE) of the area between peripheral and inner points per lesion in WLI, TXI mode 1, and TXI mode 2 were analyzed. In addition, four endoscopists independently scored the visibility of EGC images of TXI mode 1 and 2 compared with WLI.

Results

Clinicopathological characteristics were as follows: 0‐IIa/0‐IIb/0‐IIc/0‐IIa+IIc = 6/1/11/2, reddish/pale = 10/10, differentiated/undifferentiated = 18/2, median tumor size = 13.5 mm. Mean ΔE ± SD = WLI/TXI mode1/TXI mode2 = 10.3 ± 4.7, 15.5 ± 7.8, and 12.7 ± 6.1, respectively. Mean ΔE was significantly higher in TXI mode 1 than in WLI. Visibility (improved/no change/decreased) was 7/13/0 and 4/16/0 in TXI mode 1 and 2, respectively. The visibility was significantly more commonly improved in the macroscopic type 0‐IIc or 0‐IIb than in 0‐IIa or IIa+IIc in TXI mode 1.

Conclusions

TXI could improve the visibility of EGC compared with WLI.

Estimated Cost-effectiveness of Endoscopic Screening for Upper Gastrointestinal Tract Cancer in High-Risk Areas in China

IMPORTANCE

Upper gastrointestinal tract cancer, including esophageal and gastric cancers, in China accounts for 50% of the global burden. Endoscopic screening may be associated with a decreased incidence of and mortality from upper gastrointestinal tract cancer.

OBJECTIVE

To evaluate the cost-effectiveness of endoscopic screening for esophageal and gastric cancers among people aged 40 to 69 years in areas of China where the risk of these cancers is high.

DESIGN, SETTING, AND PARTICIPANTS

For this economic evaluation, a Markov model was constructed for initial screening at different ages from a health care system perspective, and 5 endoscopic screening strategies with different frequencies (once per lifetime and every 10 years, 5 years, 3 years, and 2 years) were evaluated. The study was conducted between January 1, 2019, and October 31, 2020. Model parameters were estimated based on this project, government documents, and published literature. For each initial screening age (40-44, 45-49, 50-54, 55-59, 60-64, and 65-69 years), a closed cohort of 100 000 participants was assumed to enter the model and follow the alternative strategies.

MAIN OUTCOMES AND MEASURES

Cost-effectiveness was measured by calculating the incremental cost-effectiveness ratio (ICER), and the willingness-to-pay threshold was assumed to be 3 times the per capita gross domestic product in China (US $10 276). Univariate and probabilistic sensitivity analyses were conducted to assess the robustness of model findings.

RESULTS

The study included a hypothetical cohort of 100 000 individuals aged 40 to 69 years. All 5 screening strategies were associated with improved effectiveness by 1087 to 10 362 quality-adjusted life-years (QALYs) and increased costs by US $3 299 000 to $22 826 000 compared with no screening over a lifetime, leading to ICERs of US $1343 to $3035 per QALY. Screening at a higher frequency was associated with an increase in QALYs and costs; ICERs for higher frequency screening compared with the next-lower frequency screening were between US $1087 and $4511 per QALY. Screening every 2 years would be the most cost-effective strategy, with probabilities of 90% to 98% at 3 times the per capita gross domestic product of China. The model was the most sensitive to utility scores of esophageal cancer- or gastric cancer-related health states and compliance with screening.

CONCLUSIONS AND RELEVANCE

The findings suggest that combined endoscopic screening for esophageal and gastric cancers may be cost-effective in areas of China where the risk of these cancers is high; screening every 2 years would be the optimal strategy. These data may be useful for development of policies targeting the prevention and control of upper gastrointestinal tract cancer in China.

Application of artificial intelligence-driven endoscopic screening and diagnosis of gastric cancer

The landscape of gastrointestinal endoscopy continues to evolve as new technologies and techniques become available. The advent of image-enhanced and magnifying endoscopies has highlighted the step toward perfecting endoscopic screening and diagnosis of gastric lesions. Simultaneously, with the development of convolutional neural network, artificial intelligence (AI) has made unprecedented breakthroughs in medical imaging, including the ongoing trials of computer-aided detection of colorectal polyps and gastrointestinal bleeding. In the past demi-decade, applications of AI systems in gastric cancer have also emerged. With AI's efficient computational power and learning capacities, endoscopists can improve their diagnostic accuracies and avoid the missing or mischaracterization of gastric neoplastic changes. So far, several AI systems that incorporated both traditional and novel endoscopy technologies have been developed for various purposes, with most systems achieving an accuracy of more than 80%. However, their feasibility, effectiveness, and safety in clinical practice remain to be seen as there have been no clinical trials yet. Nonetheless, AI-assisted endoscopies shed light on more accurate and sensitive ways for early detection, treatment guidance and prognosis prediction of gastric lesions. This review summarizes the current status of various AI applications in gastric cancer and pinpoints directions for future research and clinical practice implementation from a clinical perspective.

Implementation effect of institutional policy of EGD observation time on neoplasm detection

BACKGROUND AND AIMS

The observation time in EGD is associated with detection rate of premalignant or neoplastic lesions in the upper GI (UGI) tract. The aim of this study was to evaluate an institutional policy of EGD observation time on the detection rate of UGI neoplasms.

METHODS

From July 2017 to March 2019, all endoscopists were requested to comply with our institutional policy of spending more than 3 minutes of observation time in every screening EGD. Observation time was defined as the time from when the endoscope reached the duodenum to when it was withdrawn. We obtained a neoplasm detection rate (NDR) during this period and compared it with that of a baseline period from 2009 to 2015.

RESULTS

During the study period, 30,506 EGDs were performed. Mean subject age was 49.9 ± 10.5 years, and 56.5% were men. All endoscopists achieved an average EGD observation time of more than 3 minutes during this period. Mean observation time was 3:35 ± 0:50, which was significantly longer than the baseline (2:38 ± 0:21, P < .001). NDR was .33%, which was higher than the baseline (.23%, P < .001). Even after adjusting for subjects' age and gender, smoking history, and endoscopists' biopsy sampling rate, prolonged EGD observation time of more than 3 minutes increased the NDR of UGI neoplasms (odds ratio, 1.51; 95% confidence interval, 1.21-1.75).

CONCLUSIONS

This study provides evidence that implementing a protocol of a prolonged observation time could increase NDR. Observation time should be an important quality indicator of the EGD examination.

Intelligent detection endoscopic assistant: An artificial intelligence-based system for monitoring blind spots during esophagogastroduodenoscopy in real-time

BACKGROUND

Observation of the entire stomach during esophagogastroduodenoscopy (EGD) is important; however, there is a lack of effective evaluation tools.

AIMS

To develop an artificial intelligence (AI)-assisted EGD system able to automatically monitor blind spots in real-time.

METHODS

An AI-based system, called the Intelligent Detection Endoscopic Assistant (IDEA), was developed using a deep convolutional neural network (DCNN) and long short-term memory (LSTM). The performance of IDEA for recognition of gastric sites in images and videos was evaluated. Primary outcomes included diagnostic accuracy, sensitivity, and specificity.

RESULTS

A total of 170,297 images and 5779 endoscopic videos were collected to develop the system. As the test group, 3100 EGD images were acquired to evaluate the performance of DCNN in recognition of gastric sites in images. The sensitivity, specificity, and accuracy of DCNN were determined as 97.18%,99.91%, and 99.83%, respectively. To assess the performance of IDEA in recognition of gastric sites in EGD videos, 129 videos were used as the test group. The sensitivity, specificity, and accuracy of IDEA were 96.29%,93.32%, and 95.30%, respectively.

CONCLUSIONS

IDEA achieved high accuracy for recognition of gastric sites in real-time. The system can be applied as a powerful assistant tool for monitoring blind spots during EGD.

Current status and future perspective of artificial intelligence applications in endoscopic diagnosis and management of gastric cancer

Image recognition using artificial intelligence (AI) has progressed significantly due to innovative technologies such as machine learning and deep learning. In the field of gastric cancer (GC) management, research on AI-based diagnosis such as anatomical classification of endoscopic images, diagnosis of Helicobacter pylori infection, and detection and qualitative diagnosis of GC is being conducted, and an accuracy equivalent to that of physicians has been reported. It is expected that AI will soon be introduced in the field of endoscopic diagnosis and management of gastric cancer as a supportive tool for physicians, thus improving the quality of medical care.

Detecting early gastric cancer: Comparison between the diagnostic ability of convolutional neural networks and endoscopists

OBJECTIVES

Detecting early gastric cancer is difficult, and it may even be overlooked by experienced endoscopists. Recently, artificial intelligence based on deep learning through convolutional neural networks (CNNs) has enabled significant advancements in the field of gastroenterology. However, it remains unclear whether a CNN can outperform endoscopists. In this study, we evaluated whether the performance of a CNN in detecting early gastric cancer is better than that of endoscopists.

METHODS

The CNN was constructed using 13,584 endoscopic images from 2639 lesions of gastric cancer. Subsequently, its diagnostic ability was compared to that of 67 endoscopists using an independent test dataset (2940 images from 140 cases).

RESULTS

The average diagnostic time for analyzing 2940 test endoscopic images by the CNN and endoscopists were 45.5 ± 1.8 s and 173.0 ± 66.0 min, respectively. The sensitivity, specificity, and positive and negative predictive values for the CNN were 58.4%, 87.3%, 26.0%, and 96.5%, respectively. These values for the 67 endoscopists were 31.9%, 97.2%, 46.2%, and 94.9%, respectively. The CNN had a significantly higher sensitivity than the endoscopists (by 26.5%; 95% confidence interval, 14.9-32.5%).

CONCLUSION

The CNN detected more early gastric cancer cases in a shorter time than the endoscopists. The CNN needs further training to achieve higher diagnostic accuracy. However, a diagnostic support tool for gastric cancer using a CNN will be realized in the near future.

Artificial intelligence for cancer detection of the upper gastrointestinal tract

In recent years, artificial intelligence (AI) has been found to be useful to physicians in the field of image recognition due to three elements: deep learning (that is, CNN, convolutional neural network), a high-performance computer, and a large amount of digitized data. In the field of gastrointestinal endoscopy, Japanese endoscopists have produced the world's first achievements of CNN-based AI system for detecting gastric and esophageal cancers. This study reviews papers on CNN-based AI for gastrointestinal cancers, and discusses the future of this technology in clinical practice. Employing AI-based endoscopes would enable early cancer detection. The better diagnostic abilities of AI technology may be beneficial in early gastrointestinal cancers in which endoscopists have variable diagnostic abilities and accuracy. AI coupled with the expertise of endoscopists would increase the accuracy of endoscopic diagnosis.

Linked Color Imaging Focused on Neoplasm Detection in the Upper Gastrointestinal Tract : A Randomized Trial

BACKGROUND

Linked color imaging (LCI) is a new image-enhanced endoscopy technique that allows users to recognize slight differences in mucosal color.

OBJECTIVE

To compare the performance of LCI with white light imaging (WLI) in detecting neoplastic lesions in the upper gastrointestinal tract.

DESIGN

A controlled, multicenter trial with randomization using minimization. (University Hospital Medical Information Network Clinical Trials Registry: UMIN000023863).

SETTING

16 university hospitals and 3 tertiary care hospitals in Japan.

PATIENTS

1502 patients with known previous or current cancer of the gastrointestinal tract and undergoing surveillance for gastrointestinal cancer.

INTERVENTION

WLI followed by LCI examination (WLI group) or LCI followed by WLI examination (LCI group).

MEASUREMENTS

Diagnosis of 1 or more neoplastic lesions in the pharynx, esophagus, or stomach in the first examination (primary outcome) and 1 or more neoplastic lesions overlooked in the first examination (secondary outcome).

RESULTS

752 patients were assigned to the WLI group and 750 to the LCI group. The percentage of patients with 1 or more neoplastic lesions diagnosed in the first examination was higher with LCI than with WLI (60 of 750 patients or 8.0% [95% CI, 6.2% to 10.2%] vs. 36 of 752 patients or 4.8% [CI, 3.4% to 6.6%]; risk ratio, 1.67 [CI, 1.12 to 2.50; P = 0.011]). The proportion with overlooked neoplasms was lower in the LCI group than in the WLI group (5 of 750 patients or 0.67% [CI, 0.2% to 1.6%] vs. 26 of 752 patients or 3.5% [CI, 2.3% to 5.0%]; risk ratio, 0.19 [CI, 0.07 to 0.50]).

LIMITATION

Endoscopists were not blinded.

CONCLUSION

LCI is more effective than WLI for detecting neoplastic lesions in the pharynx, esophagus, and stomach.

PRIMARY FUNDING SOURCE

Fujifilm Corporation.

The endoluminal pressures during flexible gastrointestinal endoscopy

In flexible gastrointestinal (GI) endoscopy, endoscopic insufflation is crucial and directly affects visualization. Optimal visualization enables endoscopists to conduct better examinations and administer optimal treatments. However, endoscopic insufflation is typically performed manually and is subjective. We aimed to measure the GI endoluminal pressure during flexible GI endoscopy. Participants underwent esophagogastroduodenoscopy (EGD) at our endoscopy center. Pressure measurement was conducted after completing diagnostic or follow-up EGD. The endoluminal pressure in the esophagus and stomach was measured at 1-s intervals for 1 min while performing EGD for observational and diagnostic purposes. During the measurements, the endoscopists maintained what they subjectively considered to be adequate exposure for screening for lesions by dilating the lumen. Eighty patients were enrolled in this study. The upper GI endoluminal pressure was assessed during EGD without adverse events. The esophageal endoluminal pressure averaged 8.9 (- 3.0 to 20.7) mmHg, and the gastric endoluminal pressure averaged 10.0 (3.0-17.9) mmHg; the upper GI endoluminal pressures were not affected by patient-related factors or the number of endoscopists' postgraduate years. We have successfully obtained the GI endoluminal pressures during EGD. Further accumulation of these data may lead to more stable and reproducible flexible endoscopic diagnosis and intervention.

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.

Recent advances in diagnostic upper endoscopy

BACKGROUND

Esophageo-gastro-duodenoscopy (EGD) is an important procedure used for detection and diagnosis of esophago-gastric lesions. There exists no consensus on the technique of examination.

AIM

To identify recent advances in diagnostic EGDs to improve diagnostic yield.

METHODS

We queried the PubMed database for relevant articles published between January 2001 and August 2019 as well as hand searched references from recently published endoscopy guidelines. Keywords used included free text and MeSH terms addressing quality indicators and technological innovations in EGDs. Factors affecting diagnostic yield and EGD quality were identified and divided into the follow segments: Pre endoscopy preparation, sedation, examination schema, examination time, routine biopsy, image enhanced endoscopy and future developments.

RESULTS

We identified 120 relevant abstracts of which we utilized 67 of these studies in our review. Adequate pre-endoscopy preparation with simethicone and pronase increases gastric visibility. Proper sedation, especially with propofol, increases patient satisfaction after procedure and may improve detection of superficial gastrointestinal lesions. There is a movement towards mandatory picture documentation during EGD as well as dedicating sufficient time for examination improves diagnostic yield. The use of image enhanced endoscopy and magnifying endoscopy improves detection of squamous cell carcinoma and gastric neoplasm. The magnifying endoscopy simple diagnostic algorithm is useful for diagnosis of early gastric cancer.

CONCLUSION

There is a steady momentum in the past decade towards improving diagnostic yield, quality and reporting in EGDs. Other interesting innovations, such as Raman spectroscopy, endocytoscopy and artificial intelligence may have widespread endoscopic applications in the near future.

Gastric cancer missed at esophagogastroduodenoscopy in a well-defined Spanish population

BACKGROUND

Although esophagogastroduodenoscopy (EGD) is the standard procedure for the diagnosis of gastric cancer (GC), some GCs are missed. There are no published data on the missed rate of GC in Spain.

AIMS

To determine the frequency and characteristics of missed GCs and assess the quality of the EGD in a specific population with GC.

METHODS

Records of all patients diagnosed with gastric adenocarcinoma between 2012 and 2016 in a defined geographic area were reviewed. Missed GC was defined as a case with a prior negative EGD for cancer. Quality indicators from the prior EGDs were measured.

RESULTS

From 212 cases of GC, 25 cases were excluded. Seventeen out of 187 patients had a prior EGD (9.1%). Twelve of those 17 missed GC had a prior EGD with some abnormal findings. In 6 of them, biopsies were taken. Survival was no different between patients with missed and non-missed GC. Quality indicators that failed to meet standards were recording time, image documentation, and a protocol of biopsies.

CONCLUSIONS

Missed GC in an EGD in a defined population in Spain is not uncommon (9.1%). The endoscopist is an important factor in missed GC due to lack of adequate detection and sampling error. Compliance with performance of quality indicators could reduce missed GC.

Factors associated with oesophagogastric cancers missed by gastroscopy: a case-control study

INTRODUCTION

There is increasing demand for gastroscopy in the United Kingdom. In around 10% of patients, gastroscopy is presumed to have missed oesophagogastric (OG) cancer prior to diagnosis. We examine patient, endoscopist and service level factors that may affect rates of missed OG cancers.

METHODS

Gastroscopies presumed to have missed OG cancers performed up to 3 years prior to diagnosis were identified over 6 years in Sheffield, UK. Factors related to the patient, endoscopist and endoscopy lists were examined in a case-control study. Procedures which missed cancer were compared with two procedure controls: the procedures which subsequently diagnosed cancer in the same patient, and second, endoscopist matched procedures diagnostic of small benign focal lesions.

RESULTS

We identified 48 (7.7%) cases of missed OG cancer. Endoscopy lists on which OG cancer diagnoses were missed contained a greater number of total procedures compared with lists on which diagnoses were subsequently made (OR 1.42 95% CI 1.13 to 1.78) and when compared with lists during which matched endoscopists diagnosed benign small focal lesions (OR 1.25, 95% CI 1.02 to 1.52). The use of sedation, endoscopist profession and experience, or time of procedure were not associated with a missed cancer.

CONCLUSION

7.7% of patients diagnosed with OG cancer could have been diagnosed and treated earlier. Our study suggests that endoscopy lists with greater numbers of procedures may be associated with missed OG cancers. The use of sedation, endoscopist background or time of procedure did not increase the risk of missed cancer procedures.

Endoscopist biopsy rate as a quality indicator for outpatient gastroscopy: a multicenter cohort study with validation

BACKGROUND AND AIMS

The diagnosis of gastric premalignant conditions (GPCs) relies on endoscopy with mucosal sampling. We hypothesized that the endoscopist biopsy rate (EBR) might constitute a quality indicator for EGD, and we have analyzed its association with GPC detection and the rate of missed gastric cancers (GCs).

METHODS

We analyzed EGD databases from 2 high-volume outpatient units. EBR values, defined as the proportion of EGDs with ≥1 biopsy to all examinations were calculated for each endoscopist in Unit A (derivation cohort) and divided by the quartile values into 4 groups. Detection of GPC was calculated for each group and compared using multivariate clustered logistic regression models. Unit B database was used for validation. All patients were followed in the Cancer Registry for missed GCs diagnosed between 1 month and 3 years after EGDs with negative results.

RESULTS

Sixteen endoscopists in Unit A performed 17,490 EGDs of which 15,340 (87.7%) were analyzed. EBR quartile values were 22.4% to 36.7% (low EBR), 36.8% to 43.7% (moderate), 43.8% to 51.6% (high), and 51.7% and 65.8% (very-high); median value 43.8%. The odds ratios for the moderate, high, and very-high EBR groups of detecting GPC were 1.6 (95% confidence interval [CI], 1.3-1.9), 2.0 (95% CI, 1.7-2.4), and 2.5 (95% CI, 2.1-2.9), respectively, compared with the low EBR group (P < .001). This association was confirmed with the same thresholds in the validation cohort. Endoscopists with higher EBR (≥43.8%) had a lower risk of missed cancer compared with those in the lower EBR group (odds ratio, 0.44; 95% CI, 0.20-1.00; P = .049).

CONCLUSIONS

The EBR parameter is highly variable among endoscopists and is associated with efficacy in GPC detection and the rate of missed GCs.

A robust qualitative transcriptional signature for the correct pathological diagnosis of gastric cancer

BACKGROUND

Currently, pathological examination of gastroscopy biopsy specimens is the gold standard for gastric cancer (GC) diagnosis. However, it has a false-negative rate of 10-20% due to inaccurate sampling locations and/or insufficient sampling amount. A signature should be developed to aid the early diagnosis of GC using biopsy specimens even when they are sampled from inaccurate locations.

METHODS

We extracted a robust qualitative transcriptional signature, based on the within-sample relative expression orderings (REOs) of gene pairs, to discriminate both GC tissues and adjacent-normal tissues from non-GC gastritis, intestinal metaplasia and normal gastric tissues.

RESULTS

A signature consisting of two gene pairs for GC diagnosis was identified and validated in data of both biopsy specimens and surgical resection specimens pooled from publicly available datasets measured by different laboratories with different platforms. For gastroscopy biopsy specimens, 96.20% of 79 non-GC tissues were correctly identified as non-GC, and 96.84% of 158 GC tissues and six of seven adjacent-normal tissues were correctly identified as GC. For surgical resection specimens, 98.37% of 2560 GC tissues and 97.28% of 221 adjacent-normal tissues were correctly identified as GC. Especially, 97.67% of the 257 GC patients at stage I were exactly diagnosed as GC. We additionally measured 21 GC tissues from seven different GC patients, each with three specimens sampled from three tumor locations with different proportions of the tumor epithelial cell. All these GC tissues were correctly identified as GC, even when the proportion of the tumor epithelial cell was as low as 14%.

CONCLUSIONS

The qualitative transcriptional signature can distinguish both GC and adjacent-normal tissues from normal, gastritis and intestinal metaplasia tissues of non-GC patients even using inaccurately sampled biopsy specimens, which can be applied robustly at the individual level to aid the early GC diagnosis.

Blue laser imaging-bright improves the real-time detection rate of early gastric cancer: a randomized controlled study

BACKGROUND AND AIMS

Blue laser imaging-bright (BLI-bright) has shown promise as a more useful tool for detection of early gastric cancer (EGC) than white-light imaging (WLI). However, the diagnostic performance of BLI-bright in the detection of EGC has not been investigated. We aimed to compare real-time detection rates of WLI with that of BLI-bright for EGC.

METHODS

This was a prospective, randomized, controlled study in 2 Japanese academic centers. We investigated 629 patients undergoing follow-up endoscopy for atrophic gastritis with intestinal metaplasia or surveillance after endoscopic resection of EGC. Patients were randomly assigned to receive primary WLI followed by BLI-bright or primary BLI-bright followed by WLI. The real-time detection rates of EGC were compared between primary WLI and primary BLI-bright.

RESULTS

There were 298 patients in each group. The real-time detection rate of EGC with primary BLI-bright was significantly greater than that with primary WLI (93.1% vs 50.0%; P = .001). Primary BLI-bright had a significantly greater ability to detect EGCs in patients with a history of endoscopic resection for EGC, no Helicobacter pylori infection in the stomach after eradication therapy, lesions with an open-type atrophic border, lesions in the lower third of the stomach, depressed-type lesions, small lesions measuring <10 mm and 10 to 20 mm in diameter, reddish lesions, well-differentiated adenocarcinomas, and lesions with a depth of invasion of T1a.

CONCLUSIONS

BLI-bright has a higher real-time detection rate for EGC than WLI. BLI-bright should be performed during surveillance endoscopy in patients at high risk for EGC. (Clinical trial registration number: UMIN000011324.).

Three-year interval for endoscopic screening may reduce the mortality in patients with gastric cancer

BACKGROUND

Endoscopic screening has been adopted in South Korea for the national screening of gastric cancer (GC). This study aimed to assess the effect on overall survival of GC patients and determine the optimal endoscopic screening interval.

METHODS

The baseline characteristics and overall survival of GC patients treated at the National Cancer Center, Korea, between 2010 and 2016 were compared between those without a history of endoscopic evaluation (group N) and those in whom the interval between the last endoscopic evaluations and diagnosis of GC was ≤ 1, 1-2, 2-3, 3-4, or > 4 years (groups 1-5, respectively).

RESULTS

A total of 2362 patients met the criteria for the study (1060 in group N and 1302 in groups 1-5). More patients in groups 1-5 were diagnosed with stage I GC (83.7, 83.7, 71.8, 78.2, and 71.6%, respectively) than in group N (62.4%, P < 0.001) and were treated endoscopically (38.8, 33.8, 24.7, 21.8, and 15.5%, respectively, vs. 13.5%; P < 0.001). Group 2 had less-advanced tumor stages (P = 0.001) and was more likely to have received endoscopic treatments (P = 0.026) than group 3. Hazard ratios for death were significantly lower in groups 2 (0.45; 95% confidence interval [CI], 0.32-0.64) and 3 (0.57; 95% CI, 0.33-0.98) than in group N; the decrease was not significant in group 4 (0.49, 95% CI, 0.20-1.20).

CONCLUSIONS

Endoscopic screening every 3 years may reduce the mortality of GC patients, though screenings at least every 2 years may benefit patients with less-advanced stages.

Application of artificial intelligence using a convolutional neural network for detecting gastric cancer in endoscopic images

BACKGROUND

Image recognition using artificial intelligence with deep learning through convolutional neural networks (CNNs) has dramatically improved and been increasingly applied to medical fields for diagnostic imaging. We developed a CNN that can automatically detect gastric cancer in endoscopic images.

METHODS

A CNN-based diagnostic system was constructed based on Single Shot MultiBox Detector architecture and trained using 13,584 endoscopic images of gastric cancer. To evaluate the diagnostic accuracy, an independent test set of 2296 stomach images collected from 69 consecutive patients with 77 gastric cancer lesions was applied to the constructed CNN.

RESULTS

The CNN required 47 s to analyze 2296 test images. The CNN correctly diagnosed 71 of 77 gastric cancer lesions with an overall sensitivity of 92.2%, and 161 non-cancerous lesions were detected as gastric cancer, resulting in a positive predictive value of 30.6%. Seventy of the 71 lesions (98.6%) with a diameter of 6 mm or more as well as all invasive cancers were correctly detected. All missed lesions were superficially depressed and differentiated-type intramucosal cancers that were difficult to distinguish from gastritis even for experienced endoscopists. Nearly half of the false-positive lesions were gastritis with changes in color tone or an irregular mucosal surface.

CONCLUSION

The constructed CNN system for detecting gastric cancer could process numerous stored endoscopic images in a very short time with a clinically relevant diagnostic ability. It may be well applicable to daily clinical practice to reduce the burden of endoscopists.

Natural history of gastric cancer from a retrospective review of endoscopic images of older patients with interval gastric cancer

AIM

Interval gastric cancers (IGC) are defined as those diagnosed after negative results of endoscopy carried out within the past 10 years. We aimed to investigate the characteristics of IGC and the natural history of gastric cancer (GC) from a retrospective view of endoscopic images of older patients with IGC.

METHODS

We retrospectively reviewed endoscopic images of 240 patients with GC who were aged >60 years. We compared past endoscopic images with newer ones, in which GC was diagnosed. IGC were classified into two categories: missed cancers and new cancers.

RESULTS

Of the 240 patients with GC, 32 had past endoscopic images that qualified for a precise review. A total of 14 cases involved new cancers, whereas 18 involved missed cancers. Most of the IGC were stage I for at least 2 years; however, a small subset was unresectable at >2 years after a negative endoscopy. Furthermore, the rate of endoscopic treatment was significantly higher for IGC compared with that for non-IGC.

CONCLUSIONS

In people aged >60 years, most IGC remain in an early stage for at least 2 years; however, at >2 years after a negative endoscopy, some are unresectable. These results suggest that most early-stage GC will not develop into advanced cancers within 2 years; thus, a 2-year interval might be within the permissible range for patients with negative endoscopy results for any lesions. Geriatr Gerontol Int 2018; 18: 997-1002.

Longer Observation Time Increases Proportion of Neoplasms Detected by Esophagogastroduodenoscopy

BACKGROUND & AIMS

Esophagogastroduodenoscopy (EGD) is commonly used to detect upper gastrointestinal (GI) neoplasms. However, there is little evidence that longer examination time increases rate of detection of upper GI neoplasia. We investigated the association between length of time spent performing a normal screening EGD and rate of neoplasm detection.

METHODS

We performed a retrospective analysis of data from 111,962 subjects who underwent EGD as part of a comprehensive health-screening program from January 2009 to December 2015 in Korea. Endoscopy findings were extracted from reports prepared by 14 board-certified endoscopists. Endoscopists were classified as fast or slow based on their mean examination time for a normal EGD without biopsy during their first year of the study. All endoscopists used the same endoscopy unit. We obtained findings from histologic analyses of GI biopsies from patient records; positive findings were defined as the detection of neoplasms (esophageal, gastric, or duodenal lesions). We examined the association between examination time and proportions of neoplasms detected. The primary outcome measure was the rate of neoplasm detection for each endoscopist (total number of neoplastic lesions detected divided by the number of subjects screened) and as the proportion of subjects with at least 1 neoplastic lesion.

RESULTS

The mean examination time was 2 minutes 53 seconds. Using 3 minutes as a cutoff, we classified 8 endoscopists as fast (mean duration, 2:38 ± 0:21 minutes) and 6 endoscopists as slow (mean duration, 3:25 ± 0:19 minutes). Each endoscopist's mean examination time correlated with their rate of neoplasm detection (R² = 0.54; P = .046). Fast endoscopists identified neoplasms in the upper GI tract in 0.20% of patients, whereas slow endoscopists identified these in 0.28% of patients (P = .0054). The frequency of endoscopic biopsy varied among endoscopists (range, 6.9%-27.8%) and correlated with rate of neoplasm detection (R² = 0.76; P = .0015). On multivariable analysis, slow endoscopists were more likely to detect gastric adenomas or carcinomas than fast endoscopists (odds ratio, 1.52; 95% CI, 1.17-1.97).

CONCLUSIONS

In a retrospective analysis of data from more than 100,000 subjects who underwent EGD in a screening program, we found slow endoscopists detected a higher proportion of neoplasms than fast endoscopists. Examination time is therefore a useful indicator of quality for EGD.