Comparison of clinical utility of deep learning-based systems for small-bowel capsule endoscopy reading.

BACKGROUND AND AIM: Convolutional neural network (CNN) systems that automatically detect abnormalities from small-bowel capsule endoscopy (SBCE) images are still experimental, and no studies have directly compared the clinical usefulness of different systems. We compared endoscopist readings using an existing and a novel CNN system in a real-world SBCE setting. METHODS: Thirty-six complete SBCE videos, including 43 abnormal lesions (18 mucosal breaks, 8 angioectasia, and 17 protruding lesions), were retrospectively prepared. Three reading processes were compared: (A) endoscopist readings without CNN screening, (B) endoscopist readings after an existing CNN screening, and (C) endoscopist readings after a novel CNN screening. RESULTS: The mean number of small-bowel images was 14 747 per patient. Among these images, existing and novel CNN systems automatically captured 24.3% and 9.4% of the images, respectively. In this process, both systems extracted all 43 abnormal lesions. Next, we focused on the clinical usefulness. The detection rates of abnormalities by trainee endoscopists were not significantly different across the three processes: A, 77%; B, 67%; and C, 79%. The mean reading time of the trainees was the shortest during process C (10.1 min per patient), followed by processes B (23.1 min per patient) and A (33.6 min per patient). The mean psychological stress score while reading videos (scale, 1-5) was the lowest in process C (1.8) but was not significantly different between processes B (2.8) and A (3.2). CONCLUSIONS: Our novel CNN system significantly reduced endoscopist reading time and psychological stress while maintaining the detectability of abnormalities. CNN performance directly affects clinical utility and should be carefully assessed.

Muscular metastasis of superficial esophageal squamous cell carcinoma: a rare case of recurrence after endoscopic submucosal dissection with additional chemoradiotherapy.

Esophageal cancer after endoscopic treatment may recur depending on the risk. We present a case of a rare T1b esophageal cancer after endoscopic treatment plus chemoradiotherapy (CRT) that recurred with metastasis of the dorsal muscles. A 70-year-old man was referred for treatment of early-stage esophageal carcinoma. Endoscopic submucosal dissection (ESD) was performed and histopathology showed a poorly differentiated squamous cell carcinoma with invasion to the submucosal layer (sm2) with INFc-type invasion and positive venous invasion. After subsequent CRT, the patient was monitored every 6 months, using computed tomography (CT) and endoscopy. Fifteen months after the treatment, contrast CT revealed a spherical mass with 9 cm ring enhancement within the right erector spinae, that had squamous cell carcinoma confirmed by CT-guided biopsy. Radiation and systemic chemotherapy were initiated for the metastasis of the esophageal carcinoma. However, he died of respiratory failure due to rapid pleural effusion 26 months after ESD. Pathological autopsy showed diffuse squamous cell carcinoma invasion of the cystic wall, forming a lumbar mass, and absence of cancer cell remnants or recurrences in the esophagus. This case report emphasizes the need for systemic observation of superficial esophageal cancer after treatment with a high risk of recurrence.

Impact of COVID-19 on the endoscopy department since the early phase of the pandemic in 2020: A questionnaire study among patients with canceled examinations at a single Japanese institution.

INTRODUCTION: In early 2020, the Japanese government declared a nationwide state of emergency for the COVID-19 pandemic. We investigated the impact of the emergency declaration on endoscopy adherence and conducted a follow-up study of patients with canceled examinations at a tertiary endoscopy facility in Japan in 2020. METHODS: We compared the number of endoscopies performed, and cancelations at the endoscopy unit between 2019 and 2020 and used the Bayesian structural time series (BSTS) model to estimate the decrease in the number of endoscopies in 2020. We administered a questionnaire to those who had not undergone a scheduled endoscopy. RESULTS: Of 14 146 and 13 338 scheduled examinations, 1233 (8.7%) and 1403 (10.5%) were canceled in 2019 and 2020, respectively. During both years, age < 50 years, age > 80 years, upper endoscopy, and experience of endoscopy in the past 5 years were significantly associated with cancelations. In 2020, cancelations in the 14th-26th week of the year, including the period of state of emergency, increased significantly, and more women canceled. Of the 409 questionnaire-respondents, 174 (42.5%) indicated that COVID-19 had influenced their cancelation, and 315 (77.0%) had not undergone similar endoscopic examinations since then. The BSTS model predicted a decrease of 957 (95% CI -1213 to -708, P = .003) examinations. CONCLUSION: In 2020, despite low numbers of COVID-19 cases in the study site, the number of endoscopies decreased, and cancelation increased. Further research is needed on the future impact of a decrease in the number of endoscopies during a COVID-19 pandemic.

Assigning a different endoscopist for each annual follow-up may contribute to improved gastric cancer detection rates.

Background and study aims Esophagogastroduodenoscopy (EGD) is an effective and important diagnostic tool to detect gastric cancer (GC). Although previous studies show that examiner, patient, and instrumental factors influence the detection of GC, we analyzed whether assigning a different examiner to surveillance EGD would improve the detection of GC compared to assigning the same examiner as in the previous endoscopy. Patients and methods We retrospectively reviewed patients who underwent two or more consecutive surveillance EGDs at a single center between 2017 and 2019. We identified factors associated with GC detection using multivariable regression analysis and propensity-score matching. Results Among 7794 patients, 99 GC lesions in 93 patients were detected by surveillance EGD (detection rate; 1.2 %), with a mean surveillance interval of 11.2 months. Among the detected 99 lesions, 87 (87.9 %) were curatively treated with endoscopy. There were no differences in the clinicopathologic characteristics of GC detected by the same or different endoscopists. GC detection in the group examined by different endoscopists was more statistically significant than in the group examined by the same endoscopist, even after propensity-score matching (1.6 % and 0.7 %; P  < 0.05). Endoscopic experience and other factors were not statistically significant between the two groups. Conclusions In surveillance EGD, having a different endoscopist for each exam may improve GC detection rates, regardless of the endoscopist's experience.

A case of refractory esophageal stricture due to occult gastrinoma of the duodenum.

Gastrinoma may cause refractory esophageal stricture due to gastro-esophageal reflux disease (GERD), but imaging technologies have limited power in its diagnosis. A 74-year-old female with a history of peptic ulcers suffered from repeated epigastralgia, and she visited a local hospital. An esophago-gastro-duodenoscopy (EGD) demonstrated severe reflux esophagitis and multiple peptic ulcers. Blood examination revealed a high value of fasting serum gastrin. Multi-detector computed tomography showed a hypervascular and tiny nodule in duodenal bulb, although other imaging technologies did not. Short-term medication with a proton pump inhibitor or potassium-competitive acid blocker was intermittently provided, but dysphagia was repeatedly worsened, and she was referred to our division. Serum hypergastrinemia was retained, and EGD reexamination depicted esophageal stricture, treated by multiple sessions of endoscopic balloon dilatation. Primary tumor was not identified by the morphological imaging technologies, but a selective arterial secretagogue injection test suggested its existence in the duodenum or pancreatic head. Pancreaticoduodenectomy was performed, and histological study identified 2 mm-sized microgastrinoma buried in Brunner`s glands on the posterior wall of the duodenum bulb. We reported a case with difficulty in diagnosis of the smallest sporadic gastrinoma of the duodenum, which might cause refractory GERD-associated stricture.

Automatic detection of various abnormalities in capsule endoscopy videos by a deep learning-based system: a multicenter study.

BACKGROUND AND AIMS: A deep convolutional neural network (CNN) system could be a high-level screening tool for capsule endoscopy (CE) reading but has not been established for targeting various abnormalities. We aimed to develop a CNN-based system and compare it with the existing QuickView mode in terms of their ability to detect various abnormalities. METHODS: We trained a CNN system using 66,028 CE images (44,684 images of abnormalities and 21,344 normal images). The detection rate of the CNN for various abnormalities was assessed per patient, using an independent test set of 379 consecutive small-bowel CE videos from 3 institutions. Mucosal breaks, angioectasia, protruding lesions, and blood content were present in 94, 29, 81, and 23 patients, respectively. The detection capability of the CNN was compared with that of QuickView mode. RESULTS: The CNN picked up 1,135,104 images (22.5%) from the 5,050,226 test images, and thus, the sampling rate of QuickView mode was set to 23% in this study. In total, the detection rate of the CNN for abnormalities per patient was significantly higher than that of QuickView mode (99% vs 89%, P < .001). The detection rates of the CNN for mucosal breaks, angioectasia, protruding lesions, and blood content were 100% (94 of 94), 97% (28 of 29), 99% (80 of 81), and 100% (23 of 23), respectively, and those of QuickView mode were 91%, 97%, 80%, and 96%, respectively. CONCLUSIONS: We developed and tested a CNN-based detection system for various abnormalities using multicenter CE videos. This system could serve as an alternative high-level screening tool to QuickView mode.

The risks and characteristics of the delayed bleeding after endoscopic submucosal dissection for early gastric carcinoma in cases with anticoagulants.

BACKGROUND: Endoscopic submucosal dissection (ESD) is a minimally invasive treatment for early gastric carcinoma. Vitamin K antagonists and direct oral anticoagulants (DOAC) were reported to increase the risk of delayed bleeding after ESD. However, the evaluation of ESD cases taking anticoagulants is scarce. We analyzed the risk and characteristics of delayed bleeding after gastric ESD in patients on anticoagulants. METHODS: We performed a retrospective observational study at a single center. Consecutive patients who underwent ESD for early gastric carcinoma and took anticoagulants, including warfarin, rivaroxaban, dabigatran, apixaban, and edoxaban, between January 2012 and December 2018, were analyzed. We also calculated delayed bleeding rates for those without anticoagulants. RESULTS: Of 1855 eligible patients who underwent gastric ESDs, 143 took anticoagulants. Delayed bleeding occurred in 30 (21.0%) cases taking anticoagulants, with 15 (19.5%) cases in the DOAC group [rivaroxaban, seven cases (21.2%); dabigatran, four cases (20.0%); apixaban, four cases (23.5%); and edoxaban, zero cases (0%)] and 15 cases (22.7%) in the warfarin group. Furthermore, 43/344 (12.5%) patients taking antiplatelets and 76/1368 (5.6%) patients without antithrombic drugs experienced delayed bleeding. Multivariable logistic analysis revealed post-heart valve replacement (OR, 6.56; 95% CI, 1.75-24.7; p < .05) as a risk for delayed bleeding in warfarin-taking patients, while no statistically significant factor was found in DOAC-taking patients. CONCLUSIONS: Anticoagulants were associated with a high incidence of severe delayed bleeding. Careful attention should be paid to patients on anticoagulants after gastric ESD, especially those on warfarin after heart valve replacement.

Automatic detection and classification of protruding lesions in wireless capsule endoscopy images based on a deep convolutional neural network.

BACKGROUND AND AIMS: Protruding lesions of the small bowel vary in wireless capsule endoscopy (WCE) images, and their automatic detection may be difficult. We aimed to develop and test a deep learning-based system to automatically detect protruding lesions of various types in WCE images. METHODS: We trained a deep convolutional neural network (CNN), using 30,584 WCE images of protruding lesions from 292 patients. We evaluated CNN performance by calculating the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity, using an independent set of 17,507 test images from 93 patients, including 7507 images of protruding lesions from 73 patients. RESULTS: The developed CNN analyzed 17,507 images in 530.462 seconds. The AUC for detection of protruding lesions was 0.911 (95% confidence interval [Cl], 0.9069-0.9155). The sensitivity and specificity of the CNN were 90.7% (95% CI, 90.0%-91.4%) and 79.8% (95% CI, 79.0%-80.6%), respectively, at the optimal cut-off value of 0.317 for probability score. In a subgroup analysis of the category of protruding lesions, the sensitivities were 86.5%, 92.0%, 95.8%, 77.0%, and 94.4% for the detection of polyps, nodules, epithelial tumors, submucosal tumors, and venous structures, respectively. In individual patient analyses (n = 73), the detection rate of protruding lesions was 98.6%. CONCLUSION: We developed and tested a new computer-aided system based on a CNN to automatically detect various protruding lesions in WCE images. Patient-level analyses with larger cohorts and efforts to achieve better diagnostic performance are necessary in further studies.

Clinical usefulness of a deep learning-based system as the first screening on small-bowel capsule endoscopy reading.

BACKGROUND AND AIM: To examine whether our convolutional neural network (CNN) system based on deep learning can reduce the reading time of endoscopists without oversight of abnormalities in the capsule-endoscopy reading process. METHODS: Twenty videos of the entire small-bowel capsule endoscopy procedure were prepared, each of which included 0-5 lesions of small-bowel mucosal breaks (erosions or ulcerations). At another institute, two reading processes were compared: (A) endoscopist-alone readings and (B) endoscopist readings after the first screening by the proposed CNN. In process B, endoscopists read only images detected by CNN. Two experts and four trainees independently read 20 videos each (10 for process A and 10 for process B). Outcomes were reading time and detection rate of mucosal breaks by endoscopists. Gold standard was findings at the original institute by two experts. RESULTS: Mean reading time of small-bowel sections by endoscopists was significantly shorter during process B (expert, 3.1 min; trainee, 5.2 min) compared to process A (expert, 12.2 min; trainee, 20.7 min) (P < 0.001). For 37 mucosal breaks, detection rate by endoscopists did not significantly decrease in process B (expert, 87%; trainee, 55%) compared to process A (expert, 84%; trainee, 47%). Experts detected all eight large lesions (>5 mm), but trainees could not, even when supported by the CNN. CONCLUSIONS: Our CNN-based system for capsule endoscopy videos reduced the reading time of endoscopists without decreasing the detection rate of mucosal breaks. However, the reading level of endoscopists should be considered when using the system.

The sub-classification of type B2 vessels according to the magnifying endoscopic classification of the Japan Esophageal Society.

OBJECTIVES: Guidelines for magnified endoscopic diagnosis of esophageal squamous cell carcinoma (SCC) have been proposed by the Japan Esophageal Society. Type B1, B2, and B3 reflect increasing tumor invasion depths (within mucosal epithelium or into lamina propria mucosa [T1a-EP/LPM], into muscularis mucosa or superficial invasion into submucosa [T1a-MM/T1b-SM1], and into submucosa [T1b-SM2], respectively). The diagnostic accuracy of type B1 and B3 is high, but accuracy of type B2 is low. We aimed to improve the diagnostic accuracy of type B2. METHODS: We retrospectively reviewed 248 SCC lesions treated with endoscopic submucosal dissection between January 2012 and July 2018 and identified the B2 lesions. The maximum diameter of the area presenting B2 was measured and evaluated in relation to tumor invasion, for which receiver-operating characteristic (ROC) curves were generated. The optimal area size for distinguishing T1a-EP/LPM from T1a-MM or deeper invasion was determined. RESULTS: There were 78 lesions with B2, of which 26 (33%) were T1a-MM or T1b-SM1 SCCs. ROC curve analysis indicated that the optimal cut-off for the target area showing B2 was 4 mm. The invasion depth (EP/LPM: MM/SM1: SM2) of B2 observed in an area with a diameter <4 mm (B2-Narrow) and those with diameter ≥4 mm (B2-Broad) was 46:11:1 and 1:15:4, respectively. To predict T1a-MM or deeper invasion, B2-Broad had a sensitivity, specificity, positive predictive value, and negative predictive value of 61%, 98%, 95%, and 79%, respectively. CONCLUSION: The diagnostic accuracy of type B2 was improved by evaluating the area of type B2.

Artificial intelligence using a convolutional neural network for automatic detection of small-bowel angioectasia in capsule endoscopy images.

BACKGROUND AND AIM: Although small-bowel angioectasia is reported as the most common cause of bleeding in patients and frequently diagnosed by capsule endoscopy (CE) in patients with obscure gastrointestinal bleeding, a computer-aided detection method has not been established. We developed an artificial intelligence system with deep learning that can automatically detect small-bowel angioectasia in CE images. METHODS: We trained a deep convolutional neural network (CNN) system based on Single Shot Multibox Detector using 2237 CE images of angioectasia. We assessed its diagnostic accuracy by calculating the area under the receiver operating characteristic curve (ROC-AUC), sensitivity, specificity, positive predictive value, and negative predictive value using an independent test set of 10 488 small-bowel images, including 488 images of small-bowel angioectasia. RESULTS: The AUC to detect angioectasia was 0.998. Sensitivity, specificity, positive predictive value, and negative predictive value of CNN were 98.8%, 98.4%, 75.4%, and 99.9%, respectively, at a cut-off value of 0.36 for the probability score. CONCLUSIONS: We developed and validated a new system based on CNN to automatically detect angioectasia in CE images. This may be well applicable to daily clinical practice to reduce the burden of physicians as well as to reduce oversight.

Automatic detection of blood content in capsule endoscopy images based on a deep convolutional neural network.

BACKGROUND AND AIM: Detecting blood content in the gastrointestinal tract is one of the crucial applications of capsule endoscopy (CE). The suspected blood indicator (SBI) is a conventional tool used to automatically tag images depicting possible bleeding in the reading system. We aim to develop a deep learning-based system to detect blood content in images and compare its performance with that of the SBI. METHODS: We trained a deep convolutional neural network (CNN) system, using 27 847 CE images (6503 images depicting blood content from 29 patients and 21 344 images of normal mucosa from 12 patients). We assessed its performance by calculating the area under the receiver operating characteristic curve (ROC-AUC) and its sensitivity, specificity, and accuracy, using an independent test set of 10 208 small-bowel images (208 images depicting blood content and 10 000 images of normal mucosa). The performance of the CNN was compared with that of the SBI, in individual image analysis, using the same test set. RESULTS: The AUC for the detection of blood content was 0.9998. The sensitivity, specificity, and accuracy of the CNN were 96.63%, 99.96%, and 99.89%, respectively, at a cut-off value of 0.5 for the probability score, which were significantly higher than those of the SBI (76.92%, 99.82%, and 99.35%, respectively). The trained CNN required 250 s to evaluate 10 208 test images. CONCLUSIONS: We developed and tested the CNN-based detection system for blood content in CE images. This system has the potential to outperform the SBI system, and the patient-level analyses on larger studies are required.

Impact of Urgent Double-Balloon Enteroscopy on the Short-Term and Long-Term Outcomes in Overt Small Bowel Bleeding.

BACKGROUND: Double-balloon enteroscopy (DBE) is a safe and useful procedure for managing small bowel bleeding. However, there are limited studies regarding the preferable timing of DBE and its impact on long-term outcomes. AIM: We aimed to evaluate the association between the timing of DBE and the long-term outcomes of patients suspected of having overt small bowel bleeding who underwent DBE. METHODS: We retrospectively reviewed a prospectively collected database of patients who underwent DBE procedures between May 2004 and April 2016. The electronic medical records were reviewed, and interviews were conducted via mail and telephone. RESULTS: One-hundred sixty-five patients could be followed up. The bleeding source was detected during the initial DBE (DBE-positive group) for 102 patients. Sixty-three patients had no definite lesion during the initial DBE (DBE-negative group). Urgent DBE (DBE within 24 h after the last bleeding episode) was performed more often for the DBE-positive group (50/102; 49.0%) than for the DBE-negative group (10/63; 16.1%) (p < 0.0001). Nine patients in the DBE-positive group underwent curative surgery after diagnosis. Among the remaining DBE-positive patients, 38 of 93 (40.9%) had recurrent bleeding during 2675 days of follow-up. Twenty-one of 63 patients (33.3%) in the DBE-negative group had recurrent bleeding during 2490 days of follow-up. There was no significant difference between the two groups in terms of intervals without rebleeding (p = 0.17). CONCLUSION: Urgent DBE at the initial bleeding episode was useful for detecting lesions. However, the rebleeding rate was not dependent on the initial DBE results.

False-negative double-balloon enteroscopy in overt small bowel bleeding: long-term follow-up after negative results.

BACKGROUND AND AIM: Double-balloon enteroscopy (DBE) performed to investigate overt small bowel bleeding can miss the source of bleeding. We investigated the clinical outcomes of patients with negative DBE results for suspected overt small bowel bleeding, which is defined in the current guidelines as obscure gastrointestinal bleeding. METHODS: We reviewed the prospectively collected medical records of patients who underwent DBE at our hospital between May 1, 2004 and April 30, 2016. During this period, 297 patients underwent DBE for suspected overt small bowel bleeding. The first DBE yielded negative results for 83 patients (27.9%). Written interviews, telephone interviews, and medical records of these patients were reviewed in April 2017. Follow-up data were collected for 63 patients (75.9%). RESULTS: During a mean follow-up period of 83.5 months, re-bleeding occurred in 21 of 63 patients (33.3%) after a mean of 23.0 months after the first DBE yielded negative results. The bleeding source was identified in 19 of 21 patients (90.5%). In 15 of these 19 patients (78.9%), the source was the small intestine. Among these 15 patients, 14 (93.3%) had bleeding sites within reach of the first DBE and 3 (20%) experienced their first incidence of re-bleeding more than 3 years after the first DBE. The need for transfusion for the first bleeding episode was a predictor of re-bleeding (odds ratio 7.5; 95% confidence interval 1.7-33.0). CONCLUSIONS: False-negative DBE results for overt small bowel bleeding are not rare, and the first re-bleeding episode can occur 3 years later. Repeat DBE when re-bleeding occurs should be considered, even if the first DBE results were negative.

A case of a glomus tumor of the stomach resected by laparoscopy endoscopy cooperative surgery.

A 56-year-old woman who was found to have a submucosal tumor (SMT) of the stomach in a medical check-up was admitted to our hospital for a detailed investigation of the SMT. Upper gastrointestinal endoscopy revealed an SMT of 20mm at the anterior wall of the antrum of the stomach. Endoscopic ultrasonography showed a hyperechoic tumor in the fourth layer of the stomach wall. CT examination showed a strongly enhancing tumor on arterial phase images and persistent enhancement on portal venous phase images. Laparoscopy endoscopy cooperative surgery was performed with a diagnosis of SMT of the stomach highly suspicious of a glomus tumor. Immunohistochemistry revealed expression of α-SMA but no expression of desmin, c-kit, CD34, or S-100. The tumor was finally diagnosed as a glomus tumor of the stomach.