Association Between Vonoprazan and the Risk of Gastric Cancer After Helicobacter pylori Eradication.

BACKGROUND & AIMS: Potassium-competitive acid blockers (PCABs) have been increasingly used to treat upper gastrointestinal disorders, replacing proton pump inhibitors (PPIs). Whereas PPIs are associated with an increased risk of gastric cancer (GC) after Helicobacter pylori (Hp) eradication, it is uncertain whether PCABs carry the same risk. METHODS: Using a population-based claims database in Japan, we identified patients who were prescribed a clarithromycin-based first regimen of Hp eradication between 2015 and 2018. Patients who failed this regimen and those diagnosed with GC before or within 1 year after Hp eradication were excluded. We compared GC incidence between PCAB users and histamine type-2 receptor antagonist (H2RA) users, matching them on the basis of propensity scores calculated with considerations for age, sex, smoking, alcohol consumption, comorbidities, and co-administered medications. PCABs included only vonoprazan in this study. RESULTS: Among 54,055 patients, 568 (1.05%) developed GC during the follow-up period (mean, 3.65 years). The cumulative incidence of GC was 1.64% at 3 years, 2.02% at 4 years, and 2.36% at 5 years in PCAB users and 0.71% at 3 years, 1.04% at 4 years, and 1.22% at 5 years in H2RA users. The use of PCABs was associated with a higher GC risk (matched hazard ratio, 1.92; 95% confidence interval, 1.13-3.25; P = .016). Longer PCAB use and high-dose PCAB use were significantly associated with higher incidence of GC. Sensitivity analyses showed the risk of GC incidence among PCAB users was comparable with that of PPI users. CONCLUSIONS: The use of PCABs was associated with an increased risk of GC among Hp-eradicated patients, with duration/dose response effects.

High risk stigmata and treatment strategy for acute lower gastrointestinal bleeding: a nationwide study in Japan.

BACKGROUND: The rebleeding risks and outcomes of endoscopic treatment for acute lower gastrointestinal bleeding (ALGIB) may differ depending on the bleeding location, type, and etiology of stigmata of recent hemorrhage (SRH) but have yet to be fully investigated. We aimed to identify high risk endoscopic SRH and to propose an optimal endoscopic treatment strategy. METHODS: We retrospectively analyzed 2699 ALGIB patients with SRH at 49 hospitals (CODE BLUE-J Study), of whom 88.6 % received endoscopic treatment. RESULTS: 30-day rebleeding rates of untreated SRH significantly differed among locations (left colon 15.5 % vs. right colon 28.6 %) and etiologies (diverticular bleeding 27.5 % vs. others [e. g. ulcerative lesions or angioectasia] 8.9 %), but not among bleeding types. Endoscopic treatment reduced the overall rebleeding rate (adjusted odds ratio [AOR] 0.69; 95 %CI 0.49-0.98), and the treatment effect was significant in right-colon SRH (AOR 0.46; 95 %CI 0.29-0.72) but not in left-colon SRH. The effect was observed in both active and nonactive types, but was not statistically significant. Moreover, the effect was significant for diverticular bleeding (AOR 0.60; 95 %CI 0.41-0.88) but not for other diseases. When focusing on treatment type, the effectiveness was not significantly different between clipping and other modalities for most SRH, whereas ligation was significantly more effective than clipping in right-colon diverticular bleeding. CONCLUSIONS: A population-level endoscopy dataset allowed us to identify high risk endoscopic SRH and propose a simple endoscopic treatment strategy for ALGIB. Unlike upper gastrointestinal bleeding, the rebleeding risks for ALGIB depend on colonic location, bleeding etiology, and treatment modality.

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.

Long-term Risks of Recurrence After Hospital Discharge for Acute Lower Gastrointestinal Bleeding: A Large Nationwide Cohort Study.

BACKGROUND & AIMS: Currently, large, nationwide, long-term follow-up data on acute lower gastrointestinal bleeding (ALGIB) are scarce. We investigated long-term risks of recurrence after hospital discharge for ALGIB using a large multicenter dataset. METHODS: We retrospectively analyzed 5048 patients who were urgently hospitalized for ALGIB at 49 hospitals across Japan (CODE BLUE-J study). Risk factors for the long-term recurrence of ALGIB were analyzed by using competing risk analysis, treating death without rebleeding as a competing risk. RESULTS: Rebleeding occurred in 1304 patients (25.8%) during a mean follow-up period of 31 months. The cumulative incidences of rebleeding at 1 and 5 years were 15.1% and 25.1%, respectively. The mortality risk was significantly higher in patients with out-of-hospital rebleeding episodes than in those without (hazard ratio, 1.42). Of the 30 factors, multivariate analysis showed that shock index ≥1 (subdistribution hazard ratio [SHR], 1.25), blood transfusion (SHR, 1.26), in-hospital rebleeding (SHR, 1.26), colonic diverticular bleeding (SHR, 2.38), and thienopyridine use (SHR, 1.24) were significantly associated with increased rebleeding risk. Multivariate analysis of colonic diverticular bleeding patients showed that blood transfusion (SHR, 1.20), in-hospital rebleeding (SHR, 1.30), and thienopyridine use (SHR, 1.32) were significantly associated with increased rebleeding risk, whereas endoscopic hemostasis (SHR, 0.83) significantly decreased the risk. CONCLUSIONS: These large, nationwide follow-up data highlighted the importance of endoscopic diagnosis and treatment during hospitalization and the assessment of the need for ongoing thienopyridine use to reduce the risk of out-of-hospital rebleeding. This information also aids in the identification of patients at high risk of rebleeding.

Timing of colonoscopy in acute lower GI bleeding: a multicenter retrospective cohort study.

BACKGROUND AND AIMS: We aimed to determine the optimal timing of colonoscopy and factors that benefit patients who undergo early colonoscopy for acute lower GI bleeding. METHODS: We identified 10,342 patients with acute hematochezia (CODE BLUE-J study) admitted to 49 hospitals in Japan. Of these, 6270 patients who underwent a colonoscopy within 120 hours were included in this study. The inverse probability of treatment weighting method was used to adjust for baseline characteristics among early (≤24 hours, n = 4133), elective (24-48 hours, n = 1137), and late (48-120 hours, n = 1000) colonoscopy. The average treatment effect was evaluated for outcomes. The primary outcome was 30-day rebleeding rate. RESULTS: The early group had a significantly higher rate of stigmata of recent hemorrhage (SRH) identification and a shorter length of stay than the elective and late groups. However, the 30-day rebleeding rate was significantly higher in the early group than in the elective and late groups. Interventional radiology (IVR) or surgery requirement and 30-day mortality did not significantly differ among groups. The interaction with heterogeneity of effects was observed between early and late colonoscopy and shock index (shock index <1, odds ratio [OR], 2.097; shock index ≥1, OR, 1.095; P for interaction = .038) and performance status (0-2, OR, 2.481; ≥3, OR, .458; P for interaction = .022) for 30-day rebleeding. Early colonoscopy had a significantly lower IVR or surgery requirement in the shock index ≥1 cohort (OR, .267; 95% confidence interval, .099-.721) compared with late colonoscopy. CONCLUSIONS: Early colonoscopy increased the rate of SRH identification and shortened the length of stay but involved an increased risk of rebleeding and did not improve mortality and IVR or surgery requirement. Early colonoscopy particularly benefited patients with a shock index ≥1 or performance status ≥3 at presentation.

Risk and preventive factors of early-onset colorectal neoplasms: endoscopic and histological database analysis.

BACKGROUND AND AIM: The incidence of early-onset colorectal neoplasms has been increasing in both Western and Eastern countries. However, the risks and preventive factors for these neoplasms in Eastern countries remain unclear. METHODS: The data of 5580 patients who underwent colonoscopy between 2016 and 2021 were retrospectively evaluated. The primary outcome was advanced colorectal neoplasm (ACRN), defined as advanced adenomas (adenoma ≥10 mm, or with high-grade dysplasia or villous component) or adenocarcinoma. The clinical factors associated with ACRNs were determined for each age category (≤50 and >50 years), and the differences between the two categories were assessed. Odds ratios adjusted for age and sex were calculated. RESULTS: Among 1001 patients (age ≤50 years), ACRN was found in 94 (9.4%). In this younger category, male sex (adjusted odds ratio [aOR]:2.34, 95% confidence interval [CI]:1.51-3.63) and a family history of colorectal cancer (aOR:2.14, 95% CI:1.17-3.89) were significantly associated with higher odds of developing ACRNs. ACRNs were detected in 726 (15.9%) of 4579 patients (age >50 years). In the older age category, smoking (aOR:1.32, 95% CI:1.08-1.63) was significantly associated with a higher risk of ACRNs. Exercise of >3.5 metabolic equivalent of task (METs) (aOR,0.80; 95% CI:0.67-0.96) was significantly associated with a lower risk of ACRNs. CONCLUSION: The development of early-onset ACRNs was primarily associated with congenital factors, whereas that of late-onset ACRNs was associated with acquired ones. Colonoscopy is recommended for young male patients, particularly for those with a family history of colorectal cancer.

Early feeding reduces length of hospital stay in patients with acute lower gastrointestinal bleeding: A large multicentre cohort study.

AIM: No studies have compared the clinical outcomes of early and delayed feeding in patients with acute lower gastrointestinal bleeding (ALGIB). This study aimed to evaluate the benefits and risks of early feeding in a nationwide cohort of patients with ALGIB in whom haemostasis was achieved. METHODS: We reviewed data for 5910 patients with ALGIB in whom haemostasis was achieved and feeding was resumed within 3 days after colonoscopy at 49 hospitals across Japan (CODE BLUE-J Study). Patients were divided into an early feeding group (≤1 day, n = 3324) and a delayed feeding group (2-3 days, n = 2586). Clinical outcomes were compared between the groups by propensity matching analysis of 1508 pairs. RESULTS: There was no significant difference between the early and delayed feeding groups in the rebleeding rate within 7 days after colonoscopy (9.4% vs. 8.0%; p = 0.196) or in the rebleeding rate within 30 days (11.4% vs. 11.5%; p = 0.909). There was also no significant between-group difference in the need for interventional radiology or surgery or in mortality. However, the median length of hospital stay after colonoscopy was significantly shorter in the early feeding group (5 vs. 7 days; p < 0.001). These results were unchanged when subgroups of presumptive and definitive colonic diverticular bleeding were compared. CONCLUSION: The findings of this nationwide study suggest that early feeding after haemostasis can shorten the hospital stay in patients with ALGIB without increasing the risk of rebleeding.

The Association between Diverticular Rebleeding and Early-Morning Blood Pressure and Surge: A Prospective Observational Trial.

INTRODUCTION: Colonic diverticular bleeding is the major cause of lower gastrointestinal bleeding. Hypertension is a major risk factor for diverticular rebleeding. Direct evidence of an association between actual 24-h blood pressure (BP) and rebleeding is lacking. Therefore, we analyzed the association between 24-h BP and diverticular rebleeding. METHODS: We performed a prospective observational cohort trial involving hospitalized patients with colonic diverticular bleeding. We performed 24-h BP measurements (ambulatory BP monitoring [ABPM]) in the patients. The primary outcome was diverticular rebleeding. We evaluated the 24-h BP difference and the morning and pre-awaking BP surge between rebleeding and non-rebleeding patients. Morning BP surge was defined as early-morning systolic BP minus the lowest night systolic BP >45 mm Hg (highest quartile of morning BP surge). The pre-awaking BP surge was defined as the difference between morning BP and pre-awaking BP. RESULTS: Of 47 patients, 17 were excluded, leaving 30 who underwent ABPM. Of the 30 patients, 4 (13.33%) had rebleeding. The mean 24-h systolic and diastolic BP were 125.05 and 76.19 mm Hg in rebleeding patients and 129.98 and 81.77 mm Hg in non-rebleeding patients, respectively. Systolic BP at 5:00 (difference -23.53 mm Hg, p = 0.031) and 11:30 (difference -31.48 mm Hg, p = 0.006) was significantly lower in rebleeding patients than in non-rebleeding patients. Diastolic BP at 2:30 (difference -17.75 mm Hg, p = 0.023) and 5:00 (difference -16.12 mm Hg, p = 0.043) was significantly lower in rebleeding patients than in non-rebleeding patients. A morning surge was observed in one rebleeding patient and no non-rebleeding patients. The pre-awaking surge was significantly higher in rebleeding patients (28.44 mm Hg) than in non-rebleeding patients (9.30 mm Hg) (p = 0.015). CONCLUSION: Lower BP in the early-morning and a higher pre-awaking surge were risk factors for diverticular rebleeding. A 24-h ABPM can identify these BP findings and reduce the risk of rebleeding by enabling interventions in patients with diverticular bleeding.

LONG-HOSP Score: A Novel Predictive Score for Length of Hospital Stay in Acute Lower Gastrointestinal Bleeding - A Multicenter Nationwide Study.

INTRODUCTION: Length of stay (LOS) in hospital affects cost, patient quality of life, and hospital management; however, existing gastrointestinal bleeding models applicable at hospital admission have not focused on LOS. We aimed to construct a predictive model for LOS in acute lower gastrointestinal bleeding (ALGIB). METHODS: We retrospectively analyzed the records of 8,547 patients emergently hospitalized for ALGIB at 49 hospitals (the CODE BLUE-J Study). A predictive model for prolonged hospital stay was developed using the baseline characteristics of 7,107 patients and externally validated in 1,440 patients. Furthermore, a multivariate analysis assessed the impact of additional variables during hospitalization on LOS. RESULTS: Focusing on baseline characteristics, a predictive model for prolonged hospital stay was developed, the LONG-HOSP score, which consisted of low body mass index, laboratory data, old age, nondrinker status, nonsteroidal anti-inflammatory drug use, facility with ≥800 beds, heart rate, oral antithrombotic agent use, symptoms, systolic blood pressure, performance status, and past medical history. The score showed relatively high performance in predicting prolonged hospital stay and high hospitalization costs (area under the curve: 0.70 and 0.73 for derivation, respectively, and 0.66 and 0.71 for external validation, respectively). Next, we focused on in-hospital management. Diagnosis of colitis or colorectal cancer, rebleeding, and the need for blood transfusion, interventional radiology, and surgery prolonged LOS, regardless of the LONG-HOSP score. By contrast, early colonoscopy and endoscopic treatment shortened LOS. CONCLUSIONS: At hospital admission for ALGIB, our novel predictive model stratified patients by their risk of prolonged hospital stay. During hospitalization, early colonoscopy and endoscopic treatment shortened LOS.

Development and validation of a novel model for predicting stigmata of recent hemorrhage in acute lower gastrointestinal bleeding: Multicenter nationwide study.

OBJECTIVES: Stigmata of recent hemorrhage (SRH) directly indicate a need for endoscopic therapy in acute lower gastrointestinal bleeding (LGIB). Colonoscopy would be prioritized for patients with highly suspected SRH, but the predictors of colonic SRH remain unclear. We aimed to construct a predictive model for the efficient detection of SRH using a nationwide cohort. METHODS: We retrospectively analyzed 8360 patients admitted through hospital emergency departments for acute LGIB in the CODE BLUE-J Study (49 hospitals throughout Japan). All patients underwent inpatient colonoscopy. To develop an SRH predictive model, 4863 patients were analyzed. Baseline characteristics, colonoscopic factors (timing, preparation, and devices), and computed tomography (CT) extravasation were extensively assessed. The performance of the model was externally validated in 3497 patients. RESULTS: Colonic SRH was detected in 28% of patients. A novel predictive model for detecting SRH (CS-NEED score: ColonoScopic factors, No abdominal pain, Elevated PT-INR, Extravasation on CT, and DOAC use) showed high performance (area under the receiver operating characteristic curve [AUC] 0.74 for derivation and 0.73 for external validation). This score was also highly predictive of active bleeding (AUC 0.73 for derivation and 0.76 for external validation). Patients with low (0-6), intermediate (7-8), and high (9-12) scores in the external validation cohort had SRH identification rates of 20%, 31%, and 64%, respectively (P < 0.001 for trend). CONCLUSIONS: A novel predictive model for colonic SRH identification (CS-NEED score) can specify colonoscopies likely to achieve endoscopic therapy in acute LGIB. Using the model during initial management would contribute to finding and treating SRH efficiently.

Machine learning-based personalized prediction of gastric cancer incidence using the endoscopic and histologic findings at the initial endoscopy.

BACKGROUND AND AIMS: Accurate risk stratification for gastric cancer is required for optimal endoscopic surveillance in patients with chronic gastritis. We aimed to develop a machine learning (ML) model that incorporates endoscopic and histologic findings for an individualized prediction of gastric cancer incidence. METHODS: We retrospectively evaluated 1099 patients with chronic gastritis who underwent EGD and biopsy sampling of the gastric mucosa. Patients were randomly divided into training and test sets (4:1). We constructed a conventional Cox proportional hazard model and 3 ML models. Baseline characteristics, endoscopic atrophy, and Operative Link on Gastritis-Intestinal Metaplasia Assessment (OLGIM)/Operative Link on Gastritis Assessment (OLGA) stage at initial EGD were comprehensively assessed. Model performance was evaluated using Harrel's c-index. RESULTS: During a mean follow-up of 5.63 years, 94 patients (8.55%) developed gastric cancer. The gradient-boosting decision tree (GBDT) model achieved the best performance (c-index from the test set, .84) and showed high discriminative ability in stratifying the test set into 3 risk categories (P < .001). Age, OLGIM/OLGA stage, endoscopic atrophy, and history of malignant tumors other than gastric cancer were important predictors of gastric cancer incidence in the GBDT model. Furthermore, the proposed GBDT model enabled the generation of a personalized cumulative incidence prediction curve for each patient. CONCLUSIONS: We developed a novel ML model that incorporates endoscopic and histologic findings at initial EGD for personalized risk prediction of gastric cancer. This model may lead to the development of effective and personalized follow-up strategies after initial EGD.

Artificial intelligence versus expert endoscopists for diagnosis of gastric cancer in patients who have undergone upper gastrointestinal endoscopy.

AIMS: To compare endoscopy gastric cancer images diagnosis rate between artificial intelligence (AI) and expert endoscopists. PATIENTS AND METHODS: We used the retrospective data of 500 patients, including 100 with gastric cancer, matched 1:1 to diagnosis by AI or expert endoscopists. We retrospectively evaluated the noninferiority (prespecified margin 5 %) of the per-patient rate of gastric cancer diagnosis by AI and compared the per-image rate of gastric cancer diagnosis. RESULTS: Gastric cancer was diagnosed in 49 of 49 patients (100 %) in the AI group and 48 of 51 patients (94.12 %) in the expert endoscopist group (difference 5.88, 95 % confidence interval: -0.58 to 12.3). The per-image rate of gastric cancer diagnosis was higher in the AI group (99.87 %, 747 /748 images) than in the expert endoscopist group (88.17 %, 693 /786 images) (difference 11.7 %). CONCLUSIONS: Noninferiority of the rate of gastric cancer diagnosis by AI was demonstrated but superiority was not demonstrated.

Efficacy of Early Video Capsule Endoscopy for Acute Overt Lower Gastrointestinal Bleeding with Colonic Diverticulosis: A Prospective Observational Study.

BACKGROUND/AIMS: Although most patients with presumptive colonic diverticular bleeding (CDB) do not undergo a small bowel investigation in clinical practice, no prospective study supports this management. We evaluated the utility of early small bowel capsule endoscopy (CE) after negative colonoscopy results. METHODS: This prospective study evaluated the diagnostic yield of early small bowel CE (≤3 days from visit) for consecutive patients with acute-onset hematochezia, when colonoscopy found colonic diverticulosis but did not identify the definite bleeding source (n = 51; presumptive CDB). As a matched control for comparing clinical outcomes, presumptive CDB patients without CE (n = 51) were retrospectively extracted. RESULTS: On CE for the prospective cohort, the rates of total positive findings, P2 findings (high bleeding potential according to the P classification), and blood pooling in the colon were 57%, 12% (ulceration, 8%; angioectasia, 4%), and 24%, respectively. The rates of rebleeding within 30 and 365 days were 16% and 29% in the prospective cohort with CE, respectively, and were not significantly different from those in the retrospective cohort without CE (10% and 25%, respectively). In addition, thromboembolism and mortality within 30 and 365 days were not significantly different between those with and without CE. CONCLUSION: Early CE detected a suspected small bowel bleeding source in 12% of acute-onset presumptive CDB patients but did not significantly improve major clinical outcomes. Therefore, routine CE is unnecessary for presumptive CDB patients after colonoscopy (UMIN000026676).

Chemoprevention for Colorectal Cancers: Are Chemopreventive Effects Different Between Left and Right Sided Colorectal Cancers?

BACKGROUND AND AIMS: Recent studies have suggested that right- and left-sided colorectal cancers (CRCs) are molecularly distinct. In this study, we examined the association between the risk of right- and left-sided CRC and drug use to estimate their chemopreventive effects METHODS: This multicenter retrospective cohort study was conducted using the data of hospitalized patients between 2014 and 2019 from nine hospital databases. The primary outcomes were right- and left-sided CRC. We evaluated the association of CRCs with drug use and clinical factors. Odds ratios adjusted for age, sex, Charlson Comorbidity Index scores, and smoking status were calculated. We also compared the transcriptional profiling in precancerous lesions, including sessile serrated lesions (SSLs) RESULTS: A total of 307,938 patients, including 2745 with right-sided CRC and 4819 with left-sided CRC, were analyzed. The use of nonsteroidal anti-inflammatory drugs (NSAIDs), aspirin, cyclooxygenase-2 inhibitors, and steroids was associated with a lower risk of both right- and left-sided CRCs. In contrast, statins, other lipid-lowering agents, and metformin were associated with a lower risk of left-sided CRC. Transcriptomic analysis showed that SSL, which predominantly develops in the right colon, was associated with a lower expression of lipid metabolism-related genes. CONCLUSIONS: Targeting lipid metabolism may be useful for chemoprevention of left-sided CRCs, while development of right-sided CRCs may be independent of this pathway.

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.

Automatic detection of colorectal neoplasia in wireless colon capsule endoscopic images using a deep convolutional neural network.

BACKGROUND: Although colorectal neoplasms are the most common abnormalities found in colon capsule endoscopy (CCE), no computer-aided detection method is yet available. We developed an artificial intelligence (AI) system that uses deep learning to automatically detect such lesions in CCE images. METHODS: We trained a deep convolutional neural network system based on a Single Shot MultiBox Detector using 15 933 CCE images of colorectal neoplasms, such as polyps and cancers. We assessed performance by calculating areas under the receiver operating characteristic curves, along with sensitivities, specificities, and accuracies, using an independent test set of 4784 images, including 1850 images of colorectal neoplasms and 2934 normal colon images. RESULTS: The area under the curve for detection of colorectal neoplasia by the AI model was 0.902. The sensitivity, specificity, and accuracy were 79.0 %, 87.0 %, and 83.9 %, respectively, at a probability cutoff of 0.348. CONCLUSIONS: We developed and validated a new AI-based system that automatically detects colorectal neoplasms in CCE images.

Natural history of upper and lower gastrointestinal bleeding in hemodialysis patients: A dual-center long-term cohort study.

BACKGROUND AND AIM: Knowledge on the risk of gastrointestinal (GI) bleeding in hemodialysis patients is limited. We evaluated the risk of GI bleeding in hemodialysis patients compared with non-hemodialysis patients. METHODS: We performed a retrospective cohort study from 1996 to 2017 at the Graduate School of Medicine, University of Tokyo, and Horinouchi Hospital. We analyzed patients on hemodialysis for chronic renal failure and controls not on hemodialysis. The primary endpoint was GI bleeding. A survival analysis was performed to estimate the cumulative incidence and hazard ratio of GI bleeding. RESULTS: A total of 14 451 patients were analyzed (417 hemodialysis and 14 034 non-hemodialysis patients). In total, 524 GI bleeding events occurred. Upper and lower GI bleeding occurred in 432 and 92 patients in the hemodialysis and non-hemodialysis groups, respectively. The most frequent source of upper and lower GI bleeding was gastric ulcer and colonic diverticular bleeding, respectively. The cumulative incidence of GI bleeding was 4.44% at 1 year, 7.15% at 3 years, and 10.40% at 5 years in hemodialysis patients; the respective rates were 2.35%, 2.98%, and 3.79% in non-hemodialysis patients during a mean follow-up period of 3.5 years. Hemodialysis was significantly associated with an increased risk of GI bleeding after adjustment (hazard ratio 1.67, P = 0.01, 95% confidence interval 1.13-2.50). CONCLUSIONS: Hemodialysis patients had a GI bleeding rate of 10% over 5 years, and hemodialysis was a risk factor for GI bleeding.

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.

Automatic detection of different types of small-bowel lesions on capsule endoscopy images using a newly developed deep convolutional neural network.

BACKGROUND : Previous computer-aided detection systems for diagnosing lesions in images from wireless capsule endoscopy (WCE) have been limited to a single type of small-bowel lesion. We developed a new artificial intelligence (AI) system able to diagnose multiple types of lesions, including erosions and ulcers, vascular lesions, and tumors. METHODS : We trained the deep neural network system RetinaNet on a data set of 167 patients, which consisted of images of 398 erosions and ulcers, 538 vascular lesions, 4590 tumors, and 34 437 normal tissues. We calculated the mean area under the receiver operating characteristic curve (AUC) for each lesion type using five-fold stratified cross-validation. RESULTS : The mean age of the patients was 63.6 years; 92 were men. The mean AUCs of the AI system were 0.996 (95 %CI 0.992 - 0.999) for erosions and ulcers, 0.950 (95 %CI 0.923 - 0.978) for vascular lesions, and 0.950 (95 %CI 0.913 - 0.988) for tumors. CONCLUSION : We developed and validated a new computer-aided diagnosis system for multiclass diagnosis of small-bowel lesions in WCE images.

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.