Gastric cancer is a leading cause of mortality worldwide, yet the development of computer-aided diagnosis (CAD) systems for its early detection is hindered by the scarcity of high-quality datasets utilizing Magnifying Endoscopy with Narrow-Band Imaging (ME-NBI). Addressing this gap, we present EndoWLI-NBI, a large-scale dataset comprising 11,971 ME-NBI images from 868 patients, specifically curated to cover the full gastric precancerous cascade—including Intestinal Metaplasia & Chronic Gastritis, Low- and High-Grade Intraepithelial Neoplasia, and Early Gastric Cancer. Unlike existing public datasets, every image in EndoWLI-NBI is rigorously mapped to a biopsy-confirmed histopathological diagnosis, providing a definitive gold standard for label reliability. Technical validation using state-of-the-art deep learning benchmarks demonstrates the dataset’s quality and separability, while highlighting the intrinsic visual challenges in distinguishing high-grade neoplasia from early cancer. This dataset serves as a critical, medically verified resource for advancing fine-grained lesion classification, domain adaptation research, and endoscopic training in gastroenterology.

Background and Aims: Sessile serrated lesions (SSLs) are precancerous colorectal polyps with high malignant potential, yet their endoscopic recognition remains challenging. Most artificial intelligence (AI) models for polyp classification fail to specifically identify SSLs, often misclassifying them as hyperplastic polyps. This study aimed to develop and validate SERRATE (Sessile Serrated lesion Recognition and Risk Assessment Tool for Endoscopy), a deep learning model for distinguishing SSLs from conventional adenomas and hyperplastic polyps. Methods : We retrospectively collected colonoscopic images from 696T patients at Peking Union Medical College Hospital (January 2017-September 2024), yielding >20,000 images. The balanced training dataset comprised 1,120 SSL images, 1,697 conventional adenoma images, and 1,207 hyperplastic polyp images. A Swin Transformer model was trained using both white-light imaging and narrow-band imaging (NBI). Model performance was evaluated on a prospective internal validation cohort (300 patients, January 2025-April 2025; ClinicalTrials.gov NCT06773832) and the external POLAR dataset. Results : The optimal model utilized NBI images with 2×mucosal context expansion, achieving 82.42% accuracy in internal validation. For three-class classification (SSL vs. conventional adenoma vs. hyperplastic polyp), the model demonstrated 73.44% overall accuracy (precision 82.47%, recall 95.21%, specificity 91.21%, F1-score 82.17%) on prospective internal validation and 69.46% accuracy on external validation. For binary neoplasia detection, accuracy reached 91.07% across validation sets. Conclusions : SERRATE represents the first AI model specifically designed for SSL recognition with balanced training data. The model achieved clinically relevant performance for SSL differentiation, potentially improving colorectal cancer prevention through enhanced serrated lesion management.

Background/objective Prophylactic endoloop or hemoclip application and endoscopic mucosal resection (EMR) have been usually performed for large pedunculated colorectal polyps. However, their comparison remains unclear. This study aimed to compare the efficacy and safety of the prophylactic pretreatment and EMR in the large pedunculated polyps. Methods We reviewed 1994 pedunculated polyps detected in Peking Union Medical College Hospital from 2012 to 2024. Large pedunculated polyps (head size ≥2.0 cm) receiving EMR or prophylactic closure before polypectomy were included. Polyp morphology, endoscopic procedural variables, and adverse events were compared using propensity score matching and inverse probability of treatment weights. Logistic regression was performed to analyze the risk factors of adverse events. Results 204 pedunculated polyps undergoing prophylactic pretreatment and 262 polyps resected via EMR were included. Less clips (P = 0.002) and lower operational material-related costs (P < 0.001) presented in the prophylactic closure group than the EMR group, while the procedural time was comparable (prophylactic closure vs. EMR: median 13.0 vs. 10.0 min, P = 0.059). There were no differences in postpolypectomy bleeding (15.2% vs. 12.5%, P = 0.466) and postpolypectomy coagulation syndrome (1.5% vs. 0.7%, P = 0.509) between the two groups. Stalk width ≥0.6 cm (OR = 2.34, P = 0.02), number of hemoclips (OR = 1.44, P < 0.01), procedure time ≥10 min (OR = 2.03, P = 0.08), and malignant lesion (OR = 1.74, P = 0.08) were associated with high risks of immediate bleeding. The endoscopic procedure did not affect the risks of immediate or delayed bleeding. Conclusion The prophylactic closure with hemoclip or endoloop and conventional EMR are both effective and safe methods for large pedunculated colorectal polyps. Prophylactic closure has advantages in the hemoclips demand and average procedure costs compared to EMR.

Colorectal polyps are precursors to colorectal cancer, a leading cause of cancer-related mortality worldwide. While early detection and removal during colonoscopy are effective preventive measures, automated polyp classification remains challenging due to substantial appearance variability and the domain gap between curated public datasets and artifact-heavy real-world clinical frames. In this study, we propose a context-aware AI system for three-class polyp classification—adenomas, hyperplastic polyps, and sessile serrated lesions (SSLs)—using a large-scale private dataset of 4,094 real-world colonoscopy images collected under routine clinical conditions. Unlike conventional pipelines that apply uniform preprocessing across modalities, our unified framework introduces Modality-Adaptive Context Expansion, which tailors contextual scaling to modality-specific imaging characteristics. Specifically, we integrate YOLOv8-based field-of-view extraction, Mask R-CNN–based localization, and a Swin Transformer classifier with modality-specific context scaling (1.5× for WLI and 2.0× for NBI) to better balance noise suppression and discriminative contextual cues. Furthermore, we incorporate a confidence-based abstention mechanism (threshold > 0.8) to defer low-confidence predictions for human review, enabling a practical reliability–coverage tradeoff in a clinical decision-support workflow. Extensive experiments and ablation studies show that, on NBI images, the proposed system achieves high specificity (91.21%) on real-world clinical frames. Moreover, compared to standard fixed-crop pipelines, Modality-Adaptive Context Expansion improves NBI classification performance, increasing accuracy by +9.2% and specificity by +4.6%.

Gastric cancer (GC) is a leading cause of cancer-related morbidity and mortality globally. Endoscopic submucosal dissection (ESD) has improved survival rates for early gastric cancer (EGC), but complications such as delayed bleeding remain a concern. This multicenter retrospective study aimed to validate the BEST-J prediction model for post-ESD bleeding in two Beijing hospitals and identify additional risk factors influencing this complication. We analyzed data from 688 patients who underwent ESD for EGC across two institutions. Patient demographics, procedural details, and comorbidities were collected. The BEST-J score was calculated for each patient, and Firth’s penalized logistic regression was performed to identify independent risk factors for delayed bleeding. Our findings revealed that the BEST-J score had limited discriminative power in predicting post-ESD bleeding (AUC = 0.583). However, two independent risk factors—previous cerebral infarction (p = 0.012) and bile reflux (p = 0.012)—were significantly associated with delayed bleeding. This study underscores the necessity for ongoing validation of predictive models such as BEST-J, particularly across different populations. The identification of cerebral infarction and bile reflux as significant predictors highlights the need to integrate additional factors into risk assessments.

Background:Managing patients with primary biliary cholangitis (PBC) who demonstrate an inadequate response to ursodeoxycholic acid or experience intolerable side effects remains a significant clinical challenge. Objectives:This study aims to investigate the efficacy and safety of peroxisome proliferator-activated receptor (PPAR) agonists in the treatment of PBC. Design:Meta-analysis and systematic review. Methods:A systematic search of publications in PubMed, Embase, Web of Science, and the Cochrane Central Register of Controlled Trials was performed. Randomized controlled trials published in English that involved the treatment of PPAR agonists and reported on the levels of alkaline phosphatase (ALP), biochemical response rates, pruritus score, or severe and serious adverse events (AEs) were selected. The primary outcomes assessed were the effects of PPAR agonists on ALP levels and biochemical response rates. Secondary outcomes included the rates of severe or serious AEs and relief of pruritus. Results:Fourteen studies with 1137 patients were included. Compared to the control group, PPAR agonists significantly reduced ALP levels by a mean difference of -155.87 U/L (95% confidence interval (CI): -208.30 to -103.44; random-effects). Patients who received PPAR agonists showed a significantly higher biochemical response rate (risk ratio (RR), 4.42; 95% CI: 2.37-8.26; random-effects). Furthermore, there was no significant difference in the rate of severe (RR, 1.05; 95% CI, 0.49-2.28) or serious AEs (RR, 1.02; 95% CI, 0.65-1.60) between the PPAR agonists and placebo groups. Conclusion:PPAR agonists are effective and safe to treat patients with PBC. PROSPERO trial registration:CRD42024545743.

In the application of deep learning for gastric cancer detection, the quality of the data set is as important as, if not more, the design of the network architecture. However, obtaining labeled data, especially in fields such as medical imaging to detect gastric cancer, can be expensive and challenging. This scarcity is exacerbated by stringent privacy regulations and the need for annotations by specialists. Conventional methods of data augmentation fall short due to the complexities of medical imagery. In this paper, we explore the use of diffusion models to generate synthetic medical images for the detection of gastric cancer. We evaluate their capability to produce realistic images that can augment small datasets, potentially enhancing the accuracy and robustness of detection algorithms. By training diffusion models on existing gastric cancer data and producing new images, our aim is to expand these datasets, thereby enhancing the efficiency of deep learning model training to achieve better precision and generalization in lesion detection. Our findings indicate that images generated by diffusion models significantly mitigate the issue of data scarcity, advancing the field of deep learning in medical imaging.

Esophageal stricture is a common complication after esophageal endoscopic submucosal dissection (ESD). Pathological scars may predict stricture development, but clinical studies exploring their relationship are limited. This study aimed to investigate the link between concurrent pathological scarring and post-ESD esophageal stricture and develop a predictive risk model. This retrospective single-center study included patients who underwent ESD for superficial esophageal lesions (SELs) over the past decade. Data on demographics, endoscopy, pathology, and concurrent scarring were collected. Multivariate logistic regression identified risk factors, followed by validation using subgroup analysis and propensity score matching. A nomogram was developed and internally validated. Of 255 patients, 28 (11.0

Non-alcoholic fatty liver disease (NAFLD) and its advanced form, non-alcoholic steatohepatitis (NASH), are the leading causes of chronic liver disease globally. They are driven by complex mechanisms where inflammation plays a pivotal role in disease progression. Current therapies, including lifestyle changes and pharmacological agents, are limited in efficacy, particularly in addressing the advanced stages of the disease. Emerging approaches targeting inflammation, metabolic dysfunction, and fibrosis offer promising new directions, though challenges such as treatment complexity and heterogeneity persist. This review concludes the main therapeutic targets and approaches to manage inflammation currently and emphasizes the critical need for future drug development and combination therapy for NAFLD/NASH management.

Background: Rectal neuroendocrine tumors (NETs) ≤ 10 mm are commonly managed by endoscopic resection. However, the optimal technique remains controversial. We aimed to compare the efficacy and safety of a simplified endoscopic mucosal resection (sEMR) technique, performed without submucosal injection, with conventional endoscopic submucosal dissection (ESD) for small rectal NETs. Methods: This retrospective, single-center study included 74 patients with histologically confirmed rectal NETs ≤ 10 mm treated with sEMR (n = 37) or ESD (n = 37) between January 2022 and January 2025. Patients in the ESD group were matched 1:1 by age and gender. Baseline characteristics, procedural outcomes, histopathologic findings, and cost were analyzed. The primary outcome was histological complete resection (R0) rate; secondary outcomes included en bloc resection, intraoperative bleeding, tumor-to-margin distance, operation time, and material costs. Results: The R0 resection rate was significantly higher in the sEMR group compared to the ESD group (91.9% vs. 67.6%; p = 0.019). Tumor-to-margin distance was also significantly greater in the sEMR group [median (IQR): 0.2 (0.1–0.3) mm vs. 0.1 (0–0.2) mm; p = 0.024]. Intraoperative bleeding was less frequent in the sEMR group (2.7% vs. 21.6%; p = 0.028). Median operation time (409 vs. 1469 s; p < 0.001) and material cost (1486 vs. 6390 CNY; p < 0.001) were both significantly lower in the sEMR group. Conclusions: Compared with ESD, the simplified EMR technique without submucosal injection demonstrated higher R0 resection rates, lower bleeding risk, shorter operation time, and lower costs for small rectal NETs. sEMR may offer a safe, efficient, and cost-effective alternative in selected patients.

Current treatments for pancreatic ductal adenocarcinoma (PDAC) fall short of meeting clinical needs, highlighting the urgent need for a comprehensive understanding of PDAC progression, which involves not only biochemical signals but also essential biomechanical cues. Here, we used a CRISPR-Cas9 screen in an orthotopic xenograft model to explore PDAC dynamics. The RNA binding protein DEAD-box helicase 3X-linked (DDX3X) was identified as a pivotal oncogene and biomechanical checkpoint. Specifically, DDX3X up-regulation in PDAC promoted tumorigenesis and metastasis, primarily through the transcriptional repressor TLE family member 2 (TLE2). Dysregulation of DDX3X in the tumor destabilized TLE2 messenger RNA and therefore disrupted the interaction with KLF4 (KLF transcription factor 4), leading to increased expression of myosin light chain 9 (MYL9). This change remodeled F-actin, enhancing tumor cell traction forces and consequently facilitating tumor metastasis. Targeting the DDX3X-TLE2-MYL9 pathway considerably reduces PDAC progression. This research reveals a promising approach for treating PDAC by focusing on biomechanical cues.

Objective To assess the diagnostic accuracy of bowel sound analysis for irritable bowel syndrome (IBS) with a systematic review and meta-analysis. Methods We searched MEDLINE, Embase, the Cochrane Library, Web of Science, and IEEE Xplore databases until September 2023. Cross-sectional and case-control studies on diagnostic accuracy of bowel sound analysis for IBS were identified. We estimated the pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio with a 95% confidence interval (CI), and plotted a summary receiver operating characteristic curve and evaluated the area under the curve. Results Four studies were included. The pooled diagnostic sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio were 0.94 (95% CI, 0.87–0.97), 0.89 (95% CI, 0.81–0.94), 8.43 (95% CI, 4.81–14.78), 0.07 (95% CI, 0.03–0.15), and 118.86 (95% CI, 44.18–319.75), respectively, with an area under the curve of 0.97 (95% CI, 0.95–0.98). Conclusions Computerized bowel sound analysis is a promising tool for IBS. However, limited high-quality data make the results' validity and applicability questionable. There is a need for more diagnostic test accuracy studies and better wearable devices for monitoring and analysis of IBS.

ObjectiveTo explore the application value of endoscopic ultrasound fine-needle aspiration (EUS-FNA) and endoscopic retrograde cholangiopancreatography (ERCP) in the diagnosis and treatment ofpatients with pancreatic cancer combined with obstructive jaundice.MethodsClinical data of patients hospitalized in the Department of Gastroenterology of Peking Union Medical College Hospital who underwent ERCP biliary stent drainage for pancreatic cancer combined with obstructive jaundice from January 1, 2023 to February 26, 2024 were retrospectively collected. They were categorized into the fusion technology group and the simple ERCP group according to whether EUS-FNA was performed in the same endoscopic unit. The differences in pathologic diagnosis rate, ERCP drainage success rate, postoperative complication rate and patients' single hospitalization time were compared between the two groups.ResultsA total of 161 patients with pancreatic cancer combined with obstructive jaundice who underwent ERCP biliary stent drainage meeting the inclusion and exclusion criteria were enrolled, of which 80 were in the fusion technique group and 81 were in the simple ERCP group. The pathological diagnosis rate in the fusion technique group was higher than that in the simple ERCP group[92.50%(74/80) vs. 46.15%(12/26), P < 0.001], and the ERCP drainage success rate[82.50%(66/80) vs. 86.42%(70/81), P=0.360], the overall postoperative complication rate[5.00%(4/80) vs. 2.47%(2/81), P=0.443], and single hospitalization time[4 (4, 7)d vs. 5 (3, 9)d, P=0.397] were not statistically different from simple ERCP groups.ConclusionThe fusion of EUS-FNA and ERCP enhances the efficiency of diagnosis and treatment for pancreatic carcinoma, warranting widespread adoption and further research.

Twenty-four hours continuous intraocular pressure (IOP) monitoring is beneficial for glaucoma care. Contact lens sensors using LC technology can achieve non-invasive continuous IOP measuring around the clock and are relatively simple in principle and structure, thus dominating the mainstream. Among them, stretchable inductive LC contact lens sensors constructed with liquid metal have advantages in signal quality as well as wearing comfort, but currently its sensitivity is slightly insufficient. Here, we propose an LC sensor that constructed with liquid metal to form the stretchable inductance and capacitance further. The capacitive plate and inductive coil can response to IOP changes simultaneously, thus able to enhance the sensitivity in principle. We modeled the sensing mechanism and conducted design, fabrication, as well as various tests. The device exhibits good characteristics, including reliability, good signal quality, etc. Especially, it has a threefold increase in sensitivity, exceeding the current state-of-the-art contact lens sensors.

BACKGROUND:Although chronic erosive gastritis (CEG) is common, its clinical characteristics have not been fully elucidated. The lack of consensus regarding its treatment has resulted in varied treatment regimens.AIM:To explore the clinical characteristics, treatment patterns, and short-term outcomes in CEG patients in China.METHODS:We recruited patients with chronic non-atrophic or mild-to-moderate atrophic gastritis with erosion based on endoscopy and pathology. Patients and treating physicians completed a questionnaire regarding history, endoscopic findings, and treatment plans as well as a follow-up questionnaire to investigate changes in symptoms after 4 wk of treatment.RESULTS:Three thousand five hundred sixty-three patients from 42 centers across 24 cities in China were included. Epigastric pain (68.0%), abdominal distension (62.6%), and postprandial fullness (47.5%) were the most common presenting symptoms. Gastritis was classified as chronic non-atrophic in 69.9% of patients. Among those with erosive lesions, 72.1% of patients had lesions in the antrum, 51.0% had multiple lesions, and 67.3% had superficial flat lesions. In patients with epigastric pain, the combination of a mucosal protective agent (MPA) and proton pump inhibitor was more effective. For those with postprandial fullness, acid regurgitation, early satiety, or nausea, a MPA appeared more promising.CONCLUSION:CEG is a multifactorial disease which is common in Asian patients and has non-specific symptoms. Gastroscopy may play a major role in its detection and diagnosis. Treatment should be individualized based on symptom profile.