Early gastric cancer (EGC) is increasingly managed by endoscopic resection (ER); however, lymph node metastasis (LNM), which occurs in approximately 5%-10% of cases, remains the key determinant for recommending additional gastrectomy. Current guideline-based strategies, including the eCura system, provide structured risk stratification but rely on categorical decision-making and may lead to overtreatment, as nearly 90% of patients undergoing additional surgery do not have LNM. Artificial intelligence (AI) has emerged as a promising tool for improving LNM predictions. Machine learning models using clinicopathological variables have demonstrated promising discriminatory performance (area under the curve, 0.69-0.94), often outperforming conventional scoring systems such as eCura. However, these approaches rely on predefined variables and are susceptible to interobserver variability in pathological assessments. Whole-slide image-based AI directly analyzes histopathological images, offering an objective and reproducible approach. Although evidence for EGC is limited, recent multicenter data have demonstrated promising results using routine hematoxylin and eosin-stained slides. Beyond the LNM risk, treatment decisions should also consider patient factors such as age, comorbidities, and competing mortality risks. AI is expected to support treatment decision-making by integrating these multidimensional factors, enabling more personalized and risk-adapted management after ER.
Linear endoscopic ultrasound (EUS) has emerged as an essential tool in endohepatology, enabling high-resolution evaluation of the hepatic parenchyma, vascular structures, and segmental anatomy of the liver via transgastric and transduodenal imaging. As the role of linear EUS has expanded to include tissue acquisition, biliary drainage, portal pressure measurement, and targeted hepatic interventions, a thorough understanding of the liver anatomy has become indispensable to ensure procedural safety and efficacy. However, EUS-based liver evaluation remains technically challenging because of its use of dynamic imaging planes and the absence of fixed anatomical reference points. This review proposes a structured, station-based approach that integrates Couinaud's segmental anatomy with reproducible vascular and ligamentous landmarks. By correlating the transgastric and transduodenal imaging views with the corresponding portal venous, hepatic venous, and ligamentous axes, this framework provides a systematic method for accurate segmental identification. Adoption of this approach may enhance diagnostic accuracy, facilitate safe therapeutic interventions, and promote more structured and reproducible training frameworks.
Background/Aims:Computer-aided detection (CADe) improves adenoma detection; however, its performance for sessile serrated lesion (SSL) detection remains inconsistent. We hypothesized that models trained on expert-curated, histopathologically-confirmed datasets would improve SSL detection compared to models trained on public datasets without verified labels. Methods:Two CADe models with identical Visual Geometry Group 16 architectures were trained using different datasets: Model A on public datasets (~54,000 frames) and model B on a histopathologically-confirmed private dataset (~120,000 frames) derived from examinations performed by endoscopists with adenoma detection rates of >35%. Validation was conducted using 31 independent colonoscopy videos obtained from another endoscopist. Diagnostic performance was evaluated using event- and frame-based analyses. Results:Model B demonstrated a significantly higher event-based sensitivity than model A (99.7% vs. 39.9%, p<0.001). The detection of SSLs was markedly improved with model B. Frame-based sensitivity and F1-score were also higher for model B. Although model B generated more false positives per video (2.55 vs. 0.16, p<0.001), these alerts were brief and unlikely to meaningfully interfere with the endoscopic workflow. Conclusions:CADe models trained on expert-curated, histopathologically-confirmed datasets showed improved detection of neoplastic colorectal lesions, including SSLs, compared to models trained on public datasets. These findings highlight the importance of clinically curated datasets for optimizing the CADe performance for subtle colorectal lesions.
Background/Aims:Colorectal endoscopic submucosal dissection (C-ESD) is a technically demanding procedure with an extended learning curve. We aimed to characterize the learning process after introduction of the pocket-creation method (PCM) and provide practical guidance for endoscopists. Methods:We retrospectively analyzed 147 colorectal lesions treated with PCM-assisted C-ESD. Learning curves for dissection speed and procedure time were evaluated using locally weighted regression. Cumulative sum (CUSUM) analysis was used to identify the point of technical stabilization, with a predefined acceptable failure rate of 10%. Outcomes were compared between the pre- and post-stabilization phases, and multivariate logistic regression was used to identify the predictors of technical failure. Results:Dissection speed increased and procedure time decreased with experience. CUSUM analysis identified stabilization at 60 cases. Compared with the post-stabilization phase, the pre-stabilization phase showed higher technical failure rates and lower en bloc and R0 resection rates. The learning phase was the only independent predictor of technical failure. Perforation occurred in 5.4% of cases and was managed predominantly endoscopically. Conclusions:PCM-assisted C-ESD showed a distinct learning curve, with stabilization achieved after 60 cases. Technical failure is primarily driven by operator experience rather than by lesion complexity.
As endoscopic therapy has progressed, various endoscopic closure techniques have been developed to prevent adverse events and to facilitate complex therapeutic procedures. The development of closure techniques began in the 1970s with the introduction of clipping for mucosal defects after gastric polypectomy. Currently, these techniques are applied primarily to prevent post-procedural bleeding from mucosal defects resulting from endoscopic submucosal dissection; to repair iatrogenic perforations involving full-thickness defects; and to complete endoscopic full-thickness resection. They can be categorized into clipping, tissue apposition, suturing, and stapling, each with distinct advantages and disadvantages. Therefore, familiarity with each method and the ability to select the most appropriate technique for a given situation are essential.
