YMYuichi Mori

Yuichi Mori

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Yuichi Mori has a total of 623 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Development of a Novel Real-Time Computer-Assisted Colonoscopy Diagnostic Tool for Colorectal Neoplasia and Cancer to Enable Personalized Patient Management: Protocol for a Multicenter Proof-of-Concept Study

    by Olaolu Olabintan, Angelo Brunori, Natalie Halvorsen, Nils Andersen, Verónica G Zubiaurre, Mehrnaz Mirzaei, Ina Dam Pedersen, Alessandro D’APrano, Johnathan Sun, Ahmed Ghazwan Waraich, et al.

    Background: Accurate real-time classification of colorectal polyps during colonoscopy is essential for risk stratification, therapeutic decision-making, and surveillance planning, with important implications for colorectal cancer prevention, health care costs, and patient burden. Despite advances in optical diagnosis techniques, including virtual chromoendoscopy and magnification, widespread adoption remains limited because of high operator dependency and variable diagnostic accuracy. Existing computer-assisted diagnosis (CADx) systems predominantly provide binary classification of colorectal lesions, thereby constraining their clinical utility for nuanced endoscopic decision-making. Objective: This study aims to develop and validate a real-time, multiclass AI-based CADx system capable of classifying colorectal lesions into 5 categories: nonneoplastic lesions, low-grade adenomas, high-grade adenomas, sessile serrated lesions (SSL), and invasive colorectal cancer. Methods: This multicenter international study consists of 2 phases: model development and proof-of-concept validation. During the model development phase, a large, diverse dataset of >17,000 deidentified endoscopic images and video frames will be prospectively collected from 8 European centers and 1 Japanese center, capturing a broad spectrum of lesion types and patient demographics. A semisupervised learning framework using contrastive representation learning (SimCLR [simple framework for contrastive learning of visual representations]) will be used to leverage both labeled and unlabeled data, enhancing feature extraction and classification performance. Expert consensus annotation will be used for labeled data. Lesion localization will be supported by an integrated computer-assisted detection (CADe) module, while model interpretability will be facilitated through gradient-weighted class activation mapping (Grad-CAM). Proof-of-concept validation will be conducted using 250 previously unseen lesion images that will be independently assessed by expert endoscopists. The primary outcome is multiclass diagnostic accuracy, with a predefined target accuracy threshold of 85%, alongside sensitivity and specificity for invasive cancer detection. Secondary outcomes include interobserver agreement, performance across demographic and lesion subgroups, and a qualitative analysis of misclassification patterns. All data will be securely managed within a REDCap database, with strict deidentification protocols and plans to generate an open-access research resource. Results: Ethics approval has been obtained at all participating centres. Data transfer agreements have been executed by the majority of collaborating sites, with de-identified polyp images and videos currently being uploaded to a secure central database. Conclusions: This study aims to deliver a clinically applicable, real-time multiclass CADx system that extends beyond current binary AI models and supports more granular, personalized endoscopic decision-making. By integrating semisupervised learning, diverse international data, and explainable AI techniques, this approach has the potential to substantially advance the methodological foundations and clinical utility of AI-assisted optical diagnosis in routine colonoscopy, contributing to improved colorectal cancer prevention and patient care.

  • Curriculum for endoscopic treatment of esophageal neoplasia and precursor lesions: ESGE Position Statement – with competency assessment tools

    by Bas L.A.M. Weusten, Roos E. Pouw, Jacques J. Bergman, Raf Bisschops, Emmanuel Coron, Pieter Dewint, Peter Elbe, Vicente Lorenzo-Zúñiga, Marcus Hollenbach, Maximilien Barret, et al.

    Prior to the start of practical training, the trainee should be aware of different treatment modalities and their specific indications and limitations for patients with esophageal neoplasia and precursor lesions. Before starting the esophageal cancer treatment curriculum, proficiency in management of adverse events such as bleeding, perforation, and strictures, is mandatory. Competency in optical diagnosis of esophageal (pre)neoplastic lesions is a prerequisite to achieve proficiency in the endoscopic management of early esophageal neoplasia. Proficiency in both endoscopic resection and endoscopic ablation is mandatory for being considered competent in the endoscopic management of early Barrett neoplasia. Endoscopic resection in the esophagus should be learned through a structured training program consisting of lectures, live demonstrations, and hands-on training on models, followed by training in humans under strict expert supervision in a hands-on setting. Ideally, a minimum of 10 supervised cases should be done on ex vivo/in vivo models. In order to maintain competence in the endoscopic management of early esophageal neoplasia, an annual caseload of at least 10 new cases per endoscopist with esophageal endoscopic resection and 10 new cases per endoscopist with endoscopic ablation is required.

  • Open Access

    Risk of Recurrence and Adverse Events for Colorectal Endoscopic Resection: A Network Meta-Analysis of Randomized Controlled Trials

    by Roberto De Sire, Rodolfo Augusto Assis Rezende, Yuichi Mori, See Pdf

    Background and Aims: Multiple endoscopic resection techniques are available for large colorectal lesions, but comparative evidence remains limited. We performed a network meta-analysis of randomized controlled trials to define their efficacy-safety profiles.Methods: We systematically searched MEDLINE, Embase, and Scopus for RCTs comparing endoscopic mucosal resection (EMR), EMR with margin ablation, cold-EMR, underwater-EMR, endoscopic submucosal dissection, and endoscopic full-thickness resection for colorectal lesions > 10 mm. Primary outcomes were recurrence and adverse events. A frequentist random-effects network meta-analysis with benefit-risk analysis estimating numbers needed to treat (NNT) and harm (NNH) versus EMR was performed.Results: Eleven randomized controlled trials comprising 3,259 lesions (size 17-40mm) were included. Compared with EMR (recurrence 17%; adverse events 7%), endoscopic submucosal dissection ranked first for efficacy (recurrence 1%; relative risk 0.05), but last for safety (adverse events 24%; relative risk 1.66), reducing recurrence by 3% per each 1% increase in adverse events (NNT 6; NNH 19). Conversely, cold-EMR ranked first for safety (adverse events 3%, relative risk 0.37), but last for efficacy (recurrence 25%, relative risk 1.49), increasing recurrence by 2% per each 1% saving in adverse events (NNH 11; NNT 24). EMR with margin ablation showed intermediate outcomes (recurrence 6%; adverse events 7%), reducing recurrence versus standard EMR (relative risk 0.37; NNT 9) without increasing adverse events. Conclusions: Endoscopic resection techniques showed substantial trade-offs between efficacy and safety. EMR with margin ablation provided the most balanced efficacy–safety profile, whereas endoscopic submucosal dissection and cold-EMR represented opposite ends of the benefit–risk spectrum.

  • Computer-aided detection in surveillance colonoscopy: a population-based randomized trial

    by Laura Codesido-Prado, Pedro Davila-Piñón, Raquel Almazán, Luisa De Castro, Ignacio Couto-Worner, Rafael Mejuto, Alejandro Ledo, Gema Ramiro, Lidia González, Astrid I Díez-Martín, et al.

    Background and study aims: Computer-aided detection (CADe) increases adenoma detection in colonoscopy, but its effectiveness in post-polypectomy surveillance remains uncertain. This study compared CADe-assisted versus standard high-definition colonoscopy for adenoma detection rate (ADR) during surveillance. Patients and methods: We conducted a population-based, randomized controlled trial in the Galician colorectal cancer screening program. Colonoscopy days were randomized in a 1:1 ratio to high-definition colonoscopy with or without CADe. The primary endpoint was ADR. Prespecified subgroup analyses evaluated endoscopist baseline performance (median ADR in the standard arm). Effect estimates are reported as adjusted risk ratios (aRR) and confidence interval (CI). Results: Of 5,309 randomized surveillance colonoscopies, 4,824 surveillance colonoscopies were included in the final analysis (2,469 standard; 2,355 CADe-assisted) after exclusion of incomplete examinations. ADR did not differ between groups (57.4% vs 58.8%; aRR 1.02 [95% CI 0.95–1.10]). CADe increased ADR among low-performing (ADR<54.5%) endoscopists (45.5% vs 52.1%; aRR 1.15 [95% CI 1.01–1.30]), but not among high-performing endoscopists (65.9% vs 63.5%; aRR 0.96 [95% CI 0.88–1.05]). Conclusions: In a high-performing screening program, CADe did not increase ADR in surveillance colonoscopy. Its benefit may be limited to lower-performing endoscopists.


  • Open Access

    Learning and deskilling effects of artificial intelligence in colonoscopy among endoscopists with different levels of experience: a pragmatic, prospective trial

    by Tom Andre Pedersen, Yuichi Mori, Edoardo Botteri, Trond Engjom, Birgitte Seip, Georg Gjorgji Dimcevski, Roald Flesland Havre

    The effect of computer-aided detection (CADe) on performance of endoscopists with different experience levels is not well understood. This study assessed whether CADe use promotes learning or deskilling. We performed a prospective, multicenter, registry-based, pragmatic clinical trial. The primary end point was change in the proportion of colonoscopies with detection of at least one polyp ≥5 mm (PDR-5). Each endoscopist performed colonoscopies in three successive phases: 1) before CADe exposure, 2) during CADe use, and 3) after CADe removal. To assess the impact of CADe on PDR-5, we compared phase 1 with phases 2 and 3, adjusting for patient and endoscopist characteristics and endoscopy performance measures. Generalized linear mixed models were used and presented according to endoscopist experience. 13 endoscopists (7 inexperienced, 6 experienced) performed 5013 colonoscopies; median patient age was 59 years (interquartile range 44–71) and 2703 (53.9%) were women. Patient sex and age were similar across the two experience groups, whereas indications differed (e.g. more inflammatory bowel disease in the inexperienced group). CADe temporarily increased PDR-5 among inexperienced endoscopists from phase 1 to 2 (31.9% [216/678] to 39.5% [257/651]; odds ratio [OR] 1.43, 95%CI 1.11–1.84). There was no significant change between phases 1 and 3 (31.9% [216/678] to 36.3% [173/476]; OR 1.03, 95%CI 0.79–1.34). There was no significant CADe effect on PDR-5 among experienced endoscopists between phases. PDR-5 of inexperienced endoscopists increased when CADe was used. In non-CADe-assisted colonoscopy, no upskilling or deskilling was observed following a period of CADe exposure.


  • Reply to Lei et al.

    by Yuichi Mori, Uri Kopylov, Pieter Sinonquel, Alanna Ebigbo, Tony Tham

    10.1055/a-2776-5205 We thank Lei et al. [1] for their constructive correspondence regarding our recently published European Society of Gastrointestinal Endoscopy (ESGE) position statement. The goal of our statement is to facilitate the safe and effective integration of artificial intelligence (AI) tools into endoscopy training [2]. We appreciate the points raised, particularly concerning the importance of the timing of AI introduction during training and the need for bias-mitigation measures. In terms of timing, we fully agree that further research is required to determine the optimal timing for the introduction of AI in endoscopy training. The current lack of robust evidence limited our position statement to raising awareness of AI-related bias. Nevertheless, future revisions should expand guidance on when and how AI tools should be incorporated into training curricula, supported by stronger research-based evidence. We also agree with the insights regarding bias mitigation. The introduction of AI should be approached with a balanced perspective, rather than being dominated by concerns related to AI-driven bias, such as automation bias or deskilling [3]. Human progress has historically involved both gains and losses; for example, while calculators save time and enable focus on higher level tasks, they may also erode our calculation skills. Such trade-offs are usually regarded as a positive aspect of human evolution. Similarly, in medicine, the thoughtful integration of AI may allow clinicians to focus more on aspects of patient care that are of greater importance, representing a technological benefit. On the other hand, it remains essential to investigate strategies to mitigate bias and safeguard our life-critical clinical skills, such as resuscitation techniques. Potential approaches include simulation-based training, as highlighted by the authors, as well as psychological interventions addressing human behavior [4]. In conclusion, these issues are central to the safe integration of AI into endoscopy training and should be addressed in future evidence-based curricula. Article published online: 20 March 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany

  • Open Access

    Large Language Model‐Driven Analysis and Report Generation of Endoscopy Videos—A Pilot Study

    by Davide Massimi, Luca Di Stefano, Tommy Rizkala, Marco Spadaccini, Yuichi Mori, Maddalena Menini, Giulio Antonelli, Kareem Khalaf, Raf Bisschops, Daniel von Renteln, et al.

    Multimodal large language models (MLLMs) can automatically analyze clinical video, but evidence from full esophagogastroduodenoscopy (EGD) and the impact of on‐screen computer‐aided detection/diagnosis (CAD) overlays on MLLM behavior remain unclear. We tested whether an MLLM can produce clinically adequate EGD reports and whether a CAD overlay changes performance. We analyzed five complete EGD videos with Gemini 2.5 Pro in paired versions: (1) clean video and (2) the same video with a CAD overlay. Five blinded endoscopists rated report adequacy in three domains. MLLM accuracy for landmarks/lesions was further assessed by two blinded expert endoscopists using the time‐window rule (a model detection counted as correct if it occurred within ±2 s of the expert‐annotated timestamp). In this retrospective pilot study, five archived diagnostic EGD procedures from five patients were available as full‐length videos. Across five raters, MLLM Completeness was judged adequate in 56.0% (14/25 ratings) with Clean‐Video versus 48.0% (12/25 ratings) with Overlay‐Video ( p = 0.500). Visualization was identical (36.0% [9/25 ratings] for both; p = 1.000). Lesions characteristics were identical (16.0% [4/25] for both; p = 1.00). For the Landmark agreement, the overall accuracy of the MLLM with Clean‐Video vs. Overlay‐Video was: 0.55 [95% CI 0.43–0.67] vs. 0.33 [0.23–0.46], p = 0.029; sensitivity 0.53 [0.40–0.66] vs. 0.35 [0.24–0.49], p = 0.122; specificity 0.67 [0.35–0.88] vs. 0.22 [0.06–0.55], p = 0.125. In this pilot study, Gemini 2.5 Pro demonstrated inadequate performance for clinical EGD reporting. These hypothesis‐generating findings suggest substantial optimization and larger‐scale validation are required before deployment.


  • Preface

    by Yuichi Mori, Iris Lansdorp-Vogelaar