The advent of artificial intelligence (AI) in medical diagnostics has markedly transformed the landscape of gastroenterology (1–4). AI technologies, encompassing machine learning and deep learning algorithms, have been increasingly integrated into these procedures to enhance the detection and characterization of gastrointestinal lesions (5–9). These innovations aim to augment diagnostic accuracy, facilitate the early identification of abnormalities such as polyps and neoplasms, and ultimately improve patient outcomes (10–14).

Endoscopic and colonoscopic examinations are pivotal for the diagnosis, surveillance, and management of gastrointestinal diseases (15–17). Despite their clinical importance, these procedures are inherently operator-dependent, leading to variability in detection rates and diagnostic accuracy (18, 19). The incorporation of AI serves to mitigate these limitations by providing real-time image analysis and decision support, thereby standardizing examinations and reducing the incidence of missed lesions (20, 21). Furthermore, AI applications have the potential to optimize procedural efficiency and reduce the cognitive load on clinicians (22).

Bibliometric studies offer a systematic approach to quantitatively evaluating scientific publications, enabling the assessment of research performance, identification of influential contributions, and mapping of collaborative networks within a specific domain (23–25). By employing bibliometric methods alongside visualization tools, complex datasets can be transformed into interpretable visual representations, facilitating a deeper understanding of research trends and knowledge structures. A comprehensive bibliometric analysis of AI applications in endoscopy and colonoscopy is particularly valuable because of the rapid expansion, interdisciplinary nature, and global interest surrounding AI-driven medical technologies. Such analysis systematically quantifies and visually maps research contributions, collaboration patterns, influential studies, and evolving thematic trends, which traditional literature reviews might not fully capture. By objectively analyzing publication trends, author networks, and thematic clusters, bibliometric studies can identify knowledge gaps, highlight emerging research directions, and provide researchers, clinicians, and policymakers with data-driven insights necessary for guiding future research investments, policy decisions, and clinical practice advancements in the rapidly evolving field of AI-enhanced gastroenterology. The objective of this study is to conduct a thorough bibliometric and visualization analysis of global research pertaining to AI applications in endoscopy and colonoscopy.

AI has experienced significant advancements since its inception, resulting in its integration into various medical disciplines. In endoscopy and colonoscopy, the adoption of AI began in the early 2000s, primarily to enhance diagnostic accuracy and operational efficiency (52, 53). Bibliometric analysis offers a systematic and quantitative approach to evaluating research trends by analyzing key metrics such as authorship, institutional contributions, country affiliations, and citation patterns. Unlike traditional systematic reviews, bibliometric methods, when combined with tools like CiteSpace and VOSviewer, provide a more rigorous and data-driven overview, offering visually interpretable insights into the evolving scientific landscape. This study represents the first comprehensive bibliometric analysis of AI applications in endoscopy and colonoscopy, covering the period from 2002 to 2023, with the aim of mapping the research trajectory and identifying emerging trends that could influence future developments in the field.

The publication trend analysis demonstrates a clear shift toward a growing academic focus on AI in endoscopy and colonoscopy. Initially, research activity in this area was limited, but over recent years, a notable increase in publications has emerged, signaling a substantial rise in interest. This trend reflects the growing momentum in the field, with more researchers contributing to the development and exploration of AI applications. The recent rise in cumulative publications further emphasizes this increasing engagement, suggesting that the field is gaining traction and becoming a prominent focus of scientific inquiry. The slight dip in recent publication numbers may reflect ongoing indexing processes rather than a reduction in research activity, highlighting the field's continued expansion.

The geographical distribution of research highlights the dominance of certain countries and institutions in driving innovation. An analysis of the countries and institutions contributing to research on AI in endoscopy and colonoscopy reveals that China holds the first rank in terms of publications, followed by the United States in second place and Japan in third place. In terms of institutional output, Showa University takes the top spot, followed by Humanitas University in second place and the University of Oslo in third place, reflecting their major contributions to advancing AI-driven diagnostic technologies.

Regarding centrality, which measures the influence and connectedness of a country or institution within the global research network, the United States ranks first, followed by England in second place and France in third place. Among institutions, the Chinese University of Hong Kong claims the top spot, with Assistance Publique Hopitaux Paris in second place and Korea University in third place, signifying their central roles in driving international collaboration and innovation in this field. Centrality highlights how pivotal these countries and institutions are in shaping global research directions and fostering important partnerships. The higher number of publications produced by China can be attributed to substantial government investment in AI research, rapid technological advancements, and significant growth in research institutions and researchers over recent years. In contrast, the leading role of the United States regarding collaboration centrality could stem from its well-established infrastructure for international research cooperation, active engagement in cross-country research initiatives, dedicated funding opportunities specifically targeting international collaboration, and a longstanding culture of collaborative scientific research. The higher number of publications produced by China can be attributed to substantial government investment in AI research, rapid technological advancements, and significant growth in research institutions and researchers over recent years. In contrast, the leading role of the United States regarding collaboration centrality could stem from its well-established infrastructure for international research cooperation, active engagement in cross-country research initiatives, dedicated funding opportunities specifically targeting international collaboration, and a longstanding culture of collaborative scientific research. Our bibliometric visualization analysis also highlighted thicker connecting lines between the United States and other leading countries, such as Japan, Germany, and England. These visually prominent connections underscore stronger international collaborative ties, further reinforcing the influential role of U.S. research outputs in shaping global research agendas and advancing the field of AI applications in endoscopy and colonoscopy.

The impacts of a higher number of publications include accelerated innovation, rapid dissemination of new AI technologies in clinical practice, and increased visibility of research institutions and individual researchers internationally (54–56). High publication output can facilitate faster knowledge accumulation, attracting further research funding and partnerships. Conversely, stronger global collaborations have distinct advantages. Robust international research collaborations often lead to higher-quality, multidisciplinary studies that benefit from diverse perspectives and expertise, thus enhancing scientific rigor and generalizability of results (57, 58). Collaborations can also expedite knowledge transfer, standardization of methodologies, and implementation of best practices across international boundaries, contributing significantly to shaping global research priorities and clinical guidelines in AI applications in endoscopy and colonoscopy (59–61). Our findings suggest that an ideal research environment would balance both high publication productivity and strong international collaboration, thereby maximizing both innovation and the practical impact of research outcomes.

The assessment of the most frequently cited references has revealed several key studies that have significantly contributed to the advancement of AI applications in endoscopy and colonoscopy. These cited papers are crucial in highlighting various aspects of AI which providing valuable insights into the development, clinical integration, and impact of AI technologies in gastrointestinal procedures.

The most cited study by Wang et al. (62) in Gut 2019, titled “Real-time automatic detection system increases colonoscopic polyp and adenoma detection rates: a prospective randomized controlled study,” explored the efficacy of an AI-based system for detecting polyps during colonoscopy. This randomized controlled trial found that the use of AI significantly improved adenoma detection rates (ADR) by 50%, increasing ADR from 20% to 30%. The study demonstrated that AI particularly aided in identifying smaller adenomas and hyperplastic polyps, contributing to enhanced diagnostic precision in real-time clinical settings. These findings underscore the potential of AI in improving colonoscopic outcomes, especially in detecting diminutive lesions.

The second most cited study by Hirasawa et al. (46) in Gastric Cancer 2018, titled “Application of artificial intelligence using a convolutional neural network (CNN) for detecting gastric cancer in endoscopic images,” focused on developing a CNN to automatically detect gastric cancer from endoscopic images. The study trained the AI model on over 13,000 images and tested it on a separate dataset. The CNN achieved a sensitivity of 92.2% in detecting gastric cancer lesions, particularly excelling in identifying larger lesions (≥6 mm) and invasive cancers. However, it missed smaller, superficially depressed lesions that were difficult even for experienced endoscopists to identify. Despite some false positives, the CNN demonstrated a strong potential for real-time clinical applications, reducing the diagnostic burden on endoscopists and improving detection rates for gastric cancer.

The third most cited study by Urban et al. (47) in Gastroenterology 2018, titled “Deep Learning Localizes and Identifies Polyps in Real Time with 96% Accuracy in Screening Colonoscopy,” developed a CNN to detect and localize polyps during colonoscopy in real-time. The CNN achieved a 96.4% accuracy and an area under the curve (AUC) of 0.991, identifying polyps with a high degree of precision. It was particularly effective in detecting polyps in real-time scenarios, reducing the risk of missed polyps during screenings. This study demonstrated the significant potential of AI-assisted systems to increase adenoma detection rates (ADR), which are critical in colorectal cancer prevention.

This fourth most cited study by Byrne et al. (48) in Gut 2019, titled “Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model,” developed a deep convolutional neural network (DCNN) to differentiate between adenomatous and hyperplastic diminutive polyps during colonoscopy. The AI model achieved an overall accuracy of 94%, with a sensitivity of 98% for identifying adenomas and a specificity of 83% for hyperplastic polyps. These results suggest that AI can provide real-time support in differentiating diminutive polyps, potentially facilitating the 'resect and discard' strategy, which could streamline clinical decision-making during colonoscopy and reduce unnecessary pathology evaluations.

This fifth most cited study by Mori et al. (63) in Annals of Internal Medicine 2018, titled “Real-Time Use of Artificial Intelligence in Identification of Diminutive Polyps During Colonoscopy: A Prospective Study,” focused on evaluating the performance of a computer-aided diagnosis (CAD) system during colonoscopy. The study, conducted on 791 patients, demonstrated that the CAD system achieved a negative predictive value (NPV) of 96.4% for diminutive rectosigmoid adenomas. This high NPV meets the clinical threshold required for a “diagnose-and-leave” strategy, reducing the need for unnecessary polyp resections. The findings suggest that CAD systems can effectively assist in real-time diagnosis of diminutive polyps, potentially improving the cost-effectiveness and efficiency of colonoscopy procedures.

The overall findings of these highly cited article indicate that research in this field has increasingly focused on using AI technologies to enhance the accuracy, efficiency, and real-time diagnostic capabilities in endoscopy and colonoscopy procedures. These studies highlight AI's role in improving the detection of malignancies, polyps, classification of diminutive polyps, and ultimately, the early identification of cancers such as colorectal and gastric cancer. Continuous advancements in AI-based diagnostic tools, including computer-aided diagnosis (CAD) systems and deep learning models, show a clear trend toward integrating AI into clinical practice. The goal is to reduce missed lesions, improve detection rates, and ultimately optimize patient outcomes.

An in-depth evaluation of key influential studies exemplifies the growing research emphasis on integrating artificial intelligence AI technologies to improve the diagnostic performance of colonoscopy. These studies have not only demonstrated the clinical value of AI-assisted approaches, but have also served as foundational milestones in shaping methodological standards and advancing the translational trajectory of AI in gastrointestinal endoscopy. Their collective contributions reflect a shift from early technical feasibility toward clinically meaningful, scalable, and evidence-based applications.

One of the most comprehensive and methodologically impactful studies in this field is the meta-analysis by Hassan et al., which pooled data from five randomized controlled trials with over 4,300 patients (49). This analysis provided robust evidence that AI-supported colonoscopy significantly improves ADR and adenomas per colonoscopy (APC), with consistent benefits across lesion size, morphology, and location. Importantly, it highlighted that AI performance remains effective regardless of lesion characteristics typically associated with higher miss rates, such as flat morphology, diminutive size, and proximal location. By emphasizing AI's independence from conventional perceptual constraints, this study underscored the capability of algorithm-driven systems to standardize detection quality and reduce observer variability in routine clinical practice.

Similarly, Wang et al. conducted one of the first prospective tandem colonoscopy trials to directly measure AI's impact on adenoma miss rate (AMR) (50). Their study demonstrated a substantial reduction in AMR—from 40.0% in standard procedures to 13.9% with AI support—particularly in the proximal colon. Beyond its clinical findings, this study introduced a rigorous methodological framework for evaluating recognition-related errors and positioned AI as a real-time adjunct capable of supporting endoscopists across varying levels of expertise. It also contributed significantly to the understanding of how AI can mitigate human cognitive limitations without disrupting procedural flow.

Repici et al. expanded the evidence base by implementing a multicenter randomized trial in a Western population, confirming AI's generalizability and feasibility in diverse clinical settings (51). Their findings demonstrated a marked increase in ADR (54.8% vs. 40.4%) and improved APC, with no significant difference in withdrawal time or unnecessary resections. The strength of this study lies in its pragmatic design and its demonstration that AI-enhanced colonoscopy can be seamlessly incorporated into existing endoscopic workflows. In particular, its effectiveness in detecting small and non-polypoid lesions further validates the role of AI in elevating detection quality in real-world practice.

Together, these milestone studies have shaped the evidence landscape by demonstrating not only the diagnostic advantages of AI, but also the importance of robust study designs, interdisciplinary collaboration, and clinical relevance in advancing the field. They collectively represent a transition point from proof-of-concept technologies to validated, practice-ready tools.

The analysis of publication trends underscores the prominent role of a few key journals in advancing AI research in endoscopy and colonoscopy. Gastrointestinal Endoscopy leads as the primary hub for cutting-edge AI research, emphasizing its crucial role in integrating AI into clinical practice. Diagnostics follows, reflecting the growing focus on AI's impact on diagnostic improvements, while Digestive Endoscopy, ranked third, highlights its importance in AI-driven real-time imaging and lesion detection. In terms of influence, Gastrointestinal Endoscopy also holds the top spot for citations, with foundational AI studies frequently referenced. Endoscopy and Gastroenterology follow closely, reinforcing their roles in promoting high-impact AI research and accelerating the adoption of AI technologies in gastroenterology.

In the keyword analysis of AI research in endoscopy and colonoscopy, diagnosis emerged as the most frequently mentioned term, reflecting the field's strong focus on enhancing the early detection of gastrointestinal lesions, particularly polyps and tumors. AI technologies significantly improve diagnostic accuracy, leading to better patient outcomes by facilitating timely interventions (64–66). Cancer, is another critical area of focus due to its global prevalence. AI plays a pivotal role in early detection during colonoscopy, reducing missed diagnoses, and improving overall screening efficiency (67–69). The frequent mention of classification further highlights AI's crucial role in distinguishing between benign and malignant lesions. AI-powered classification systems, driven by machine learning and deep learning algorithms, greatly enhance the precision and reliability of diagnostic assessments in clinical practice (70–72).

In the analysis of central keywords, particular emphasis is placed on diagnosis, which is further strengthened by AI-driven classification systems (73, 74). Innovations such as capsule endoscopy enable non-invasive, precise detection of gastrointestinal abnormalities, allowing AI to perform real-time classification of lesions (75–77). Keywords related to cancer and malignant conditions, such as colorectal cancer, colorectal lesions, adenoma detection rate, and colonic polyps, were frequently identified in the analysis. These terms highlight AI's critical role in detecting and classifying high-risk lesions (78, 79). AI technologies are especially effective in distinguishing malignant from benign polyps, improving the accuracy of early diagnosis and supporting clinicians in making more informed treatment decisions (80, 81). Although our analysis included 1,571 studies, the explicit keyword “Artificial Intelligence” appeared with a frequency of 158 occurrences (approximately 4%). It is crucial to emphasize that this frequency reflects only explicit occurrences of the term “Artificial Intelligence” itself, rather than indicating the number of studies using AI. The analyzed literature included numerous related keywords, some directly describing AI techniques (e.g., “neural network,” “classification,” “machine learning”), others describing AI's diagnostic applications (e.g., “diagnosis,” “validation”), and others specifically focused on gastrointestinal terminology (e.g., “endoscopy,” “lesion,” “cancer”). This diversity in terminology explains the relatively lower frequency of the specific phrase “Artificial Intelligence,” despite all analyzed studies being directly relevant to AI applications in endoscopy and colonoscopy.

The cluster analysis highlights four key research areas where AI is making significant contributions in endoscopy and colonoscopy. Gastric cancer emerges as a major focus, with AI technologies enhancing early detection and diagnosis, which is crucial for improving treatment outcomes (82–84). Polyp segmentation is another critical area, where AI automates the identification and delineation of polyps, helping to standardize detection and reduce missed polyps during colonoscopy, a vital step in preventing colorectal cancer (85–87). In wireless capsule endoscopy, AI optimizes the analysis of vast image datasets, improving diagnostic accuracy while reducing the workload for clinicians (88). Lastly, adenoma detection benefits from AI's ability to increase detection rates of these precancerous polyps, significantly enhancing screening outcomes and reducing the risk of colorectal cancer (89–91). Similar to previous results, these areas emphasize the impact of AI on diagnosis, cancer detection, and AI-assisted classification in various types of malignancies.

The time trend analysis of clusters reveals that research in adenoma detection, polyp segmentation, wireless capsule endoscopy, and gastric cancer has gained significant momentum in recent years. This growing focus highlights the increasing importance of AI in enhancing diagnostic precision and early detection in gastrointestinal procedures. The rising trend in adenoma detection reflects AI's role in improving the identification of precancerous polyps, a key factor in preventing colorectal cancer (92–94). Polyp segmentation has also seen increased attention, as AI-driven models enhance the accuracy of polyp detection and reduce variability across clinicians (95, 96). Similarly, wireless capsule endoscopy is benefiting from AI's ability to analyze large image datasets efficiently, contributing to non-invasive diagnostic improvements (97, 98). Lastly, the surge in research on gastric cancer demonstrates the field's focus on leveraging AI for early diagnosis, which is critical for improving survival rates (99–101). These trends indicate a broad recognition of AI's potential to revolutionize gastrointestinal diagnostics, particularly in detecting and classifying various types of malignancies.

While AI technologies offer transformative potential for enhancing endoscopy and colonoscopy, their global clinical implementation faces challenges and ethical considerations (102, 103). A critical challenge involves standardizing AI algorithms across diverse populations, medical practices, and healthcare settings (104). Differences in patient demographics, disease prevalence, imaging equipment, and clinical workflows can lead to performance variability of AI models when applied outside their training environments (105–108). Ensuring AI generalizability requires extensive validation in multicentric and diverse cohorts worldwide.

Furthermore, ethical concerns such as algorithm transparency, accountability, patient privacy, and data security pose considerable obstacles (109–112). The “black-box” nature of AI algorithms raises important questions regarding transparency, potentially leading to mistrust among clinicians and patients (113–115). Establishing accountability for AI-driven clinical decisions remains complex and currently lacks clear international standards or regulatory guidance.

Additionally, the widespread clinical integration of AI raises potential issues of equity and fairness (116, 117). Disparities in access to advanced AI technologies, especially in low- and middle-income countries, could exacerbate existing global healthcare inequalities (118–120). Addressing these concerns requires international collaborative efforts to develop clear ethical frameworks, rigorous clinical validation standards, transparent reporting guidelines, and equitable technology dissemination strategies.

While this study offers a thorough bibliometric analysis of AI applications in endoscopy and colonoscopy, several limitations should be noted. First, the analysis was confined to articles indexed in the Web of Science Core Collection. Although this is a comprehensive resource, it may not cover relevant research from other databases. Additionally, only English-language publications were considered, which may have overlooked significant research in other languages, introducing a geographic bias that underrepresents contributions from non-English-speaking regions. Future research could leverage large language models (LLMs) to incorporate and examine non-English literature. Employing LLMs would enhance the depth of bibliometric studies, enabling the identification of nascent research clusters and offering a fuller representation of worldwide contributions to AI in endoscopy and colonoscopy.

Another limitation concerns citation bias. Bibliometric analysis heavily relies on citation counts, which can be affected by factors such as the reputation of journals, the timing of publication, and practices like self-citation. Therefore, citation-based metrics may not entirely capture the true influence of specific papers or authors. Moreover, given the rapidly advancing nature of AI research, newer studies may not have gathered substantial citations yet, potentially underrepresenting emerging research trends. Another important consideration, not extensively covered in our current analysis, involves the practical application and limitations of AI technologies in clinical practice worldwide or on a country-specific basis. Although our study provided a comprehensive overview of global research trends, future analyses could specifically examine variations in clinical adoption, practical challenges, regulatory environments, ethical considerations, cost-effectiveness, and healthcare infrastructure that influence how AI is implemented differently across countries. Such studies would be invaluable for understanding the real-world impact, identifying barriers to clinical integration, and guiding targeted policy decisions and resource allocation for broader and more equitable implementation of AI-driven endoscopy and colonoscopy practices globally. Finally, our analysis only includes papers published up to September 2024, meaning further developments in the field may not be reflected. The reported slight decline in publications observed in 2024 should be interpreted cautiously. Since the dataset was analyzed up to September 2024, the data for 2024 represents only a partial year, potentially leading readers to inaccurately conclude that research activity in AI applications for endoscopy and colonoscopy is decreasing. Therefore, this trend may simply reflect incomplete data rather than an actual reduction in research output.

This bibliometric analysis provides insights into the rapidly growing field of AI applications in endoscopy and colonoscopy. The analysis highlights key contributors, such as China and the United States, with the latter leading in global research collaborations. Institutions like Showa University and influential authors such as Mori Y have emerged as pivotal in advancing the field. Leading journals, particularly Gastrointestinal Endoscopy, play a critical role in disseminating cutting-edge research.

The cluster analysis revealed emerging trends focused on critical areas such as adenoma detection, polyp segmentation, and wireless capsule endoscopy, reflecting the increasing clinical emphasis on enhancing real-time diagnostic capabilities. The most frequently used keywords—“diagnosis,” “classification,” and “cancer”—underscore the central role of AI in improving the accuracy and efficiency of endoscopic procedures. As AI technologies become more integrated into clinical practice, they are anticipated to significantly improve diagnostic accuracy by providing real-time image analysis, reducing inter-operator variability, and enhancing the detection rates of critical gastrointestinal conditions, such as colorectal cancer and adenomas. These advancements are expected to lead to earlier and more precise diagnoses, reduce missed lesions, and ultimately improve patient outcomes by facilitating timely and effective treatment interventions in gastroenterology.