Comment on the clinical trial landscape of artificial intelligence applications in gastrointestinal endoscopy

Dear Editor,

We read with great interest the article by Li et al, which provides a comprehensive analysis of the clinical trial landscape of artificial intelligence (AI) applications in gastrointestinal (GI) endoscopy. This study offers significant insights into the rapid evolution of AI in this field, shedding light on its current applications, challenges, and future directions. As the authors note, AI has demonstrated remarkable potential in improving diagnostic accuracy, particularly in colorectal cancer detection. We fully support their assessment of the promising role of AI and would like to extend their discussion on several key aspects, particularly the implications for global clinical implementation and the integration of AI with emerging technologies.

The analysis of 124 trials, with a substantial proportion focused on colorectal adenoma detection, offers strong evidence for AI’s ability to enhance diagnostic precision in GI endoscopy. The marked growth in AI-related trials from 2018 to 2021, as highlighted by the authors, underscores the growing recognition of AI’s transformative potential in endoscopy. However, we believe it is crucial to expand the discussion on the need for global generalizability of AI algorithms. While the concentration of AI research in China is undoubtedly a driving force behind the rapid technological advancements in this area, there is a need for broader international collaboration to assess how these algorithms perform in diverse clinical settings, with different patient demographics, healthcare systems, and regulatory environments. As the authors correctly point out, single-center studies dominate the current landscape, and this could potentially limit the external validity of findings. Multi-center, international studies are needed to ensure that AI models can be universally applied and validated across varying healthcare systems and patient populations^[1^].

Moreover, it is important to consider that the generalizability of AI algorithms is not just a matter of geographic distribution but also of the technological diversity across clinical environments. Variations in device manufacturers, image quality standards, and operator expertise can all affect the performance of AI algorithms. For instance, recent studies have indicated that AI-assisted colonoscopy systems, although promising in controlled environments, show variability in performance when applied to heterogeneous clinical settings, due to differences in hardware, software, and procedural protocols^[2^]. Thus, the implementation of AI in GI endoscopy must be accompanied by rigorous testing in varied clinical contexts to ensure consistent performance and to address concerns related to algorithmic bias and variability.

In addition to geographic and clinical diversity, we believe that future AI applications in GI endoscopy should focus on integrating multi-modal data. While AI-driven image analysis, such as polyp detection, has shown significant promise, combining AI with other emerging technologies – such as biomarkers, patient history, and clinical data – could lead to more personalized and accurate diagnostic and therapeutic strategies. Multi-modal AI systems that incorporate not just endoscopic images, but also patient-specific genetic and clinical data, could revolutionize how we approach disease management, particularly in early-stage cancers and complex GI conditions. Several studies have already demonstrated the potential benefits of such integrated approaches in oncology, suggesting that this model could be adapted for GI endoscopy as well^[3,4^].

Furthermore, as AI systems in endoscopy advance, it will be crucial to establish standardized evaluation metrics for clinical validation. While the authors correctly highlight the need for rigorous trials, we would like to emphasize the importance of including comprehensive cost-effectiveness analyses in future studies. As AI algorithms transition from research to clinical use, assessing the economic feasibility of implementing these technologies on a broad scale will be essential. This includes evaluating the costs associated with AI systems, the potential savings from earlier and more accurate diagnoses, and the overall impact on patient outcomes. Studies that address these factors will be pivotal in facilitating the widespread adoption of AI in clinical practice^[5^].

Finally, we commend the authors for their focus on the regulatory challenges surrounding AI in GI endoscopy. As the authors suggest, the rapid evolution of AI technology necessitates regulatory frameworks that can keep pace with innovation while ensuring patient safety. Harmonization of regulatory standards, particularly at the international level, will be crucial for the successful global deployment of AI-powered endoscopic systems. The approval of AI-driven devices in different regions must align with stringent safety and efficacy criteria, which will require ongoing collaboration between regulatory bodies, industry stakeholders, and healthcare providers^[6^].

In conclusion, we support the authors’ findings and call for further international collaboration, multi-center trials, and the integration of multi-modal data to enhance the clinical utility of AI in GI endoscopy. As the field continues to evolve, it will be critical to focus on the real-world implementation of AI technologies, ensuring that these advancements are not only scientifically robust but also accessible and beneficial to patients worldwide.

Ethical approval

Not applicable.

Consent

Not applicable.

Sources of funding

This work was supported by the Clinical Specialized Discipline of Health System of PuTuo District in Shanghai (Grant Number: 2021tszk01), Putuo District Central Hospital Project (No.2023-BSH-02).

Author contributions

The three authors contributed to the development and organization of this letter.

Conflicts of interest disclosure

The authors declare no competing interests.

Research registration unique identifying number (UIN)

Not applicable.

Guarantor

The first author assumes complete responsibility for the integrity of the work, from its inception to the final published article.

Provenance and peer review

Not commissioned.

Data availability statement

No data were used in this Letter to the Editor.

Generative AI statement

The authors declare that no generative artificial intelligence (AI) was used in the creation of this manuscript. No AI tools were used for data generation, analysis, or conclusions. The authors verified all content. This manuscript follows the TITAN Guidelines 2025.