Indexed In : Science Citation Index Expanded(SCIE), MEDLINE,
Pubmed/Pubmed Central, Elsevier Bibliographic, Google Scholar,
Databases(Scopus & Embase), KCI, KoreaMed, DOAJ
Gut and Liver is an international journal of gastroenterology, focusing on the gastrointestinal tract, liver, biliary tree, pancreas, motility, and neurogastroenterology. Gut atnd Liver delivers up-to-date, authoritative papers on both clinical and research-based topics in gastroenterology. The Journal publishes original articles, case reports, brief communications, letters to the editor and invited review articles in the field of gastroenterology. The Journal is operated by internationally renowned editorial boards and designed to provide a global opportunity to promote academic developments in the field of gastroenterology and hepatology. +MORE
Yong Chan Lee |
Professor of Medicine Director, Gastrointestinal Research Laboratory Veterans Affairs Medical Center, Univ. California San Francisco San Francisco, USA |
Jong Pil Im | Seoul National University College of Medicine, Seoul, Korea |
Robert S. Bresalier | University of Texas M. D. Anderson Cancer Center, Houston, USA |
Steven H. Itzkowitz | Mount Sinai Medical Center, NY, USA |
All papers submitted to Gut and Liver are reviewed by the editorial team before being sent out for an external peer review to rule out papers that have low priority, insufficient originality, scientific flaws, or the absence of a message of importance to the readers of the Journal. A decision about these papers will usually be made within two or three weeks.
The remaining articles are usually sent to two reviewers. It would be very helpful if you could suggest a selection of reviewers and include their contact details. We may not always use the reviewers you recommend, but suggesting reviewers will make our reviewer database much richer; in the end, everyone will benefit. We reserve the right to return manuscripts in which no reviewers are suggested.
The final responsibility for the decision to accept or reject lies with the editors. In many cases, papers may be rejected despite favorable reviews because of editorial policy or a lack of space. The editor retains the right to determine publication priorities, the style of the paper, and to request, if necessary, that the material submitted be shortened for publication.
Correspondence to: Jung Ho Bae
ORCID https://orcid.org/0000-0001-7669-1213
E-mail bjh@snuh.org
See “Artificial Intelligence Models May Aid in Predicting Lymph Node Metastasis in Patients with T1 Colorectal Cancer” by Ji Eun Baek, et al. on page 69, Vol. 19, No. 1, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Gut Liver 2025;19(1):3-5. https://doi.org/10.5009/gnl250001
Published online January 15, 2025, Published date January 15, 2025
Copyright © Gut and Liver.
The landscape of colorectal cancer (CRC) management has evolved significantly with improved screening programs and advanced endoscopic techniques.1 While these developments have led to increased detection of early-stage CRC, they have also highlighted a critical challenge: accurate prediction of lymph node metastasis (LNM) in T1 CRC. The current paradigm, based on established guidelines such as those from the Japanese Society for Cancer of the Colon and Rectum, faces a substantial overtreatment issue, with 85% to 90% of high-risk patients showing no LNM in surgical specimens after additional surgery.2,3
Artificial intelligence (AI) has emerged as a promising solution to this clinical challenge. Recent systematic reviews indicate that more than 10 studies on AI-based LNM prediction in early CRC have been published, primarily conducted in Japan and Korea since 2018.4,5 These studies have retrospective design and employed a variety of AI models, ranging from machine learning (ML) algorithms such as random forest and support vector machine to deep learning algorithms like convolutional neural networks. Commonly reported performance metrics include sensitivity, specificity, accuracy, and area under the curve.
Broadly, these studies can be categorized into two main approaches: computer vision-based analysis of whole slide images (WSIs) and ML-based analysis of clinicopathological features.4 The utilization of digital pathology through advanced deep learning architectures allows for the extraction of intricate spatial and textural features from WSIs, which may not be discernible through traditional histological evaluation. On the other hand, ML-based models leverage clinical and pathological variables retrieved from electronic medical records, such as patient demographics, tumor location, size, and histopathological characteristics, to provide a comprehensive and integrative risk assessment framework.
The first AI model for predicting LNM in T1 CRC, developed by Ichimasa et al.,6 utilized a support vector machine algorithm. This model, which incorporated 45 clinicopathological factors, achieved a sensitivity of 100%, matching existing guidelines, while significantly improving specificity to 66% compared to 44%, 0%, and 0% for the American, European, and Japanese guidelines, respectively. As a result, the AI model substantially reduced the rate of unnecessary additional surgeries compared to traditional approaches.
Building on this foundation, recent work by Baek et al.7 represents a significant advancement in the field by employing ML models that integrate both endoscopic and pathological features for LNM prediction. This retrospective study analyzed data from 1,386 patients who underwent surgical resection of T1 CRC, with or without prior endoscopic resection, between 2010 and 2018. The authors incorporated four expert-assigned endoscopic characteristics, including hardness, white spots, ulceration/depression, and expansion, alongside four adverse pathological features, including tumor histology, lymphovascular invasion, depth of invasion, and tumor budding, to train four ML algorithms.
A notable strength of the ML approach for clinicopathological factors lies in its interpretability. Baek et al.7 utilized SHAP (SHapley Additive exPlanations) plots to illustrate the impact of individual variables on LNM prediction, enabling physicians to gain valuable insights into AI-derived conclusions. These models demonstrated improved risk assessment for LNM compared to the Japanese Society for Cancer of the Colon and Rectum guidelines, offering a more nuanced and accurate approach to clinical decision-making. This integration of diverse clinical data highlights the growing potential of AI to transform T1 CRC management by improving diagnostic accuracy and reducing unnecessary surgical interventions.
Despite these advancements, current ML models by Baek et al.7 require further refinement. The best-performing model’s area under the receiver operating characteristic curve (AUROC) of 0.679 remains modest compared to other AI-based approaches, which have achieved AUROC values exceeding 0.72 to 1.00.5 Performance was particularly limited in the subgroup of patients who underwent initial endoscopic resection, with an AUROC of only 0.582, highlighting a critical area for improvement.
While the inclusion of both endoscopic and pathological features is a strength, the study is hindered by persistent issues related to interobserver variability in the assessment of histopathologic risk factors.4,8-10 Endoscopic findings, which are assessed by individual experts, can be subjective and prone to inconsistencies. Features such as “hardness” and “expansion” derived from the review of endoscopic images, lack standardized definitions, further exacerbating the potential for interobserver bias.7 This subjectivity not only limits the reliability of the data but also poses challenges for the reproducibility and scalability of the models in diverse clinical settings.
To address these limitations, future efforts should prioritize enhancing model performance by integrating deep learning architectures capable of directly analyzing endoscopic images alongside WSI analysis. Combining these approaches with ML-based analysis of clinicopathological features in a hybrid model could leverage the strengths of each methodology, enabling a more comprehensive and accurate prediction system. This hybrid approach would minimize observer bias by automating the analysis of visual data while incorporating detailed clinical and pathological variables to provide a holistic risk assessment framework.
Expanding datasets through multicenter collaborations and balancing the dataset by increasing the representation of LNM-positive cases through data augmentation or optimized hyperparameter tuning are critical.10 Furthermore, external validation with independent data is essential to improve model robustness and generalizability. These refinements will be pivotal in translating the promising findings of Baek et al. into clinically viable tools for accurate LNM prediction in T1 CRC, ultimately reducing unnecessary surgeries.
In conclusion, AI represents a transformative advancement in predicting LNM in T1 CRC, encompassing ML models that leverage clinicopathological data and innovative approaches utilizing pathological image analysis. These technologies have shown significant potential to enhance diagnostic accuracy, reduce unnecessary surgeries, and improve patient outcomes. However, their adoption also raises ethical considerations, particularly regarding critical decision-making based on AI-derived predictions. Additionally, the integration of AI models into hospital systems necessitates careful planning, including interoperability with existing electronic medical records, clinician training, and continuous performance monitoring. Addressing these challenges through collaborative efforts involving clinicians, data scientists, and policymakers will be crucial to ensure the seamless incorporation of AI-driven tools into clinical workflows without compromising patient safety or care quality.
No potential conflict of interest relevant to this article was reported.
Gut and Liver 2025; 19(1): 3-5
Published online January 15, 2025 https://doi.org/10.5009/gnl250001
Copyright © Gut and Liver.
Department of Internal Medicine and Healthcare Research Institute, Healthcare System Gangnam Center, Seoul National University Hospital, Seoul, Korea
Correspondence to:Jung Ho Bae
ORCID https://orcid.org/0000-0001-7669-1213
E-mail bjh@snuh.org
See “Artificial Intelligence Models May Aid in Predicting Lymph Node Metastasis in Patients with T1 Colorectal Cancer” by Ji Eun Baek, et al. on page 69, Vol. 19, No. 1, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
The landscape of colorectal cancer (CRC) management has evolved significantly with improved screening programs and advanced endoscopic techniques.1 While these developments have led to increased detection of early-stage CRC, they have also highlighted a critical challenge: accurate prediction of lymph node metastasis (LNM) in T1 CRC. The current paradigm, based on established guidelines such as those from the Japanese Society for Cancer of the Colon and Rectum, faces a substantial overtreatment issue, with 85% to 90% of high-risk patients showing no LNM in surgical specimens after additional surgery.2,3
Artificial intelligence (AI) has emerged as a promising solution to this clinical challenge. Recent systematic reviews indicate that more than 10 studies on AI-based LNM prediction in early CRC have been published, primarily conducted in Japan and Korea since 2018.4,5 These studies have retrospective design and employed a variety of AI models, ranging from machine learning (ML) algorithms such as random forest and support vector machine to deep learning algorithms like convolutional neural networks. Commonly reported performance metrics include sensitivity, specificity, accuracy, and area under the curve.
Broadly, these studies can be categorized into two main approaches: computer vision-based analysis of whole slide images (WSIs) and ML-based analysis of clinicopathological features.4 The utilization of digital pathology through advanced deep learning architectures allows for the extraction of intricate spatial and textural features from WSIs, which may not be discernible through traditional histological evaluation. On the other hand, ML-based models leverage clinical and pathological variables retrieved from electronic medical records, such as patient demographics, tumor location, size, and histopathological characteristics, to provide a comprehensive and integrative risk assessment framework.
The first AI model for predicting LNM in T1 CRC, developed by Ichimasa et al.,6 utilized a support vector machine algorithm. This model, which incorporated 45 clinicopathological factors, achieved a sensitivity of 100%, matching existing guidelines, while significantly improving specificity to 66% compared to 44%, 0%, and 0% for the American, European, and Japanese guidelines, respectively. As a result, the AI model substantially reduced the rate of unnecessary additional surgeries compared to traditional approaches.
Building on this foundation, recent work by Baek et al.7 represents a significant advancement in the field by employing ML models that integrate both endoscopic and pathological features for LNM prediction. This retrospective study analyzed data from 1,386 patients who underwent surgical resection of T1 CRC, with or without prior endoscopic resection, between 2010 and 2018. The authors incorporated four expert-assigned endoscopic characteristics, including hardness, white spots, ulceration/depression, and expansion, alongside four adverse pathological features, including tumor histology, lymphovascular invasion, depth of invasion, and tumor budding, to train four ML algorithms.
A notable strength of the ML approach for clinicopathological factors lies in its interpretability. Baek et al.7 utilized SHAP (SHapley Additive exPlanations) plots to illustrate the impact of individual variables on LNM prediction, enabling physicians to gain valuable insights into AI-derived conclusions. These models demonstrated improved risk assessment for LNM compared to the Japanese Society for Cancer of the Colon and Rectum guidelines, offering a more nuanced and accurate approach to clinical decision-making. This integration of diverse clinical data highlights the growing potential of AI to transform T1 CRC management by improving diagnostic accuracy and reducing unnecessary surgical interventions.
Despite these advancements, current ML models by Baek et al.7 require further refinement. The best-performing model’s area under the receiver operating characteristic curve (AUROC) of 0.679 remains modest compared to other AI-based approaches, which have achieved AUROC values exceeding 0.72 to 1.00.5 Performance was particularly limited in the subgroup of patients who underwent initial endoscopic resection, with an AUROC of only 0.582, highlighting a critical area for improvement.
While the inclusion of both endoscopic and pathological features is a strength, the study is hindered by persistent issues related to interobserver variability in the assessment of histopathologic risk factors.4,8-10 Endoscopic findings, which are assessed by individual experts, can be subjective and prone to inconsistencies. Features such as “hardness” and “expansion” derived from the review of endoscopic images, lack standardized definitions, further exacerbating the potential for interobserver bias.7 This subjectivity not only limits the reliability of the data but also poses challenges for the reproducibility and scalability of the models in diverse clinical settings.
To address these limitations, future efforts should prioritize enhancing model performance by integrating deep learning architectures capable of directly analyzing endoscopic images alongside WSI analysis. Combining these approaches with ML-based analysis of clinicopathological features in a hybrid model could leverage the strengths of each methodology, enabling a more comprehensive and accurate prediction system. This hybrid approach would minimize observer bias by automating the analysis of visual data while incorporating detailed clinical and pathological variables to provide a holistic risk assessment framework.
Expanding datasets through multicenter collaborations and balancing the dataset by increasing the representation of LNM-positive cases through data augmentation or optimized hyperparameter tuning are critical.10 Furthermore, external validation with independent data is essential to improve model robustness and generalizability. These refinements will be pivotal in translating the promising findings of Baek et al. into clinically viable tools for accurate LNM prediction in T1 CRC, ultimately reducing unnecessary surgeries.
In conclusion, AI represents a transformative advancement in predicting LNM in T1 CRC, encompassing ML models that leverage clinicopathological data and innovative approaches utilizing pathological image analysis. These technologies have shown significant potential to enhance diagnostic accuracy, reduce unnecessary surgeries, and improve patient outcomes. However, their adoption also raises ethical considerations, particularly regarding critical decision-making based on AI-derived predictions. Additionally, the integration of AI models into hospital systems necessitates careful planning, including interoperability with existing electronic medical records, clinician training, and continuous performance monitoring. Addressing these challenges through collaborative efforts involving clinicians, data scientists, and policymakers will be crucial to ensure the seamless incorporation of AI-driven tools into clinical workflows without compromising patient safety or care quality.
No potential conflict of interest relevant to this article was reported.