Artificial Intelligence

AI and U.S. Healthcare Costs: NEJM Catalyst | July 2026

Introduction: Artificial intelligence is rapidly transforming healthcare through drug discovery, clinical decision support, remote monitoring, and administrative automation. While AI is widely expected to reduce healthcare costs, this perspective argues that current payment models and market dynamics may instead increase healthcare spending despite improving quality and access. Key Takeaways: AI is likely to improve healthcare quality, efficiency, and patient access, but these benefits will not automatically translate into lower healthcare costs. Under the current fee-for-service (FFS) payment model, AI may actually increase healthcare utilization and total spending by expanding diagnostic testing, patient monitoring, and clinical services. AI has the potential to accelerate drug discovery and personalized medicine, but these innovations may initially increase pharmaceutical expenditures. Administrative automation may improve operational efficiency, yet cost savings may not be passed on to patients because of existing healthcare market structures. Value-based care provides a stronger financial framework for AI to reduce unnecessary care, improve outcomes, and slow long-term spending growth. Highly consolidated hospital systems and insurance markets may limit the ability of AI-driven efficiencies to reduce healthcare expenditures. Policymakers and payers will need to redesign payment models, reimbursement policies, and regulatory frameworks if AI is expected to generate meaningful cost savings. Without aligning financial incentives, AI is more likely to improve healthcare than reduce its cost. Clinical Impact: AI should be viewed primarily as a tool to enhance clinical outcomes, expand access, and improve efficiency rather than an immediate solution for reducing healthcare expenditure. The economic impact of AI will depend more on healthcare policy and reimbursement reform than on technological capability alone. Bottom Line: Artificial intelligence alone will not solve the healthcare cost crisis. Its ability to reduce spending depends on transitioning from fee-for-service to value-based care, supported by policies that align AI innovation with better outcomes rather than greater healthcare utilization.

Large AI Models and Healthcare: Nature Medicine | June 2026

Introduction: Large frontier AI models such as GPT-5 and Gemini have achieved impressive results across numerous healthcare benchmarks. However, high benchmark scores alone may not reflect real-world clinical reliability. This study systematically evaluated the robustness of leading AI models using adversarial testing and clinician-guided assessment. Why was this study needed? AI models are increasingly being proposed for clinical decision support. Benchmark performance may overestimate real-world clinical capability. Robustness and reliability are essential before deployment in healthcare. Multimodal medical reasoning remains insufficiently validated. Better evaluation frameworks are needed to ensure safe AI implementation. Results: Leading frontier AI models demonstrated significant vulnerability to simple adversarial changes, often producing incorrect answers despite previously excellent benchmark performance. AI systems could sometimes guess correct answers even when critical clinical information was removed, while becoming confused by minor prompt modifications and generating convincing—but incorrect—reasoning. Current healthcare benchmarks vary considerably in what they actually measure, highlighting a substantial gap between benchmark success and true clinical readiness. Clinical Impact: This study serves as an important reminder that high benchmark accuracy does not guarantee safe clinical performance. Before widespread adoption, healthcare AI should undergo rigorous real-world validation, stress testing, and clinician-led evaluation to ensure robustness, transparency, and patient safety. Bottom Line: AI is highly capable—but not yet fully reliable for autonomous clinical decision-making. Robustness, consistency, and real-world validation should become the next benchmarks before frontier AI models are integrated into routine healthcare.

AI Ethics From Silicon Valley to the Vatican: JAMA | July 2026

Introduction: Artificial intelligence is rapidly transforming medicine, but its influence extends far beyond healthcare. This JAMA AI Conversations article explores how AI ethics has become a global societal issue, engaging technology leaders, policymakers, healthcare professionals, and even the Vatican. Why was this article needed? As AI becomes increasingly embedded in clinical practice and everyday life, concerns about transparency, safety, fairness, accountability, and human values have moved beyond the technology sector. Broader ethical guidance is needed to ensure AI serves humanity responsibly. What did the article show? AI ethics is evolving from a technical discussion into a global societal dialogue. The Vatican's engagement highlights the growing recognition that AI development is fundamentally a human and ethical challenge. Interpretability, transparency, governance, and safety remain central to trustworthy AI. Healthcare is emerging as one of the most important sectors for responsible AI implementation. Collaboration between technology companies, clinicians, ethicists, governments, and international organizations is essential for developing safe AI systems. Future AI regulation should balance innovation with patient safety, human dignity, equity, and public trust. Clinical Impact: As AI becomes integrated into clinical decision-making, healthcare professionals will need to understand not only how AI performs but also how it is developed, governed, and ethically deployed. Responsible AI implementation will require transparency, human oversight, and patient-centered values. Take-Home Message: The future of medical AI will be determined not only by technological advances but also by the ethical principles that guide its development. Building trustworthy AI requires collaboration among clinicians, scientists, policymakers, industry, and society to ensure innovation remains aligned with human values.

Physician-Complementing AI in Oncology: The ASCO Post | June 2026

Introduction: Artificial intelligence is rapidly transforming oncology, evolving from image interpretation and pathology analysis to supporting complex clinical decision-making. This perspective argues that AI should enhance the capabilities of oncologists rather than replace their expertise. Why was this article needed? As AI becomes increasingly integrated into cancer care, there is growing debate over whether it should substitute physicians or function as a tool that augments clinical judgment. The authors advocate for a physician-centered model of AI adoption. What did the article show? The authors distinguish physician-substituting AI from physician-complementing AI, favoring the latter approach. AI can reduce administrative burden through ambient documentation and automated electronic medical record workflows. AI can synthesize vast clinical, genomic, laboratory, and real-world data beyond human cognitive capacity. Decision-support AI may help oncologists personalize treatment by integrating multiple patient-specific variables. AI can incorporate patient preferences, quality-of-life goals, and comorbidities into therapeutic decision-making. Rather than replacing oncologists, AI enables clinicians to spend more time on patient communication, empathy, and shared decision-making. The future of precision oncology will depend on close collaboration between physician expertise and AI-driven intelligence. Clinical Impact: Physician-complementing AI has the potential to improve clinical efficiency, personalize treatment decisions, and support evidence-based oncology while preserving the essential role of physician judgment and the doctor–patient relationship. Take-Home Message: The greatest value of AI in oncology lies not in replacing physicians but in amplifying their expertise. By combining human clinical judgment with AI-powered decision support, cancer care can become more precise, efficient, and patient-centered.

AI-Based Clinical Trial End Points: A New Era in Drug Development: NEJM AI | July 2026

Introduction Clinical trial endpoints have traditionally relied on expert human interpretation, particularly for pathology-based outcomes. However, variability between observers, cost, and time remain important limitations. This NEJM AI perspective discusses how artificial intelligence is beginning to transform endpoint assessment in clinical trials. Why Was AI Needed? * Histological assessment is often subjective. * Significant interobserver variability exists among expert pathologists. * Manual review is labour-intensive and time-consuming. * Inconsistent interpretation may affect trial outcomes and regulatory decisions. * Standardised and reproducible assessment has long been a major unmet need. What Has Changed? * In 2025, the FDA and EMA endorsed **AIM-NASH**, the first AI-enabled clinical trial end point. * AIM-NASH uses deep learning to evaluate liver biopsy histology in MASH clinical trials. * AI demonstrated high agreement with expert pathologists while improving consistency. * This represents the first regulatory acceptance of an AI-based efficacy end point. * The approval establishes a regulatory pathway for future AI-enabled biomarkers. Potential Clinical Impact * More objective assessment of treatment response. * Reduced observer variability. * Improved patient selection. * Faster central pathology review. * Shorter and potentially less expensive clinical trials. * Better standardisation across international multicenter studies. * Potential expansion to oncology, gastroenterology, cardiology, radiology, and other specialities. Take-Home Messages * AI has moved from decision support to regulatory-accepted clinical trial evaluation. * AIM-NASH represents a historic milestone in AI-assisted drug development. * AI is expected to complement—not replace—clinical experts and pathologists. * Rigorous validation and regulatory oversight remain essential. * This approval may fundamentally reshape the design and conduct of future clinical trials.

Medical AI Assistant: Publication or Medical Device?: NEJM AI | July 2026

Introduction: As artificial intelligence becomes increasingly integrated into clinical practice, an important question arises: should AI assistants be regulated as medical devices or viewed as evidence-based clinical methodologies? This NEJM AI perspective proposes a new framework that could influence the future regulation and adoption of medical AI. Why was this article needed?: Current regulatory pathways primarily treat clinical AI systems as medical devices. However, this approach may not be appropriate for open-source AI assistants that are transparent, peer-reviewed, and used by physicians as decision-support tools. The author argues that an alternative regulatory model is needed. What did the article show?: The author introduces the concept of a Medical Artificial Intelligence Assistant (MAIA)—an open-weight generative AI model integrated with a patient's health records through a secure vector database and used by physicians via an open-source interface. If the AI architecture, retrieval methods, and clinical validation are published in peer-reviewed journals, MAIA should be regarded as a published clinical methodology rather than a commercial medical device. Physicians using such validated systems are applying evidence from the medical literature, similar to adopting a published clinical guideline or risk score. Clinical Impact: This framework could encourage greater transparency, peer review, independent validation, and open scientific collaboration while reducing dependence on proprietary "black-box" AI systems. It also highlights the importance of physician oversight, reproducibility, and evidence-based implementation as AI becomes part of routine clinical practice. Take-Home Message: This perspective challenges the traditional view that every clinical AI tool should be regulated as a medical device. Instead, it proposes that transparent, peer-reviewed, open-source AI assistants may evolve into evidence-based clinical methodologies, potentially reshaping AI regulation, physician adoption, and the future standard of care.

LLMs Rapidly Transform European Gastroenterology Practice : Gut | June 2026

Introduction: Large language models (LLMs), exemplified by ChatGPT and similar artificial intelligence platforms, are rapidly reshaping healthcare delivery, medical education, and scientific research. In gastroenterology, these technologies have the potential to support clinical decision-making, streamline documentation, enhance endoscopy workflows, and improve academic productivity. Despite their growing visibility, real-world data on how gastroenterologists are using LLMs in daily practice have been limited. Problem Statement: The widespread adoption of LLMs has occurred faster than the development of formal training programs, governance frameworks, and specialty-specific guidance. Understanding current patterns of use, perceived benefits, concerns, and barriers is essential for ensuring safe and effective integration of these technologies into gastroenterology practice. Summary: The EuroGI-AI project provides the largest assessment to date of LLM adoption among European gastroenterologists. The survey demonstrated that more than half of respondents already use LLMs in their clinical or academic activities, with many engaging with these tools on a weekly or daily basis. Educational purposes, clinical decision support, and scientific writing emerged as the most common applications. Among academic users, literature summarization and language refinement were particularly valued, highlighting the growing role of AI in research productivity. Most respondents considered LLMs reliable and believed they improve professional efficiency and clinical outcomes. Notably, many clinicians also recognized potential applications within endoscopy practice. However, concerns regarding inaccuracies remained common and were associated with reduced trust in AI-generated outputs. Despite widespread use, formal training was uncommon, yet there was overwhelming support for structured educational programs and incorporation of AI competencies into gastroenterology curricula. Institutional restrictions, financial costs, technological complexity, and time limitations were identified as key barriers to broader adoption. Overall, the findings suggest that LLMs are rapidly becoming integrated into European gastroenterology practice. Future success will depend on specialty-specific development, structured training, robust governance frameworks, and ongoing evaluation to ensure safe, ethical, and clinically meaningful implementation of AI technologies.

Competing Risk Analysis in HCT & IEC Research : Transplant Cell Ther | May 2026

Introduction Competing risks are frequently encountered in hematopoietic cell transplantation (HCT) and immune effector cell (IEC) therapy research, particularly when mutually exclusive outcomes coexist. A classical example is treatment-related mortality (TRM) competing with disease relapse, where early non-relapse death precludes the possibility of relapse occurrence. Mismanagement of competing risks can produce biased survival estimates and misleading hazard interpretations, making proper statistical handling essential in transplantation and cellular therapy studies. Problem Statement Competing risks are commonly analyzed incorrectly using standard Kaplan–Meier (KM) methods that censor competing events, violating the assumption of non-informative censoring. This leads to systematic overestimation of event probabilities. In addition, Fine–Gray subdistribution hazard models are frequently misinterpreted as causal models despite their prognostic—not etiologic—nature. Many clinicians remain unfamiliar with the conceptual distinctions between cumulative incidence, cause-specific hazards and subdistribution hazards, limiting the quality and interpretability of transplant outcome research. Summary This practical review provides a clinician-oriented primer on competing risk methodology with a detailed stepwise guide for implementation using the open-access R and RStudio statistical environment. The authors first explain the limitations of naïve Kaplan–Meier analysis in competing-risk settings. When competing events are censored using standard KM methods, the resulting 1-KM estimator inflates the true probability of the event of interest because it ignores the dependency between mutually exclusive outcomes. The review emphasizes the importance of the cumulative incidence function (CIF), which appropriately accounts for the probability of remaining free from competing events. Unlike KM estimation, CIF conditions the risk of the primary event on the absence of competing outcomes over time, generating more accurate cumulative risk estimates. Gray’s test is highlighted as the preferred statistical method for comparing CIF curves between groups. The manuscript also clarifies the distinction between cause-specific hazard ratios (HRcs) and subdistribution hazard ratios (HRsd). Cause-specific Cox proportional hazard models censor competing events and estimate instantaneous event rates among individuals still at risk, making them more suitable for mechanistic or etiologic inference. In contrast, Fine–Gray models retain competing events in the risk set and directly model cumulative incidence, making HRsd more useful for patient-level prognostic prediction rather than causal interpretation. A key conceptual message is that Fine–Gray models may generate misleading associations when variables strongly influence competing events. For example, a covariate associated with higher TRM may appear to reduce relapse incidence simply because patients die before relapse can occur. The authors demonstrate through simulations that HRsd estimates become increasingly biased as the association with competing events strengthens, reinforcing the need for cautious interpretation of Fine–Gray regression outputs. The second half of the review provides a practical tutorial for conducting competing risk analyses in R. Using a real-world allogeneic HCT dataset in acute myeloid leukemia, the authors illustrate cumulative incidence estimation, visualization, Gray’s testing, cause-specific Cox regression and Fine–Gray regression modeling. Multiple R packages are introduced, including tidycmprsk, survival, ggsurvfit, and gtsummary, enabling generation of publication-ready cumulative incidence plots and regression tables. The review concludes by emphasizing that Fine–Gray models are best suited for prognostic estimation, whereas cause-specific Cox models are preferable for etiologic or mechanistic research questions. The authors strongly advocate collaboration with trained biostatisticians to minimize methodological bias and improve rigor in transplantation and cellular therapy research involving competing risks.

AI Meets Endoscopy for early ESCC Detection: Endoscopy | April 2026

Introduction Esophageal squamous cell carcinoma is a highly aggressive malignancy where early detection and accurate assessment of invasion depth are critical for determining optimal treatment strategies. Current tools such as magnifying endoscopy and endoscopic ultrasonography (EUS) are effective but operator-dependent. Advances in deep learning offer an opportunity to enhance diagnostic accuracy through automated image analysis. Problem Statement Existing diagnostic methods for early ESCC are limited by variability in expertise and lack a reliable, integrated system for simultaneous detection and invasion depth assessment. Summary This study introduces MUMA-EDx, a novel multimodal deep learning model that integrates magnifying endoscopy and EUS imaging to improve both detection and staging of early ESCC. The model was trained on a large retrospective dataset and validated prospectively, demonstrating excellent performance. For tumor detection, the model achieved outstanding accuracy with an AUC of 0.94 in retrospective validation and 1.00 in prospective testing. For invasion depth classification—a more complex and clinically critical task—the model maintained strong performance, outperforming novice endoscopists and approaching expert-level accuracy. The key innovation lies in combining multiple imaging modalities, which significantly improved diagnostic precision compared to single-modality approaches. Clinically, this tool has the potential to standardize ESCC diagnosis, reduce operator dependency, and guide more accurate treatment decisions. Key takeaway: Multimodal AI can match expert performance and may redefine early cancer detection and staging in endoscopy.

AI-Based Risk Prediction for HCC: Cancer Discovery | April 2026

Hepatocellular carcinoma (HCC) remains one of the most lethal cancers globally, largely due to late diagnosis and inadequate risk stratification. Current clinical risk scores have limited predictive accuracy and often fail to identify high-risk individuals early. With the increasing availability of large-scale healthcare data, machine learning offers an opportunity to improve early detection using routinely collected clinical information. Problem Statement Existing HCC risk prediction models are insufficient in accuracy, lack generalizability across populations, and often rely on limited variables. There is a need for a scalable, interpretable, and robust model that can integrate diverse real-world clinical data to accurately stratify HCC risk and enable early detection at a population level. Summary This large multicentric study developed an interpretable machine-learning model (PRE-Screen-HCC) using data from over 900,000 individuals across UK Biobank and the All of Us cohort. The model integrated demographics, lifestyle factors, clinical records, laboratory data, genomics, and metabolomics. It significantly outperformed existing risk scores in predicting HCC risk across diverse populations. Importantly, the model is transparent, externally validated, and made accessible via a web-based calculator, making it clinically applicable. This study represents a major step toward precision screening and early detection of HCC using real-world data and AI.

AI for Endoscopic and Histologic Assessment in IBD Trials: J Crohn’s and Colitis, March 2026

Introduction Accurate assessment of disease activity in inflammatory bowel disease (IBD) clinical trials relies on central reading of endoscopic and histologic images. Although considered the current gold standard, this process is limited by interobserver variability, operational delays, high costs, and scalability challenges. With rapid advances in artificial intelligence (AI) and machine learning (ML), automated image analysis has emerged as a promising tool to improve the accuracy, consistency, and efficiency of endpoint assessment in clinical trials. To address this evolving field, the International Organisation for the Study of IBD (IOIBD) developed an evidence-based consensus on the appropriate use of AI-assisted evaluation of endoscopic and histologic endpoints in IBD trials. Summary This consensus initiative involved a literature review (2018–2025) on AI applications in IBD endoscopy and histopathology. Based on the available evidence, the steering committee formulated 36 statements, which were voted on by 72 IOIBD experts, with ≥80% agreement required for consensus. A total of 45 experts completed the survey, and consensus was achieved for 28 statements. The panel concluded that AI-assisted central reading could significantly enhance diagnostic accuracy, improve reproducibility, reduce costs, and accelerate image interpretation in IBD trials. Importantly, the consensus strongly favoured a hybrid model combining human expert review with AI support, rather than complete replacement of human interpretation. However, several challenges remain. These include limited external validation of AI models, concerns regarding generalizability across populations and endoscopy platforms, and dependence on high-quality human-annotated training datasets. Conclusion The IOIBD consensus supports the integration of AI/ML tools into endoscopic and histologic assessment in IBD clinical trials, emphasising AI-augmented human decision-making. Future efforts should focus on validation, regulatory guidance, and multimodal data integration, which may ultimately enable broader use of AI in both clinical research and routine IBD care.

AI-Enabled Imaging for Predicting Postoperative Recurrence in Crohn’s Disease: Gut, March 2026

Postoperative recurrence (POR) remains a major challenge in Crohn’s disease (CD), occurring in up to 70% of patients within the first year after intestinal resection. Current surveillance strategies—primarily ileocolonoscopy and fecal calprotectin—have limited accuracy and lack standardized monitoring algorithms. Emerging technologies are reshaping the approach to POR detection. Advanced endoscopic imaging techniques, including confocal laser endomicroscopy and endocytoscopy, allow real-time microscopic evaluation of the anastomosis and early mucosal changes. Intestinal ultrasound and cross-sectional imaging provide non-invasive transmural assessment, improving monitoring beyond mucosal disease. Simultaneously, multi-omics approaches (genomics, transcriptomics, proteomics, metabolomics, and microbiome profiling) are identifying biological pathways associated with recurrence risk. Integration of these datasets using artificial intelligence (AI) may enable predictive multimodal models that combine clinical, imaging, and molecular data. Such AI-driven strategies could allow personalized risk stratification, earlier detection of recurrence, and tailored postoperative therapy, moving Crohn’s disease management toward precision medicine. However, prospective validation and clinical workflow integration are still required.

ACG Delphi Consensus on AI in GI- AMJ Feb.26

This American College of Gastroenterology (ACG)–led modified Delphi consensus provides a comprehensive framework for responsible integration of artificial intelligence (AI) into gastroenterology, hepatology, and endoscopy practice. A multidisciplinary task force of 32 experts and 12 industry partners reviewed the literature across five domains: endoscopy, clinical/practice management applications, IBD and liver disease, training/education, and ethics/equity. Out of 43 proposed statements, 100% ultimately reached consensus (≥70% agreement). In endoscopy, strong evidence supports computer-aided detection (CADe) for improving adenoma detection rates and reducing miss rates in controlled settings, though real-world impact and long-term outcomes (e.g., interval cancer reduction) remain uncertain. The panel emphasizes robust validation using heterogeneous datasets to mitigate bias. Beyond endoscopy, promising applications include ambient AI scribes, natural language processing for coding, workflow optimization, and prior authorization support. In IBD and hepatology, AI shows potential in improving diagnostic accuracy, risk stratification, and therapeutic guidance. Importantly, the consensus stresses that AI should augment—not replace—clinical expertise. Recommendations include structured AI curricula for trainees while preserving independent procedural competence to prevent “deskilling.” Ethical priorities focus on data governance, chain-of-custody protections, bias mitigation, specialty oversight, and equitable reimbursement models. Future priorities include prospective pragmatic trials, multi-institutional data-sharing consortia, and transparent validation standards to ensure safe, effective, and equitable AI adoption in gastrointestinal practice.

Cholangioscopy in biliary tract disease(Endoscopy, Jan-2026)

Cholangioscopy is an advanced endoscopic procedure that allows direct visualization of the bile ducts and is commonly used to evaluate biliary tract diseases, including strictures, stones, and malignancies. It is particularly valuable in cases where traditional imaging modalities, such as ERCP, fail to provide definitive diagnoses. Cholangioscopy can facilitate targeted biopsies, therapeutic interventions (e.g., lithotripsy for bile duct stones), and aid in distinguishing benign from malignant biliary strictures. Recent advancements in artificial intelligence (AI) have significantly improved the diagnostic accuracy of cholangioscopy. For example, studies have shown that AI systems analyzing cholangioscopy footage outperform traditional sampling methods like brush cytology and forceps biopsy in identifying malignancy. AI systems, with sensitivity and specificity rates as high as 80.0% and 90.3%, respectively, can enhance clinical decision-making and reduce diagnostic uncertainty.

Cholangioscopy in biliary tract disease (Thieme, Jan-2026)

Cholangioscopy is a specialized endoscopic procedure that allows direct visualization of the bile ducts, aiding in the diagnosis and management of biliary tract diseases such as strictures, stones, and malignancies. It is particularly valuable when conventional imaging and sampling techniques, such as brush cytology or forceps biopsy during ERCP, fail to provide definitive results. Cholangioscopy facilitates targeted biopsies and enhances diagnostic accuracy. Recent advancements include the integration of artificial intelligence (AI) systems to analyze cholangioscopy footage. Studies, such as the one described in the provided context, demonstrate that AI systems can significantly outperform traditional ERCP sampling modalities in diagnosing malignancy. The AI system reviewed was shown to achieve higher sensitivity, specificity, and overall accuracy in classifying biliary strictures, making it a promising tool for improving diagnostic confidence and patient outcomes.

Laboratory-Based ML Model for Predicting Advanced Fibrosis in MASLD

The study described focuses on the development of a laboratory-based machine learning (ML) model aimed at predicting advanced liver fibrosis in patients with metabolic dysfunction-associated steatotic liver disease (MASLD). Below is a detailed explanation of the model and its significance: ### **Background** MASLD is a condition characterized by fat accumulation in the liver due to metabolic dysfunction, and it can silently progress to advanced liver fibrosis, which poses severe health risks. Traditionally, liver fibrosis is diagnosed using invasive liver biopsy, which is not only uncomfortable for patients but also costly and resource-intensive. Noninvasive methods like imaging and blood-based tests exist but have limitations in accuracy and accessibility. To address these challenges, the researchers developed a machine learning-based model that utilizes routine clinical and laboratory data to predict advanced fibrosis. ### **Study Design** The research was conducted as a retrospective analysis involving patients with biopsy-confirmed MASLD. The key components of the study included: - **Data Integration**: Combining demographic data, clinical history, and standard laboratory test results to create a comprehensive dataset. - **Algorithm Evaluation**: Testing multiple machine learning algorithms to identify the optimal model for distinguishing advanced fibrosis from non-advanced stages. - **Feature Selection**: Reducing the complexity of the model by identifying a minimal set of variables that maintain predictive accuracy. ### **Key Findings** - **Best Performing Algorithm**: The Extra Trees classifier emerged as the most effective machine learning model. It demonstrated robust predictive performance and consistency across various validation settings, making it suitable for clinical use. - **Simplified Model**: To improve practicality, the model was refined to rely on a small set of commonly available clinical and laboratory parameters. This simplification ensures easy integration into routine care without requiring specialized tools or expertise. - **Web-Based Tool**: The finalized model was deployed as an accessible, web-based application. This allows clinicians to assess the risk of advanced fibrosis in MASLD patients noninvasively and efficiently. ### **Clinical Significance** 1. **Noninvasive Assessment**: The ML model eliminates the need for invasive liver biopsy, reducing patient discomfort and healthcare costs. 2. **Accessibility**: By relying on routinely available laboratory data, the model can be used in resource-limited settings where specialized diagnostic equipment may not be available. 3. **Early Detection**: The ability to identify advanced fibrosis at earlier stages can improve patient outcomes by enabling timely intervention and management. 4. **Scalability**: The web-based deployment ensures that the tool can be widely adopted in clinical practice, supporting scalable fibrosis assessment. ### **Future Implications** With further external validation, this machine learning approach has the potential to transform the evaluation of liver fibrosis in MASLD patients. It could be integrated into routine workflows to assist in: - Risk stratification for MASLD patients. - Guiding clinical decision-making, such as prioritizing patients for more intensive monitoring or treatment. - Enhancing care in underserved areas where invasive diagnostic methods are impractical. ### **Conclusion** The laboratory-based ML model for predicting advanced fibrosis in MASLD represents a significant step forward in noninvasive, accessible, and scalable liver disease evaluation. By leveraging machine learning and commonly available clinical data, this tool has the potential to improve the detection and management of advanced fibrosis, ultimately enhancing patient outcomes and reducing healthcare burdens.

Trial of Artificial Intelligence and Adjunctive Polyp Detection - J of JGH - Jan,26

The study referenced, titled "Trial of Artificial Intelligence and Adjunctive Polyp Detection," published in the Journal of Gastroenterology and Hepatology (JGH) on January 26, evaluates the effectiveness of artificial intelligence (AI)-assisted polyp detection systems in improving adenoma detection rates during colonoscopy procedures. The trial focused on whether combining AI technology with established colonoscopy techniques could further enhance detection outcomes. ### Key Details of the Study: 1. **Objective**: The study aimed to determine the incremental value of AI-based polyp detection systems when used alongside traditional colonoscopy practices, such as extended withdrawal time, retroflexion, patient positioning adjustments, and advanced imaging techniques. 2. **Methodology**: - **Design**: A prospective randomized controlled trial conducted at a single hospital. - **Participants**: Multiple experienced endoscopists performed colonoscopies, with patients randomized into two groups: one using AI-assisted detection and the other following conventional procedures. - **Adjunctive Techniques**: Endoscopists were allowed to use supplementary detection-enhancing techniques based on clinical judgment rather than strict protocols. 3. **Findings**: - **AI's Impact**: Colonoscopies supported by AI demonstrated a trend toward improved adenoma detection rates compared to standard procedures. - **Screening Colonoscopy Benefits**: The use of AI was particularly beneficial in patients undergoing screening colonoscopies, leading to higher numbers of detected polyps and overall detection performance. - **Role of Conventional Practices**: Traditional practices such as extended withdrawal time and advanced imaging techniques were strongly associated with improved outcomes. AI remained an independent contributor to better adenoma detection rates when combined with these practices. 4. **Conclusion**: - AI-assisted polyp detection systems significantly enhance adenoma detection, even when used by experienced endoscopists. - The effectiveness of AI is maximized when integrated with established procedural strategies, emphasizing the importance of combining technological advancements with meticulous colonoscopy practices to achieve optimal detection performance. This study underscores the potential of AI in advancing colorectal cancer prevention by improving adenoma detection rates, particularly in screening settings. It advocates for the integration of AI technology with high-quality procedural techniques to optimize patient outcomes.

Perihilar Cholangiocarcinoma and Deep Learning

### Perihilar Cholangiocarcinoma (pCCA) and Deep Learning #### **What is Perihilar Cholangiocarcinoma (pCCA)?** Perihilar cholangiocarcinoma (pCCA) is a type of bile duct cancer that arises near the liver's hilum, where the bile ducts exit the liver. It is the most common type of cholangiocarcinoma (bile duct cancer) and is often associated with primary sclerosing cholangitis (PSC), a chronic liver disease characterized by inflammation and scarring of the bile ducts. Unfortunately, pCCA is frequently diagnosed at a late stage, making it one of the leading causes of mortality in PSC patients. Early detection of pCCA is critical because it allows for curative treatments such as liver transplantation or surgical resection. --- #### **What is Deep Learning?** Deep learning is a subset of artificial intelligence (AI) and machine learning that uses neural networks with multiple layers to analyze complex datasets. A deep learning model "learns" patterns and features from raw data, such as medical images, without requiring explicit programming. It is particularly useful in image analysis, where it can identify subtle, complex patterns that may be missed by human experts. In this study, a specific deep learning architecture called **3D DenseNet-121** was used. This is a convolutional neural network (CNN) designed to process three-dimensional (3D) medical imaging data, such as MRI scans. DenseNet-121 is known for its efficiency in feature extraction and its ability to learn hierarchical representations of the data, making it well-suited for detecting abnormalities in medical images. --- #### **How MRI Helps Identify the Severity of PSC** Magnetic Resonance Imaging (MRI), particularly contrast-enhanced MRI, is a non-invasive imaging technique that provides detailed visualization of the liver, bile ducts, and surrounding tissues. In patients with PSC, MRI can help identify: 1. **Bile Duct Abnormalities**: PSC causes strictures (narrowing) and dilation of bile ducts, which can be visualized on MRI. 2. **Mass Lesions**: The presence of a mass or thickening in the bile ducts may indicate cholangiocarcinoma. 3. **Liver Damage**: MRI can assess liver fibrosis, cirrhosis, or other complications of PSC. 4. **Vascular Involvement**: MRI can reveal whether cancer has invaded nearby blood vessels, which is critical for staging and treatment planning. However, early-stage pCCA is often challenging to detect using MRI alone because the cancer may not form a distinct mass or may blend with the background fibrosis and inflammation caused by PSC. This is where deep learning can play a transformative role. --- #### **How Deep Learning Analyzes MRI to Predict Cancer in PSC** Deep learning models, like the 3D DenseNet-121 used in this study, analyze MRI images to detect patterns indicative of early-stage pCCA. Here's how the process works: 1. **Data Input**: - The model is trained on a dataset of MRI images from patients with and without pCCA. - Images are labeled based on clinical diagnoses, such as "pCCA present" or "pCCA absent." 2. **Feature Extraction**: - The model automatically learns to identify features associated with pCCA, such as subtle changes in bile duct structure, tissue texture, or contrast enhancement patterns. - Unlike traditional methods, the model does not require manual feature engineering by radiologists. 3. **Prediction**: - Once trained, the model analyzes new MRI images and predicts whether pCCA is present or absent. - The model outputs sensitivity (ability to detect true positives) and specificity (ability to avoid false positives). 4. **Performance Comparison**: - In this study, the deep learning model significantly outperformed expert radiologists in detecting early-stage pCCA: - **Sensitivity**: The model detected 87.9% of pCCA cases compared to 50% by radiologists. - **Specificity**: The model had a specificity of 84.1%, slightly lower than the radiologists' 100%, but still clinically acceptable. - **Area Under the Receiver Operating Curve (AUC)**: The model achieved an AUC of 86.0%, compared to 75.0% for radiologists. 5. **Mass-Independent Detection**: - Even in cases where no visible mass was present (a common scenario in early-stage pCCA), the model achieved a higher sensitivity (91.6%) than radiologists (50.6%). --- #### **Advantages of Deep Learning in pCCA Detection** 1. **Early Detection**: The model can identify subtle changes in MRI that may indicate early-stage pCCA, even before a mass is visible. 2. **Improved Sensitivity**: Fewer cases of cancer are missed compared to radiologists. 3. **Efficiency**: The model can analyze large volumes of MRI data quickly and consistently. 4. **Non-Invasive**: Deep learning enhances the diagnostic power of existing imaging techniques without requiring additional invasive procedures. --- #### **Conclusion** The application of deep learning, specifically the 3D DenseNet-121 model, represents a significant advancement in the detection of perihilar cholangiocarcinoma in patients with primary sclerosing cholangitis. By leveraging MRI data, the model can detect early-stage pCCA with higher sensitivity and comparable specificity to expert radiologists. This technology has the potential to improve outcomes by enabling earlier diagnosis and timely curative interventions for PSC patients at risk of developing pCCA.

Multicenter Validation of an AI-Based Cholangioscopy System for Biliary Disease Evaluation

The multicenter validation study focused on assessing the performance of an artificial intelligence (AI)-based system for evaluating biliary tract disease using cholangioscopy video footage. Accurate differentiation between benign and malignant biliary strictures is a major clinical challenge, as traditional diagnostic methods, such as endoscopic retrograde cholangiopancreatography (ERCP)-based brush cytology and forceps biopsy, are limited by their low sensitivity. Cholangioscopy offers direct visualization and targeted sampling of biliary pathology, but its reliance on human interpretation introduces diagnostic errors. ### Key Objectives: 1. **Evaluate AI System Performance:** The study aimed to validate the ability of the AI system to analyze unedited cholangioscopy recordings and accurately classify biliary strictures as benign or malignant. 2. **Compare Diagnostic Accuracy:** The AI system's predictions were compared to traditional ERCP-based sampling techniques (brush cytology, forceps biopsy, and their combined use). 3. **Assess Generalizability:** The study examined the system's robustness across multiple institutions and diverse patient populations. ### Methods: - **Data Collection:** Cholangioscopy videos were gathered from multiple academic centers. - **AI Analysis:** The AI system processed the videos without retraining and independently generated diagnostic predictions. - **Comparison:** AI predictions were compared to diagnostic results obtained from conventional ERCP sampling methods. ### Results: - **Superior Diagnostic Accuracy:** The AI system consistently outperformed traditional ERCP-based techniques in classifying biliary strictures as benign or malignant. - **Generalizability:** The system demonstrated strong performance across different institutions and patient populations, confirming its robustness and applicability. - **Enhanced Clinical Utility:** The AI system showed potential as an adjunctive tool, improving early detection of malignancies and reducing reliance on less sensitive methods. ### Implications: 1. **Improved Diagnostic Accuracy:** AI-assisted cholangioscopy analysis significantly enhances the ability to differentiate between benign and malignant biliary strictures, addressing a critical clinical challenge. 2. **Streamlined Workflow:** The system integrates seamlessly into existing procedural workflows without requiring additional retraining or modifications. 3. **Potential for Early Detection:** By improving malignancy detection rates, the AI system could lead to earlier interventions and better patient outcomes. 4. **Reduced Diagnostic Errors:** The AI system minimizes reliance on subjective visual interpretation by clinicians, enhancing reliability. ### Conclusion: This multicenter validation study confirms that AI-based analysis of cholangioscopy footage is a powerful tool for biliary disease evaluation. Its high diagnostic accuracy, generalizability, and ability to integrate into current clinical workflows suggest that AI systems could play a transformative role in the diagnosis and management of biliary tract diseases.

AI algorithm for early identification of MASLD

The AI algorithm developed for the early identification of Metabolic Dysfunction–Associated Steatotic Liver Disease (MASLD) is a significant advancement in leveraging artificial intelligence and natural language processing (NLP) technologies to detect this prevalent yet often underdiagnosed condition. Here's a detailed explanation of the algorithm and its application: ### Purpose: The primary goal of the AI algorithm is to identify patients with MASLD early, especially since the disease is often asymptomatic in its early stages and may only become evident when it progresses to cirrhosis. Early detection can enable timely interventions, prevent disease progression, and improve patient outcomes. ### Methodology: 1. **Natural Language Processing (NLP):** - The algorithm incorporates an NLP component that scans electronic health records (EHRs), particularly abdominal imaging reports, to identify signs of hepatic steatosis (fatty liver). - It analyzes free-text imaging reports to detect mentions of hepatic steatosis with a high degree of precision. 2. **MASLD Criteria Application:** - The algorithm applies clinical criteria for MASLD, which includes excluding other causes of liver disease such as significant alcohol use. - It utilizes additional data from the EHR, such as alcohol consumption history, to ensure that identified cases meet the MASLD definition. 3. **Validation:** - The algorithm's performance was validated through manual review of patient cohorts generated monthly. - It achieved a positive predictive value (PPV) of over 93% for identifying MASLD and a PPV of up to 99.4% for detecting hepatic steatosis in imaging reports. - The algorithm also demonstrated a ~95% negative predictive value for excluding patients with alcohol use. ### Results: - Over a 6-month period, the algorithm identified 957 individuals with MASLD from EHR data. - Interestingly, only 14.6% of these patients (n=140) had a MASLD-related diagnosis code, highlighting the significant underdiagnosis of the condition in clinical practice. - This demonstrates the algorithm's potential to uncover a hidden population of patients with MASLD who might otherwise go undetected. ### Implications: - **Enhanced Detection:** The AI algorithm enables large-scale, automated screening for MASLD across healthcare systems, overcoming the limitations of manual chart reviews and reliance on diagnosis codes alone. - **Targeted Interventions:** By identifying patients earlier, healthcare providers can implement interventions such as lifestyle modifications, weight management, and monitoring to prevent disease progression. - **Adaptability:** The algorithm can be adapted by other healthcare institutions to improve MASLD detection and track alcohol use patterns in their patient populations. ### Conclusion: This AI algorithm offers a powerful tool for the early identification of MASLD with high accuracy and reliability. It addresses the challenge of underdiagnosis by leveraging EHR data and advanced NLP techniques. By integrating this tool into clinical workflows, healthcare providers can enhance MASLD detection, optimize patient care, and reduce the burden of liver disease on healthcare systems.

GEMA-AI: Gender-Equity Model for Liver Transplant Waiting List Prioritization

GEMA-AI, or the Gender-Equity Model for Liver Transplant Waiting List Prioritization, is an innovative artificial intelligence-based model designed to improve fairness and accuracy in liver transplant allocation, with a specific focus on addressing gender disparities. Below is a detailed explanation of the model, its features, and its potential impact: ### 1. **Goal of GEMA-AI** - The primary goal of GEMA-AI is to reduce disparities in access to liver transplants, particularly for women, who have historically been disadvantaged under traditional prioritization models. - The model aims to improve equity and ensure that transplant candidates with the highest clinical urgency are prioritized appropriately. ### 2. **Key Features of GEMA-AI** - **Same Inputs as Existing Models**: GEMA-AI uses the same core laboratory inputs as widely used liver allocation scores, such as MELD-Na and MELD 3.0. These inputs include INR (International Normalized Ratio), bilirubin, sodium, and RFH-GFR (a renal function estimator). - **Explainable Artificial Intelligence (AI)**: Unlike traditional "black-box" AI models, GEMA-AI is built as an explainable artificial neural network. This ensures transparency in decision-making, which is critical for clinical applications. - **Nonlinear Modeling**: GEMA-AI captures nonlinear relationships between variables and mortality risk, such as the "U-shaped" risk associated with sodium levels. This allows it to more accurately assess patients with extreme lab values or severe clinical conditions. - **International Development and Validation**: The model was developed using UK liver transplant registry data and externally validated using Australian transplant cohorts, ensuring its robustness across different populations. - **Large-Scale Cohorts**: The combined dataset used for training and validation included 9,320 adult liver transplant candidates, providing a solid foundation for the model's development. ### 3. **Clinical Endpoint and Performance** - **Primary Clinical Endpoint**: The model focuses on predicting 90-day mortality or delisting due to clinical deterioration, which reflects the urgency of a patient’s need for a transplant. - **Improved Discrimination**: GEMA-AI demonstrated better discriminatory performance compared to existing models like MELD-Na, MELD 3.0, and GEMA-Na. This means it can more accurately predict which patients are at the highest risk. - **Stronger Benefit for Women**: The model’s advantage was particularly pronounced among women, addressing a known bias in previous allocation systems. This could lead to a significant reduction in gender disparities in transplant access. ### 4. **Handling Extreme Cases** - **Better Calibration for High-Risk Patients**: GEMA-AI showed superior calibration in patients with extreme lab values or severe clinical features, such as ascites (fluid buildup in the abdomen) and poor renal function. - **Extreme Lab Values**: Traditional linear models often mis-rank the sickest patients due to their inability to handle extreme values. GEMA-AI’s nonlinear approach avoids this issue, ensuring fair prioritization for these individuals. ### 5. **Impact on Patient Prioritization** - **Meaningful Reclassification**: GEMA-AI was able to re-rank patients by clinically meaningful score differences compared to existing models. This reprioritization often benefited patients with more severe clinical profiles. - **Potential Mortality Reduction**: Modeling suggests that GEMA-AI could prevent a significant number of waiting-list deaths, with the greatest impact observed among women. - **Prioritizing Sicker Profiles**: Patients with worse renal function and more severe symptoms were often prioritized higher by GEMA-AI, reflecting its ability to identify those in greatest need. ### 6. **Implementation Considerations** - **Reassessment Over Time**: The authors recommend reassessing patients at least every three months to ensure that prioritization remains accurate as clinical conditions change. - **Flexibility in Implementation**: GEMA-AI’s performance remained strong even when certain variables (e.g., ascites) were excluded, suggesting it can be adapted to different healthcare systems. - **Further Validation**: While the model has shown strong results in the UK and Australia, additional validation is needed before it can be adopted in other regions or healthcare systems. ### 7. **Broader Implications** - **Advancing Organ Allocation**: GEMA-AI demonstrates the potential of explainable machine learning to improve fairness and accuracy in organ allocation. By addressing biases and leveraging nonlinear modeling, it represents a significant advancement over traditional regression-based systems. - **Gender Equity in Healthcare**: The model underscores the importance of addressing gender disparities in medical decision-making and provides a framework for achieving more equitable outcomes in other areas of healthcare. ### 8. **Conclusion** GEMA-AI is a groundbreaking approach to liver transplant prioritization that combines explainable AI, robust modeling, and a focus on equity. By addressing longstanding gender disparities and improving the prioritization of critically ill patients, it has the potential to save lives and set a new standard for fairness in organ allocation systems worldwide.

AI-assisted versus conventional reading in pan-intestinal capsule endoscopy

The comparison between AI-assisted pan-intestinal capsule endoscopy (AI-PCE) and conventional reading of pan-intestinal capsule endoscopy (CR-PCE) highlights significant advancements in the diagnostic capabilities of AI technology in the context of suspected mid-lower gastrointestinal bleeding (MLGIB). Here’s a detailed breakdown: ### Overview of Pan-Intestinal Capsule Endoscopy (PCE) PCE is a minimally invasive diagnostic tool used to evaluate the gastrointestinal (GI) tract, particularly for detecting potentially haemorrhagic lesions (PHLs) in cases of suspected MLGIB. While effective, the traditional method of reading PCE (CR-PCE) is labor-intensive, time-consuming, and prone to variability and missed lesions due to human error. ### AI-Assisted PCE (AI-PCE) AI-PCE employs artificial intelligence, specifically a convolutional neural network (CNN), to assist in detecting lesions within the GI tract. This technology automates the lesion detection process, potentially improving diagnostic accuracy and efficiency. --- ### Key Findings from the Study #### 1. **Improved Sensitivity and Negative Predictive Value (NPV)** AI-PCE demonstrated significantly higher sensitivity and NPV compared to CR-PCE: - **Sensitivity**: AI-PCE achieved a sensitivity of 95% overall, compared to 67% for CR-PCE. This means AI-PCE was much more effective at detecting lesions. - **Negative Predictive Value (NPV)**: AI-PCE had an NPV of 92% versus 63% for CR-PCE, indicating a reduced likelihood of missing lesions. #### 2. **Performance by Intestinal Segment** - **Small Bowel**: - Sensitivity: 96% (AI-PCE) vs. 59% (CR-PCE). - NPV: 97% (AI-PCE) vs. 76% (CR-PCE). - **Colon**: - Sensitivity: 90% (AI-PCE) vs. 68% (CR-PCE). - NPV: 94% (AI-PCE) vs. 86% (CR-PCE). #### 3. **Lesion Detection** AI-PCE outperformed CR-PCE in detecting various lesion types: - **Vascular Lesions**: 51% detection rate with AI-PCE vs. 33% with CR-PCE. - **Ulcers/Erosions**: 16% detection rate with AI-PCE vs. 7% with CR-PCE. - **Protuberant Lesions**: Comparable detection rates (5% vs. 4%). - **Active Bleeding**: Comparable detection rates (7% vs. 7%). #### 4. **Comparison with Colonoscopy** AI-PCE also outperformed traditional colonoscopy: - Sensitivity: 90% (AI-PCE) vs. 32% (colonoscopy). - Positive Predictive Value (PPV): 100% (AI-PCE) vs. 65% (colonoscopy). - NPV: 94% (AI-PCE) vs. 65% (colonoscopy). --- ### Advantages of AI-PCE Over CR-PCE and Colonoscopy 1. **Higher Diagnostic Accuracy**: AI-PCE provides significantly better sensitivity and NPV, reducing the risk of missed lesions. 2. **Minimally Invasive**: Unlike colonoscopy, PCE is non-invasive, making it a more comfortable option for patients. 3. **Consistency and Reliability**: AI reduces reader dependency, variability, and the likelihood of human error in lesion detection. 4. **Time Efficiency**: Automating the review process with AI can save time for clinicians. --- ### Implications for Research, Practice, and Policy 1. **Redefining Diagnostic Standards**: AI-PCE may become the first-line diagnostic tool for suspected MLGIB, reducing the reliance on invasive procedures like colonoscopy. 2. **Improved Patient Outcomes**: By reducing false negatives, AI-PCE can decrease the need for repeated procedures and enhance early detection of critical lesions. 3. **Integration of AI into Clinical Workflows**: The study supports the incorporation of AI into capsule endoscopy to improve diagnostic accuracy and efficiency. 4. **Future Research**: Further validation studies and cost-effectiveness analyses are necessary to confirm the widespread applicability of AI-PCE. --- ### Conclusion AI-assisted PCE represents a significant advancement over conventional reading methods and even colonoscopy in the diagnosis of suspected MLGIB. With its superior sensitivity, reliability, and minimally invasive nature, AI-PCE has the potential to revolutionize the diagnostic approach to gastrointestinal bleeding and set a new standard for clinical practice. However, further research is needed to validate these findings and address the cost implications of integrating AI into routine diagnostic workflows.

Machine learning in GI Endoscopy

Machine learning (ML) has revolutionized gastrointestinal (GI) endoscopy by improving diagnostic precision, reducing variability, and streamlining workflows. ML algorithms analyze endoscopic images or videos to detect patterns, identify lesions, and assist in diagnosing, characterizing, or localizing GI diseases such as polyps, cancers, inflammatory bowel disease, and ulcers. This integration enhances lesion detection rates, predicts disease progression, and standardizes examination quality, reducing operator-dependent variability. Key applications of ML in GI endoscopy include: 1. **Computer-Assisted Detection (CADe):** Identifying lesions such as polyps or bleeding during endoscopy. 2. **Computer-Assisted Diagnosis (CADx):** Characterizing and classifying lesions to aid in diagnosis. 3. **Therapeutic Assistance:** Supporting interventional decisions, such as predicting lymph node metastasis (LNM) in colorectal cancer (CRC) to reduce unnecessary surgeries. 4. **Technical Support:** Enhancing procedural performance, such as scope insertion guidance for more precise and comfortable colonoscopies. Despite these advancements, challenges remain. High-quality, diverse, and annotated training datasets are essential, but current datasets often lack diversity, contain biases, and underrepresent rare diseases. Manual annotation is time-consuming, and low-quality images or artefacts can hinder accuracy. Additionally, balancing model accuracy with speed, cost, and computational efficiency is crucial. Ethical and regulatory concerns, such as patient privacy, algorithmic bias, data security, and transparency, are significant barriers to adoption. Real-time deployment is also limited by technical infrastructure, false positives, and workflow disruptions. Future success will require standardized protocols, clinician training, interdisciplinary collaboration, and advancements in explainable AI to build trust and ensure safe, effective integration of ML into GI endoscopy.

Submucosal vessel detection during third-space endoscopy - Role of AI

The role of artificial intelligence (AI) in submucosal vessel detection during third-space endoscopy, such as endoscopic submucosal dissection (ESD) and peroral endoscopic myotomy (POEM), has been explored to enhance procedural safety and efficiency. These advanced endoscopic techniques involve navigating the submucosal layer, where unseen blood vessels pose a risk of accidental injury and significant bleeding. AI has shown promise in improving the detection of submucosal vessels during these procedures. By assisting endoscopists, AI enhances vessel identification accuracy and reduces the time required for detection, thereby facilitating smoother procedural workflows. While AI may occasionally generate false positives, these are brief and unlikely to disrupt the procedure. The findings suggest that AI can serve as a valuable tool for improving safety and reducing complications in complex therapeutic endoscopy. Although further clinical trials are necessary, this study highlights AI's potential to support endoscopists and improve outcomes in third-space endoscopic procedures.

CADe colonoscopy in colorectal cancer screening - ESGE Postional Statement

The ESGE (European Society of Gastrointestinal Endoscopy) Position Statement on computer-assisted detection (CADe) in colonoscopy for colorectal cancer (CRC) screening and post-polyp surveillance provides a cautious but favorable recommendation for its use. Here is a comprehensive summary of the statement and its implications: ### What is CADe in Colonoscopy? CADe, or computer-assisted detection, is an artificial intelligence (AI) tool designed to enhance the detection of polyps during colonoscopy procedures. Polyps are growths in the colon or rectum that can sometimes develop into colorectal cancer if left untreated. CADe systems analyze the video feed from the colonoscope in real-time, helping endoscopists identify polyps that might otherwise be missed, particularly small or flat polyps that are more challenging to detect. ### ESGE's Evaluation Process The ESGE Position Statement was developed through a structured evaluation process, which included: 1. **Systematic Reviews**: A thorough review of existing evidence on CADe's effectiveness and safety. 2. **Microsimulation Modeling**: Simulations to predict the potential impact of CADe on colorectal cancer incidence and mortality. 3. **Patient Values and Preferences**: Consideration of patient perspectives regarding the use of this technology in screening and surveillance. A panel of European experts assessed the evidence, weighing the potential benefits and harms of CADe use during colonoscopy. The final vote on December 18, 2024, showed a majority (68.4%, or 13 out of 19 members) in favor of recommending CADe, while 31.6% (6 members) voted against it, highlighting ongoing uncertainty in the field. ### ESGE's Recommendation The ESGE issued a **weak but favorable recommendation** for the use of CADe in colonoscopy for CRC screening and post-polyp surveillance. The key points of the recommendation are as follows: 1. **Potential Benefits**: - **Improved Polyp Detection**: CADe enhances the detection of small and flat polyps, which are often missed during traditional colonoscopy. This improvement could help prevent the progression of these polyps into colorectal cancer. - **Reduction in CRC Incidence and Mortality**: Evidence suggests that CADe may lead to a small but meaningful reduction in the rates of colorectal cancer and related deaths. 2. **Concerns and Limitations**: - **Limited Evidence**: The current body of research supporting CADe is still limited and carries significant uncertainty. - **Modest Absolute Benefit**: While CADe shows promise, the overall reduction in cancer cases and deaths is considered modest. - **Risk of Overdiagnosis**: CADe increases the detection of non-threatening polyps, which could lead to overdiagnosis, unnecessary follow-up colonoscopies, and increased patient burden. 3. **Patient Preferences**: - The panel concluded that most well-informed patients who have already decided to undergo colonoscopy for screening or surveillance would likely prefer CADe-assisted procedures, given the potential for improved detection. ### Overall Conclusion The ESGE cautiously supports the use of CADe during colonoscopy, recognizing its potential to improve polyp detection and reduce colorectal cancer incidence and mortality. However, the recommendation is classified as weak due to the limited strength of current evidence, the modest absolute benefits, and the potential downsides, such as overdiagnosis and increased surveillance burden. The ESGE emphasizes the importance of balancing these benefits with patient preferences and the risks of unnecessary procedures. Further high-quality research is needed to strengthen the evidence base and address the uncertainties surrounding CADe's role in colorectal cancer screening and surveillance.

AI for submucosal vessel detection during third-space endoscopy

Submucosal vessel detection is critically important in third-space endoscopy because this advanced procedure involves creating a pathway within the layers of the gastrointestinal wall to treat conditions like achalasia, tumors, or gastrointestinal leaks. During this process, endoscopists work near delicate blood vessels embedded in the submucosal layer. If these vessels are accidentally injured, it can lead to severe bleeding, complications, or even life-threatening situations. Accurate identification of submucosal vessels helps the endoscopist navigate safely, avoid vessel injury, and perform the procedure with greater precision. Artificial intelligence (AI) has the potential to significantly enhance vessel detection during third-space endoscopy. AI algorithms, trained on large datasets of endoscopic images, can automatically identify and outline submucosal vessels in real-time. This reduces the cognitive burden on endoscopists, allowing them to focus on the procedure while relying on AI to highlight high-risk areas. AI can improve safety by offering consistent vessel detection even in challenging conditions like poor visibility or anatomical variations. Additionally, AI could assist less experienced trainees by acting as a "second set of eyes," enhancing their ability to recognize vessels accurately. However, for AI to be effectively integrated into clinical practice, it must demonstrate high accuracy, robustness, and reliability through rigorous testing in real-world scenarios.

Cancer Recurrence in Patients With CRC - Role of AI

The role of AI in detecting and analyzing cancer recurrence in patients with colorectal cancer (CRC) has been transformative, particularly with advancements like the DFCI-imaging-student model. Below is a detailed explanation of how AI contributes to this area: ### 1. **Automated Recurrence Detection Using NLP** - The DFCI-imaging-student model utilizes natural language processing (NLP) to extract cancer recurrence information from unstructured radiology reports. This eliminates the need for manual record reviews, which are often time-intensive and prone to human error. - By processing large volumes of radiology reports efficiently, the AI model ensures that recurrence data is captured with high precision and reliability. ### 2. **Study Design for Validation** - The model was applied to a cohort of 200 colorectal cancer patients diagnosed with stage III disease, alongside 200 breast cancer patients, as part of a longitudinal study at Kaiser Permanente Northern California. - Patients were followed for up to 19 years (2005–2024), providing a robust dataset for validating the AI model's performance in detecting recurrence. ### 3. **Model Performance in Colorectal Cancer** - The AI model demonstrated high sensitivity (94.3%) and specificity (86.9%) in detecting recurrence for colorectal cancer patients. - Sensitivity refers to the model's ability to correctly identify patients with recurrence, while specificity indicates its accuracy in identifying those without recurrence. These metrics confirm the reliability of the model in clinical settings. ### 4. **Precision in Estimating Time-to-Recurrence** - Among correctly identified recurrence cases in colorectal cancer patients, the median error in estimating the timing of recurrence was minimal—just 0.48 months. - This near-exact alignment with manual reviews highlights the model's ability to provide precise and actionable information regarding when recurrence occurs. ### 5. **Clinical and Research Impact** - The AI model enables efficient, large-scale analysis of recurrence data, which is critical for improving cancer surveillance and patient outcomes. - Real-time monitoring of recurrence trends allows clinicians to tailor treatment plans and follow-up strategies more effectively. - The model also supports multicenter oncology research, fostering collaboration and enabling the analysis of recurrence patterns across diverse patient populations. ### 6. **Future Implications** - AI-driven recurrence detection models like this one pave the way for advancements in personalized medicine. By understanding recurrence patterns, researchers can develop targeted therapies and interventions to prevent or manage recurrence more effectively. - Additionally, such models can be integrated into electronic health record (EHR) systems to provide automated alerts for clinicians, enhancing early detection and improving patient care. In summary, AI plays a crucial role in colorectal cancer recurrence detection by automating data extraction, providing accurate and timely insights, and supporting clinical decision-making and research efforts. The DFCI-imaging-student model exemplifies how AI can revolutionize oncology care, particularly in improving outcomes for CRC patients.

AI ML-based nomogram for mortality risk stratification in cirrhotic patients

The AI/ML-based nomogram developed in this study serves as a predictive tool for estimating in-hospital mortality risk among cirrhotic patients with sepsis. This retrospective single-center study analyzed data from 264 patients admitted between January 2018 and July 2025, dividing them into Survivor and Non-survivor groups, with 28.4% succumbing to the condition during hospitalization. To ensure robust model performance, the dataset was split into a training set (70%) and validation set (30%). Key predictors of mortality identified included alcoholic cirrhosis, Child-Pugh score, mechanical ventilation, total bilirubin (TBiL), and heart rate (HR). These factors were selected using LASSO regression and further optimized with multivariate logistic regression. A nomogram was constructed to visually quantify individual mortality risks based on the weighted contribution of these predictors. The model demonstrated excellent discrimination with an AUC of 0.81 in the training set and 0.83 in the validation set, while calibration plots showed strong agreement between predicted and observed outcomes. Alcoholic cirrhosis was a significant risk factor, with a 3.7 times higher mortality rate compared to non-alcoholic cases. Higher Child-Pugh scores indicated worsening prognosis, mechanical ventilation doubled mortality risk, and elevated TBiL levels reflected impaired hepatic function. Extreme heart rate abnormalities also correlated with poor outcomes. Infection complications like spontaneous bacterial peritonitis and esophageal variceal bleeding were more frequent among non-survivors. The nomogram provides clinicians with a quantitative tool for mortality risk stratification, enabling early intervention strategies. Its findings align with global trends in cirrhosis-related mortality, emphasizing its clinical relevance and utility in managing high-risk patients effectively.

Deep learning : Predicting HCC surgery success with multimodal imaging

Deep learning is a subset of machine learning that uses artificial neural networks to model and analyze complex data patterns. It is particularly effective in tasks involving large datasets, such as image recognition, natural language processing, and medical diagnostics. Deep learning algorithms excel at automatically extracting meaningful features from raw data, making them highly suitable for applications in healthcare, such as predicting hepatocellular carcinoma (HCC) surgery success. ### Predicting HCC Surgery Success with Multimodal Imaging #### **What is Multimodal Imaging?** Multimodal imaging refers to the integration of multiple imaging techniques to provide comprehensive information about a patient's condition. In the context of HCC, multimodal imaging may involve combining triphasic computed tomography (CT), magnetic resonance imaging (MRI), and other imaging modalities. Each modality provides unique insights into tumor characteristics, vascular involvement, and surrounding tissue conditions. Triphasic CT, for instance, captures images in three phases (arterial, venous, and delayed), highlighting liver vascularization and tumor perfusion patterns. MRI, on the other hand, can provide detailed information about soft tissue and molecular composition. By combining these imaging modalities, clinicians can gain a more holistic understanding of tumor behavior and treatment outcomes. #### **Role of Deep Learning in Predicting HCC Surgery Success** Deep learning models like **Recur-NET** leverage multimodal imaging data to predict the likelihood of HCC recurrence after surgical resection. The integration of multimodal imaging and deep learning offers several advantages: 1. **Automated Feature Extraction**: - Traditional radiomics methods rely on manually defined features, which might miss subtle patterns in imaging data. Deep learning models autonomously extract meaningful features from multimodal imaging data, capturing complex patterns that might escape human detection. - For example, Recur-NET uses a residual neural network (ResNet) architecture to analyze triphasic CT images and identify imaging biomarkers associated with recurrence risk. 2. **Improved Accuracy**: - Deep learning models like Recur-NET have demonstrated high accuracy in predicting microvascular invasion (MVI), a key factor influencing HCC recurrence. The model achieved AUROC (Area Under the Receiver Operating Characteristic) values ranging from 0.77 to 0.86, showcasing its ability to discriminate recurrence risks effectively. 3. **Integration of Clinical Variables**: - In addition to imaging data, Recur-NET incorporates clinical variables such as patient demographics, tumor size, and liver function. This multimodal approach enhances the model's predictive power, as it accounts for both imaging and non-imaging factors affecting surgical outcomes. 4. **Risk Stratification**: - Deep learning models can stratify patients into different risk categories based on recurrence likelihood. This enables personalized post-surgical management, such as tailoring surveillance protocols or additional therapies for high-risk patients. 5. **Precision Medicine**: - By analyzing multimodal imaging data and clinical inputs, deep learning models contribute to precision medicine. They help refine treatment planning, improve surgical outcomes, and reduce recurrence rates, ultimately enhancing patient survival and quality of life. #### **Challenges and Future Directions** While deep learning models like Recur-NET show great promise, there are challenges to overcome for widespread clinical adoption: 1. **Generalizability**: - Recur-NET was trained on data primarily from Asian populations and triphasic CT imaging. Validation in Western populations and the incorporation of MRI-based models are needed to ensure broader applicability. 2. **Infrastructure Requirements**: - Clinical implementation requires robust infrastructure for processing large imaging datasets, integrating multimodal data, and validating models across diverse healthcare systems. 3. **Ethical and Regulatory Considerations**: - The use of AI in healthcare involves addressing ethical concerns, such as data privacy, algorithmic bias, and regulatory approval. #### **Future Implications** The integration of deep learning and multimodal imaging represents a significant advancement in predicting HCC surgery success. Models like Recur-NET exemplify the potential of AI in precision medicine, paving the way for: - Enhanced recurrence prediction. - Personalized treatment planning. - Improved surgical outcomes. - Reduced mortality rates in HCC patients. As AI technologies continue to evolve, deep learning models will play an increasingly vital role in transforming HCC management and advancing healthcare as a whole.

Artificial intelligence-assisted colonoscopy improves adenoma detection rates

Yes, artificial intelligence (AI)-assisted colonoscopy has been shown to improve adenoma detection rates (ADR). Adenoma detection is a critical measure in colonoscopy, as higher ADRs are directly linked to reduced colorectal cancer (CRC) incidence and mortality. Despite advancements, up to 20% of adenomas are still missed due to human limitations in visual detection. A study conducted at Hallym University Dongtan Sacred Heart Hospital evaluated AI-assisted colonoscopy using the SmartEndo CADe system, a deep-learning tool that highlights potential polyps in real time through bounding boxes and auditory alerts. The study included 474 patients in the AI-assisted group and compared them with 474 patients undergoing standard colonoscopy. Propensity score matching ensured both groups were balanced for clinical and procedural characteristics. The AI-assisted colonoscopy significantly improved ADR (35.9% vs. 26.4%, p=0.002), demonstrating its effectiveness in routine practice. It also increased the polyp detection rate (53.2% vs. 46.2%, p=0.038) and the mean adenomas per colonoscopy (0.69 vs. 0.43, p<0.001). AI systems were particularly beneficial for trainees, boosting their ADR from 27.2% to 51.5% (p=0.023). Additionally, AI-assisted colonoscopy proved valuable in detecting subtle, small, or flat lesions, even among high-performing endoscopists. While AI improves detection, challenges such as false positives and increased procedure time persist. However, AI assistance standardizes detection quality, reduces operator variability, and enhances early CRC prevention. Despite limitations like the study’s retrospective nature, AI-assisted colonoscopy represents a significant advancement in improving colonoscopy quality and outcomes.

Esophageal Squamous Cell Carcinoma, Post operative Recurrence and Machine Learning

Esophageal squamous cell carcinoma (ESCC) is a common and aggressive form of cancer, with a high risk of postoperative recurrence. Accurate prediction of recurrence is essential for optimizing follow-up care and tailoring treatment strategies to improve patient outcomes. Traditional models for predicting overall survival in ESCC often lack focus on postoperative recurrence, which is a critical factor affecting long-term survival. A recent study developed and validated a machine learning model based on support vector machines (SVM) to predict postoperative recurrence in ESCC patients. The study analyzed clinical data from 310 patients who underwent surgery, using preoperative, intraoperative, and postoperative variables such as tumor markers, inflammatory markers, TNM stage, and complications. Key risk factors identified included age, ECOG performance status, neutrophil-to-lymphocyte ratio (NLR), C-reactive protein-to-prealbumin ratio (CPR), CY211, TNM stage, and postoperative complications. The best-performing model, SVM6+8, achieved high sensitivity (94% in the test cohort) and strong predictive accuracy. A nomogram based on the SVM6+TNM model was developed to estimate 1-, 3-, and 5-year disease-free survival (DFS). Kaplan-Meier survival analysis showed significantly improved DFS in low-risk patients. This machine learning approach provides clinicians with a reliable tool to stratify patients into high- and low-risk groups, enabling personalized follow-up schedules and adjuvant therapy planning. While the study demonstrated robust internal validation, external validation is needed to confirm its generalizability. Nonetheless, the integration of machine learning with clinical data represents a significant advancement in ESCC management.

Machine learning (ML) in GI Endoscopy

Machine learning (ML) in gastrointestinal (GI) endoscopy is revolutionizing the field by improving diagnostic precision, reducing human variability, and enhancing workflow efficiency. Here is a detailed explanation of the role, applications, challenges, and future outlook for ML in GI endoscopy: ### Key Benefits of ML in GI Endoscopy: 1. **Enhanced Diagnostic Precision**: - ML algorithms, particularly deep learning models, can analyze endoscopic images with greater accuracy than conventional methods. - They excel in detecting, classifying, and predicting gastrointestinal diseases, including polyps, adenomas, and cancers. 2. **Reduction of Human Variability**: - Inter-operator variability in lesion detection and diagnosis is a common challenge in endoscopy. - ML ensures more consistent and standardized results across different clinical settings, leading to improved diagnostic reliability. ### Core Applications of ML in GI Endoscopy: ML is utilized in four primary domains: 1. **Computer-Assisted Detection (CADe)**: - CADe systems use deep convolutional neural networks (CNNs) like ENDO-AID and RetinaNet to identify lesions in real-time during endoscopy. - These systems have improved adenoma detection rates by 7–10%, significantly reducing missed lesions during colonoscopy. 2. **Computer-Assisted Diagnosis (CADx)**: - CADx models help classify lesions and predict their malignancy, aiding in the accurate diagnosis of GI diseases. - Algorithms such as XGBoost and CatBoost have been used to predict outcomes like polyp recurrence and gastrointestinal bleeding risks, outperforming traditional statistical models. 3. **Therapeutic Assistance**: - ML tools guide therapeutic interventions, such as polyp removal or biopsy, by identifying optimal treatment strategies based on real-time data. 4. **Technical Support for Workflow Efficiency**: - ML systems optimize workflow by automating processes like image analysis, annotation, and report generation, reducing the burden on clinicians and improving efficiency in busy endoscopy suites. ### Challenges in ML Implementation: 1. **Heterogeneity of GI Diseases**: - The diversity of gastrointestinal diseases and demographic variations across populations make it difficult for ML algorithms to generalize effectively. - Algorithms trained on limited datasets may perform poorly in diverse clinical settings. 2. **Dataset Quality and Bias**: - Retrospective data is commonly used for training, which can introduce selection and spectrum biases, reducing model reliability in real-world practice. - High-quality, diverse, and annotated datasets are essential for robust ML training. 3. **Annotation Bottleneck**: - Creating large, expert-annotated datasets is time-consuming and expensive. - Tools like Label Studio and Segment Anything Model (SAM) help automate and accelerate the annotation process, improving efficiency. 4. **Image Artefacts and Reflections**: - Artefacts and reflections in endoscopic images can distort model training and predictions. - Techniques like CLAHE (Contrast Limited Adaptive Histogram Equalization), data augmentation, and transformer networks enhance image clarity and model robustness. 5. **Real-Time Integration Limitations**: - Deploying ML systems in real-time during endoscopy faces technical hurdles, including data latency, limited computational resources, and integration challenges in busy clinical environments. 6. **Ethical and Privacy Concerns**: - Compliance with regulations like GDPR ensures ethical data usage, anonymization, informed consent, and transparency in AI-assisted diagnostics. - Algorithmic bias and inequitable outcomes must be addressed to ensure fairness. 7. **Cost–Accuracy Trade-off**: - Using junior endoscopists for data annotation reduces costs but may compromise accuracy. - Combining junior and expert reviewers provides a balanced, cost-effective approach. ### Explainability and Clinician Trust: - **Explainable Artificial Intelligence (XAI)** tools like SHAP (SHapley Additive exPlanations) and saliency maps are crucial for clinician trust. - These methods make ML decisions interpretable, allowing clinicians to understand and act on AI-generated insights. ### Future Outlook: 1. **Integration of Large Language Models (LLMs)**: - The combination of LLMs and computer vision systems is expected to further enhance diagnostic accuracy, workflow efficiency, and patient-centric care in GI endoscopy. 2. **Standardisation and Collaboration**: - Standardized data protocols and interdisciplinary collaboration among clinicians, engineers, and regulators are essential for developing reliable and scalable ML systems. 3. **Clinician Training**: - Endoscopists need specialized training to interpret AI outputs, handle false positives, and integrate ML tools effectively into clinical workflows. 4. **Global Applicability**: - Models trained on diverse, multi-regional datasets can generalize better to global populations, ensuring equitable and effective diagnostic solutions. ### Conclusion: Machine learning is transforming GI endoscopy into a more precise, efficient, and patient-centred diagnostic field. While challenges like data heterogeneity, bias, and real-time integration exist, advancements in technology, collaboration, and ethical practices are paving the way for widespread adoption. With the integration of large language models and computer vision systems, ML has the potential to redefine the future of gastrointestinal healthcare.

AI in health and health care: summary of the JAMA Summit

The JAMA Summit on AI in health and health care highlighted the transformative impact of artificial intelligence on the field, emphasizing its potential to improve access, quality, and affordability of care while addressing its associated challenges and risks. Below is a detailed summary of the key points discussed during the summit: ### **1. Transformative Impact of AI in Health Care** Artificial intelligence is revolutionizing health care at an unprecedented scale. It is reshaping clinical practices, health management, and patient engagement. AI applications span diagnostic tools, administrative operations, and hybrid systems that integrate clinical and business functions. The summit underscored how AI is altering the landscape of health care delivery, with the potential to improve efficiency, accuracy, and outcomes. ### **2. Broad Application Spectrum** AI is being deployed across diverse domains in health care: - **Clinical Decision Support:** Examples include sepsis alerts, diabetic retinopathy screening, and predictive analytics to identify high-risk patients. - **Mobile Health Applications:** AI-powered apps are enhancing patient engagement and self-management of chronic conditions. - **Administrative Tools:** AI is streamlining operations such as scheduling, billing, and documentation, reducing administrative burden and costs. These applications demonstrate AI's deep integration into daily health care operations, making it a critical tool for providers and patients alike. ### **3. Regulatory Gaps** A significant concern raised during the summit was the lack of consistent regulatory oversight for AI tools in health care. Many AI systems operate outside the purview of the FDA, meaning their clinical impact and safety are often unevaluated. This regulatory gap raises concerns about unverified outcomes, patient safety, and the potential for unintended harm. ### **4. Evaluation Challenges** Assessing the effectiveness of AI tools remains a major challenge. The outcomes of these tools depend heavily on factors such as: - The quality of the user interface. - The level of training provided to clinicians and users. - The specific clinical context in which the AI is deployed. Even high-performing algorithms can fail in real-world settings if deployment conditions are suboptimal. This highlights the need for robust evaluation methods to ensure AI tools perform effectively across various environments. ### **5. Safety vs. Effectiveness Focus** Current oversight and compliance efforts primarily monitor **safety issues**, such as detecting errors or “hallucinations” in AI systems. However, there is limited focus on measuring **effectiveness**, which refers to whether AI tools actually improve patient outcomes, care quality, or operational efficiency. The summit emphasized the importance of shifting the focus toward evaluating both safety and effectiveness. ### **6. Multistakeholder Collaboration** The successful deployment of AI in health care requires collaboration among multiple stakeholders: - **Developers:** To design tools that prioritize safety, equity, and clinical value. - **Clinicians:** To ensure AI tools are practical and usable in real-world scenarios. - **Regulators:** To establish standards for safety and effectiveness. - **Health Systems:** To integrate AI into workflows and monitor its performance post-implementation. A lifecycle approach—spanning creation, testing, implementation, and post-market monitoring—was identified as essential for AI’s success in health care. ### **7. Development of Evaluation Infrastructure** The summit stressed the need for standardized measurement and monitoring tools to assess AI effectiveness. These tools would enable rapid, efficient, and evidence-based evaluations across various health care settings, ensuring AI tools meet clinical and operational benchmarks. ### **8. National Data Ecosystem** Establishing a nationally representative data infrastructure was deemed critical for equitable evaluation of AI tools. Such an ecosystem would: - Provide generalizable insights into AI’s health impacts. - Ensure diverse populations and environments are included in evaluations. - Promote equity in AI deployment and outcomes. ### **9. Incentive and Policy Alignment** Effective AI deployment requires aligning incentives and policies to encourage developers and institutions to prioritize: - **Safety:** Ensuring AI tools do not harm patients. - **Equity:** Addressing disparities in access and outcomes. - **Clinical Value:** Focusing on improving care quality and efficiency rather than speed or profit. Policy reforms and market-based incentives were identified as key drivers for fostering responsible AI innovation. ### **10. Future Outlook** The summit concluded with discussions about the future of AI in health care. AI is expected to disrupt every component of health care delivery, presenting both opportunities and risks. Realizing its full potential will depend on creating a robust ecosystem that fosters: - Transparent and evidence-driven innovation. - Equitable access to AI tools and their benefits. - Collaboration among stakeholders to address challenges and optimize outcomes. In summary, while AI offers transformative opportunities for health care, its successful integration requires addressing regulatory gaps, evaluation challenges, and equity concerns. The summit emphasized the need for collaboration, standardized evaluation methods, and policy alignment to ensure AI improves access, quality, and affordability while minimizing risks.

Integrating Machine Learning and Bioinformatics to Develop a Gene-Based Prognostic Model for Gastric Cancer

Integrating machine learning and bioinformatics has proven to be a transformative approach in developing a gene-based prognostic model for gastric cancer. This study utilized transcriptomic and clinical data from The Cancer Genome Atlas (TCGA) and validation datasets from the Gene Expression Omnibus (GEO) to identify genes associated with patient survival outcomes. Advanced machine learning techniques, such as random survival forest and generalized boosted regression modeling, were applied to analyze the data and pinpoint seven key genes—CGB5, FEM1A, MATN3, ZNF101, MARCKS, BRI3BP, and APOD—that are closely linked to prognosis in gastric cancer patients. The identified genes were used to construct a high-precision risk score model capable of predicting patient survival outcomes with accuracy. Validation methods, including Kaplan-Meier survival analysis, receiver operating characteristic (ROC) curves, and Cox regression analysis, confirmed the model’s robustness and demonstrated that the risk score is an independent prognostic factor for gastric cancer. Immunohistochemical analysis further revealed that several hub genes, such as CGB5, MATN3, MARCKS, and APOD, were more highly expressed in cancerous tissues compared to normal tissues, highlighting their correlation with disease pathology. This integration of machine learning with gene expression profiling enables precise risk stratification, paving the way for personalized and targeted treatment strategies in gastric cancer management. The study underscores the potential of combining computational tools and molecular insights to advance cancer prognosis and therapy.

Artificial neural network-predicted PPG and HVPG with measured PPG in decompensated cirrhosis

The study compared the performance of artificial neural network (ANN)-predicted portal pressure gradient (PPG) and hepatic venous pressure gradient (HVPG) with measured PPG in patients with decompensated cirrhosis. Here's a detailed breakdown of the findings: ### **Correlation Results in Group A (Real-World Conditions):** - **HVPG vs. Measured PPG:** HVPG demonstrated negligible correlation with measured PPG (correlation coefficient r = 0.014). This indicates that HVPG is highly unreliable in estimating actual portal pressures in real-world conditions, likely due to the influence of hepatic venous collaterals. - **ANN-Predicted PPG vs. Measured PPG:** ANN-predicted PPG showed moderate correlation with measured PPG (r = 0.437, P < 0.001), suggesting that the ANN model provides a more accurate and consistent estimate of portal pressure compared to HVPG. ### **Agreement Analysis:** - Bland–Altman plots revealed that ANN-predicted PPG had **narrower limits of agreement** with measured PPG (−8.45 to 8.51 mmHg) compared to HVPG (−22.17 to 10.03 mmHg). This signifies that the ANN model is more precise and less prone to large deviations from actual portal pressure values. ### **Group B Findings (Optimized Conditions):** - In Group B, which excluded patients with a high coefficient of variation (>30%) to approximate optimized conditions: - **HVPG vs. Measured PPG:** HVPG showed moderate correlation (r = 0.457). - **ANN-Predicted PPG vs. Measured PPG:** ANN-predicted PPG demonstrated a slightly better correlation (r = 0.476). - This indicates that under optimized conditions, both methods perform comparably in estimating portal pressures. ### **Etiology Subgroup Findings:** - **Hepatitis B–Related Cirrhosis:** The ANN model significantly outperformed HVPG in hepatitis B–related cirrhosis, with a correlation coefficient of r = 0.716 for ANN-predicted PPG vs. r = 0.472 for HVPG. - **Alcohol-Related Cirrhosis:** HVPG performed better than ANN-predicted PPG in alcohol-related cirrhosis. - **Autoimmune Cirrhosis:** Both methods showed weak performance in autoimmune cirrhosis, indicating the need for further refinement in these specific cases. ### **Child–Pugh Subgroup Results:** - **Class A and B Patients (Mild to Moderate Cirrhosis):** - Both HVPG and ANN-predicted PPG correlated moderately with measured PPG, showing comparable accuracy in less severe stages of cirrhosis. - **Class C Patients (Severe Cirrhosis):** - ANN-predicted PPG maintained correlation with measured PPG even in severe cirrhosis cases. - HVPG completely failed to correlate with measured PPG in class C patients, highlighting its unreliability in advanced disease stages. ### **Clinical Relevance:** - ANN-predicted PPG offers a **stable and reliable estimate** of portal pressure in patients with venous collaterals or advanced disease, where HVPG often yields inaccurate results. - In real-world settings (Group A), HVPG significantly **underestimated PPG**, leading to potential clinical misinterpretations of portal pressure severity. In contrast, ANN-predicted values closely matched measured PPG, reducing diagnostic errors. ### **Advantages of ANN Method:** - **Noninvasive:** Unlike HVPG, which requires invasive procedures, the ANN model uses clinical and imaging data to predict PPG. - **Reproducible:** The ANN approach is less affected by procedural variability and anatomical complexities. - **Improved Accuracy:** Especially in complex clinical conditions like severe cirrhosis, the ANN model demonstrated superior agreement and precision compared to HVPG. ### **Limitations:** - The study was retrospective, which limits the statistical strength and generalizability of the findings. - Subgroup sizes were limited, particularly for alcohol-related and autoimmune cirrhosis, which may impact the reliability of results in these etiologies. - The study focused only on patients with decompensated cirrhosis, excluding compensated cases. ### **Conclusion:** While HVPG remains the clinical gold standard, the ANN-predicted PPG demonstrated **superior stability, agreement, and practicality** under complex clinical conditions, making it a promising noninvasive complement for evaluating portal hypertension in cirrhosis patients. Future research should focus on validating the ANN model in prospective, multicenter studies across diverse etiologies and disease stages, including compensated cirrhosis.

Evaluation of NLP Algorithms for Identifying IBD from Clinical Records

The study conducted a comprehensive evaluation of 15 natural language processing (NLP) algorithms to identify patients with inflammatory bowel disease (IBD) from free-text secondary care clinical records. Here are the detailed findings related to the evaluation: ### 1. **Study Objectives** The primary goal was to compare the performance of various NLP algorithms spanning 50 years of evolution, ranging from traditional rule-based models to advanced transformer-based and large language models (LLMs). The evaluation focused on multiple aspects, including accuracy, fairness, cost-efficiency, explainability, and environmental impact. --- ### 2. **Dataset** The study utilized a robust dataset of 9311 labeled clinical documents from 1612 patients. This dataset ensured high diversity and statistical power, allowing for reliable model comparisons. --- ### 3. **Algorithm Scope** The algorithms evaluated included: - **Traditional NLP models**: Rule-based regex, TF-IDF, and Bag-of-Words (BoW). - **Transformer-based models**: BERT, DistilBERT, sBERT. - **Large language models (LLMs)**: DeepSeek, Mistral, Qwen. --- ### 4. **Performance Results** #### **Top Performer** - **DistilBERT_IBD**: Achieved the highest accuracy with a micro F1 score of 93.54%, closely followed by sBERT with a score of 93.05%. However, both models exhibited moderate specificity limitations. #### **LLM Performance** - LLMs like Mistral, Qwen, and DeepSeek performed well without exposure to training data, achieving micro F1 scores between 86.47% and 92.20%. However, they required longer runtimes and higher computational resources. #### **Traditional Models** - Regex and spaCy negation models were fast and cost-effective but had poor specificity, making them unsuitable for precise patient identification. - TF-IDF and BoW demonstrated solid baseline performance (~92% F1) but lacked contextual understanding for complex medical text. #### **Document vs Patient-Level Accuracy** - Document-level accuracy was higher compared to patient-level accuracy due to misalignment between document predictions and overall patient diagnoses, which reduced recall and calibration. --- ### 5. **Fairness Findings** - All models exhibited biases, particularly against women, wealthier individuals, and patients of African ethnicity. - LLMs were more balanced in terms of gender bias but displayed mild age-related bias, overpredicting IBD in older age groups. --- ### 6. **Economic and Environmental Analysis** #### **Cost Efficiency** - Simple models like regex and TF-IDF were the fastest and cheapest, with processing times under 2 minutes and carbon emissions below 2g CO₂. - In contrast, the 70B parameter LLM consumed over 33,000g CO₂, highlighting sustainability concerns for large-scale deployment. #### **Energy Consumption** - Transformer-based and LLM models required significantly more energy (up to 163 kWh) compared to traditional machine learning methods. --- ### 7. **Explainability** - Explainability tools such as SHAP and LIME were used to analyze model decision-making patterns and detect overfitting. - LLMs posed unique challenges for interpretability, requiring the development of new frameworks beyond traditional feature attribution methods. --- ### 8. **Calibration** - sBERT-Base demonstrated the best calibration with the lowest Brier scores, indicating reliable probability predictions compared to other models. --- ### 9. **Document Contribution** - Clinic letters were found to be the most predictive of IBD, with higher odds ratios (22.69–23.38) compared to endoscopy or pathology reports, reflecting their richer contextual information. --- ### 10. **Clinical Implications** The study’s findings have significant implications for healthcare: - **Transformer-based models** like BERT and DistilBERT offer the best trade-off between accuracy, efficiency, and interpretability, making them ideal for clinical NLP deployment. - Open-source availability on platforms like Hugging Face and GitHub allows healthcare institutions to replicate or fine-tune models for their local electronic health record (EHR) systems. --- ### 11. **Future Outlook** The study predicts that once issues related to cost, performance, and bias are addressed, LLMs will become the standard for clinical data retrieval and patient identification within the next decade. --- ### 12. **Study Strengths** The study is notable for its: - Transparency and open-source code. - Multi-metric evaluation encompassing accuracy, fairness, cost, energy consumption, and explainability. - Robust dataset and diverse algorithm comparison, making it a benchmark for clinical NLP validation. --- In summary, the evaluation highlighted the strengths and limitations of different NLP algorithms in identifying IBD from clinical records. While transformer-based models like DistilBERT and sBERT emerged as top performers, challenges such as fairness, environmental impact, and interpretability remain critical areas for improvement.

Large language models (LLMs) with script concordance testing (SCT)

The benchmarking study that evaluated the clinical reasoning performance of large language models (LLMs) using Script Concordance Testing (SCT) provided several insights into their capabilities and limitations in medical applications. Below is a detailed summary of the findings and their implications: --- ### What is Script Concordance Testing (SCT)? SCT is a validated method for assessing clinical reasoning under uncertainty. Unlike multiple-choice questions (MCQs) that focus on knowledge recall, SCT evaluates how new information influences diagnostic or management decisions. It is particularly suited for scenarios where there is no single "correct" answer, but instead a range of plausible responses that depend on context. --- ### Key Findings from the Study #### 1. **Clinical Context and Need for SCT** - Current benchmarks for LLMs in medicine, like licensing exams, primarily assess knowledge recall rather than real-time diagnostic reasoning under uncertainty. - SCT is better aligned with real-world clinical reasoning, making it a valuable tool for evaluating LLMs in medical contexts. #### 2. **Benchmark Scope** - The study created a 750-question SCT benchmark using datasets from 10 international sources spanning specialties like pediatrics, neurology, internal medicine, emergency medicine, surgery, and physiotherapy. - This broad scope ensured that the test covered diverse clinical scenarios and minimized the risk of data leakage. #### 3. **Human vs. Model Comparison** - The SCT performance of LLMs was compared to 1,070 medical students, 193 residents, and 300 attending physicians. - This direct comparison provided a clear understanding of how LLMs measure up to human clinicians at different levels of expertise. #### 4. **Performance of LLMs** - Among the 10 LLMs tested, OpenAI’s o3 model performed best, achieving a score of 67.8%, followed by GPT-4o at 63.9%. - These scores approached, but did not match, the performance of senior residents or attending physicians. - Other models, such as Google’s Gemini 2.5, scored lower (52.1%), and reasoning-optimized models like DeepSeek R1 (55.5%) underperformed expectations. #### 5. **Comparison with Medical Students** - LLMs performed on par with or better than medical students in many cases. - However, they consistently underperformed compared to senior residents and attending physicians, indicating that LLMs still have limitations in advanced clinical reasoning. #### 6. **Overconfidence Bias** - LLMs demonstrated a tendency to choose extreme Likert-scale ratings (+2 or −2) more often than neutral responses (0). - This overconfidence in probability adjustments is a significant deviation from the balanced approach used by expert clinicians. #### 7. **Performance of Reasoning-Optimized Models** - Surprisingly, models fine-tuned for explicit reasoning (e.g., chain-of-thought models) performed worse on SCT. - This suggests that such tuning may impair the models' ability to exercise flexible judgment under uncertainty. #### 8. **Differences in Reasoning Patterns** - Human clinicians tended to use more balanced probability shifts, while LLMs exaggerated belief changes, often missing cases where new information should not alter the decision. - This highlights a key limitation in the probabilistic reasoning of LLMs. #### 9. **Benchmark Validity** - SCT demonstrated a performance gradient consistent with human expertise levels (students < residents < attendings), reinforcing its validity as a measure of clinical reasoning. #### 10. **Advantages of SCT Over MCQs** - SCT emphasizes probabilistic, context-sensitive reasoning, which is critical in clinical practice but often not captured in MCQ-style tests. - This makes SCT a more rigorous and realistic benchmark for evaluating LLMs in medicine. #### 11. **Challenges and Limitations** - SCT is abstracted and does not fully replicate the complexities of real-world clinical reasoning, such as interpersonal dynamics or system-level decision-making. - The scoring system may inflate results for models that mimic conservative responses and penalize models that reason well but deviate from consensus. --- ### Implications for the Use of LLMs in Medicine 1. **Clinical Decision Support** - LLMs show promise as tools for supporting medical decision-making, particularly for medical students and junior clinicians. - However, their overconfidence and inflexibility highlight the need for human oversight to ensure safe deployment. 2. **Training and Refinement** - The study underscores the importance of refining LLM training to improve probabilistic reasoning and reduce overconfidence. - Future efforts should focus on aligning model reasoning more closely with expert clinician patterns. 3. **Benchmarking and Validation** - The public release of the SCT benchmark (https://www.concor.dance) will enable researchers to develop better prompts, improve model training, and integrate LLMs more effectively into clinical workflows. 4. **Limitations in Current Models** - The study highlights that LLMs are not yet ready to independently handle complex clinical reasoning tasks, particularly under uncertainty. - Their performance on SCT suggests that they still lack the nuanced judgment required for advanced medical practice. 5. **Educational Applications** - The weak correlation between SCT and MCQ scores in both humans and models supports the idea that SCT measures reasoning skills beyond factual knowledge. - LLMs could potentially be used as educational tools to help medical students and trainees develop clinical reasoning skills. --- ### Future Directions - **Model Improvement:** Researchers can use the insights from this study to refine LLMs, focusing on probabilistic reasoning and reducing overconfidence. - **Expanded Testing:** Additional benchmarks that incorporate real-world complexities, such as interpersonal and systemic factors, could further evaluate LLM capabilities. - **Integration with Clinicians:** LLMs may be better suited for augmenting, rather than replacing, human decision-making in clinical settings. --- In summary, while LLMs demonstrate promising capabilities in clinical reasoning as evaluated by SCT, they are not yet on par with senior clinicians. Their overconfidence, exaggerated belief changes, and difficulty with nuanced judgment under uncertainty limit their current utility in independent clinical practice. However, the SCT benchmark has provided valuable insights that can guide the development of more robust and reliable medical AI systems.

TRIALSCOPE

**TRIALSCOPE** is a cutting-edge framework designed for clinical trial simulation using real-world data (RWD). It leverages advanced artificial intelligence (AI) and causal inference techniques to extract, clean, and analyze patient data sourced from electronic medical records (EMRs). The framework is specifically designed to overcome the challenges posed by confounding factors, which can otherwise compromise the reliability of data derived from observational studies. ### Key Features and Capabilities: 1. **AI-Powered Data Extraction and Cleaning**: - TRIALSCOPE uses AI to automate the extraction and cleaning of patient data from EMRs, reducing the need for manual intervention and minimizing errors. 2. **Causal Inference**: - The framework applies causal inference methodologies to address confounding variables, ensuring that the insights derived from the data are robust and reliable. 3. **Reproduction of Randomized Trial Results**: - TRIALSCOPE has demonstrated its efficacy by successfully reproducing the results of randomized clinical trials (RCTs) for diseases like lung and pancreatic cancer. This capability validates its accuracy and reliability. 4. **Simulation of Virtual Clinical Trials**: - One of its most innovative features is the ability to simulate "virtual" clinical trials. This enables researchers to test hypotheses, evaluate interventions, and generate evidence without the need for actual patient recruitment or physical trials. 5. **Scalability**: - The framework offers a scalable solution for generating real-world evidence (RWE), making it a valuable tool for researchers and healthcare organizations. 6. **Reduction in Manual Curation**: - By automating much of the data processing and analysis, TRIALSCOPE significantly reduces the reliance on manual data curation, saving time and resources. ### Applications: - **Clinical Research**: Enables researchers to test hypotheses and evaluate treatment efficacy using real-world data. - **Drug Development**: Assists pharmaceutical companies in assessing drug performance and safety in a virtual environment before conducting physical trials. - **Healthcare Policy**: Provides insights for policymakers to make evidence-based decisions using reliable RWE. ### Significance: TRIALSCOPE represents a transformative approach to clinical research by bridging the gap between real-world data and traditional randomized controlled trials. Its ability to generate reliable evidence from observational data has the potential to accelerate medical discoveries, reduce the costs associated with clinical trials, and improve patient outcomes. In summary, TRIALSCOPE is a powerful, scalable framework that combines AI, causal inference, and real-world data to revolutionize how clinical trials are conducted and simulated.

CADe and CRC

Computer-Aided Detection (CADe) and Colorectal Cancer (CRC) are connected through the use of artificial intelligence (AI) technologies to improve the detection of polyps during colonoscopy procedures, which is a critical step in CRC prevention. Here's an in-depth explanation based on the context provided: --- ### **What is CADe?** CADe refers to computer-aided detection systems that use AI algorithms to assist gastroenterologists during colonoscopy procedures. These systems analyze real-time images captured during the procedure to highlight potential polyps or abnormalities in the colon that could be missed by human observation alone. --- ### **Role of CADe in CRC Prevention** Colorectal cancer typically develops from precancerous polyps, such as adenomas or serrated lesions. Early detection and removal of these polyps during colonoscopy can significantly reduce the risk of developing CRC. CADe aims to enhance the detection rate of these polyps, thereby improving the effectiveness of colonoscopy as a preventive measure. --- ### **Evidence from Recent Studies** The American Gastroenterological Association (AGA) reviewed data from 41 randomized controlled trials involving over 32,000 patients to evaluate the effectiveness of CADe systems for CRC prevention. Key findings include: 1. **Increased Adenoma Detection Rate (ADR):** CADe-assisted colonoscopy showed an 8% increase in ADR, which is a critical metric for assessing the quality of colonoscopy procedures. Higher ADRs are associated with reduced risks of interval colorectal cancer (cancer that develops between regular screenings). 2. **Detection of Advanced Adenomas and Serrated Lesions:** CADe demonstrated modest improvements in detecting advanced adenomas and serrated lesions, which are more likely to progress to cancer. 3. **Concerns About Overdiagnosis:** While CADe increases the number of polyps detected, many of these polyps are diminutive and unlikely to progress to cancer. This raises concerns about overdiagnosis, unnecessary follow-up procedures, increased costs, and resource strain. --- ### **Uncertain Long-Term Outcomes** Despite the short-term benefits in polyp detection, it remains unclear whether CADe-assisted colonoscopy leads to a reduction in colorectal cancer incidence or mortality. The available evidence is of "very low certainty," meaning that more robust, long-term studies are needed to establish whether CADe translates into fewer cancers or deaths. --- ### **Challenges and Considerations** 1. **Overdiagnosis and Resource Strain:** Detecting diminutive polyps that are clinically insignificant could lead to unnecessary surveillance and increased healthcare costs. 2. **Cost-Effectiveness:** The financial implications of implementing CADe systems in clinical practice are still uncertain, especially given the modest improvements in detection rates. 3. **Patient-Centered Outcomes:** The impact of CADe on patient satisfaction, anxiety, and overall experience during colonoscopy has not been thoroughly studied. 4. **Implementation Variability:** Some centers already use CADe systems like GI Genius, but adoption varies widely across healthcare settings. --- ### **Future Directions** The AGA highlights the need for: 1. **Better Long-Term Data:** Studies focusing on interval cancer reduction, cost-effectiveness, and patient-centered outcomes are essential to determine the true value of CADe in CRC prevention. 2. **Improved Software:** Future iterations of CADe systems, trained on larger datasets and optimized for detecting clinically significant polyps, may enhance performance and address current limitations. 3. **Dynamic Evidence-Based Practice:** The AGA plans to revisit its recommendations as more robust evidence emerges, underscoring the importance of balancing innovation with proven clinical benefits. --- ### **Current Status of CADe in CRC Prevention** At present, CADe is considered a promising but unproven tool in the fight against colorectal cancer. While it shows potential for improving polyp detection rates, its impact on long-term outcomes like cancer incidence and mortality remains uncertain. Healthcare providers and researchers must continue to evaluate its effectiveness while striving to optimize its use in clinical practice.

AI-Based Computer-Aided Detection in Colonoscopy

AI-based computer-aided detection (CADe) in colonoscopy refers to the use of artificial intelligence (AI) systems to assist healthcare professionals in identifying abnormalities, such as polyps or adenomas, in the colon during a colonoscopy procedure. These systems aim to enhance the accuracy, efficiency, and reliability of colorectal cancer screening and prevention by providing real-time support to endoscopists. ### Key Points About AI-Based CADe in Colonoscopy: #### 1. **Purpose and Importance**: - **Colorectal Cancer Prevention**: Colorectal cancer is one of the leading causes of cancer-related deaths globally. Early detection and removal of precancerous polyps or adenomas during colonoscopy significantly reduce the risk of developing colorectal cancer. - **Addressing Human Limitations**: Even experienced endoscopists can miss lesions during colonoscopy due to factors like fatigue or subtle lesion appearances. AI-based CADe systems aim to reduce these miss rates, improving overall diagnostic accuracy. #### 2. **How AI-Based CADe Works**: - The system uses advanced algorithms, often powered by deep learning and computer vision, to analyze the video feed from the colonoscope in real time. - It highlights suspicious areas or potential lesions (e.g., polyps) on the screen, prompting the endoscopist to investigate further. #### 3. **Consensus-Driven Metrics for Evaluation**: A recent study used a modified Delphi process, involving international experts and industry representatives, to establish standardized metrics for evaluating AI-driven CADe systems in colonoscopy. These metrics aim to ensure consistency, transparency, and clinical value. The six prioritized metrics are: - **Sensitivity**: The ability of the system to correctly identify true positives (e.g., actual polyps or adenomas). - **Independent Validation**: The requirement for independent, external validation of the system's performance to ensure reliability. - **Adenoma Detection Rate (ADR)**: A critical clinical outcome metric that measures the proportion of colonoscopies in which at least one adenoma is detected. - **False-Positive Rate**: The rate at which the system incorrectly flags normal areas as suspicious, which can lead to unnecessary interventions or distractions. - **Latency**: The time delay between the video input and the system's output, which is crucial for real-time usability during live colonoscopy procedures. - **Adenoma Miss Rate (AMR)**: The proportion of adenomas missed by the system, which is an important indicator of its diagnostic accuracy. #### 4. **Benefits of Standardized Metrics**: - **Clinical Performance**: Ensures that the AI systems are effective in improving diagnostic accuracy and reducing adenoma miss rates. - **Reliability**: Independent validation and consistent evaluation criteria build trust in the system's performance. - **Technical Efficiency**: Metrics like latency ensure that the system is practical and user-friendly for real-time use in clinical settings. #### 5. **Impact on Clinical Practice**: - The adoption of these standardized metrics is expected to improve the development, evaluation, and adoption of AI-based CADe systems in colonoscopy. - By enhancing the consistency and transparency of performance evaluation, these metrics will help healthcare providers and policymakers make informed decisions about integrating AI into routine clinical practice. - Ultimately, the use of AI-driven CADe systems has the potential to improve adenoma detection rates, reduce colorectal cancer incidence and mortality, and enhance the overall quality of care. #### 6. **Challenges and Considerations**: - **False Positives**: High false-positive rates can lead to unnecessary interventions and distract endoscopists. - **Implementation Costs**: The cost of integrating AI systems into clinical practice may be a barrier for some healthcare facilities. - **Training and Acceptance**: Endoscopists need to be trained on how to use these systems effectively, and there may be resistance to adopting new technologies. - **Ethical and Legal Concerns**: Issues related to data privacy, liability in case of missed diagnoses, and regulatory approvals need to be addressed. ### Conclusion: AI-based computer-aided detection systems in colonoscopy represent a promising advancement in the field of gastroenterology and colorectal cancer prevention. The establishment of consensus-driven metrics, such as sensitivity, ADR, false-positive rate, and latency, provides a standardized framework for evaluating these systems. As these technologies continue to evolve and gain adoption, they have the potential to significantly enhance the accuracy and efficiency of colonoscopy, ultimately improving patient outcomes and reducing the burden of colorectal cancer worldwide.