Informatics in Medicine Unlocked最新文献

{"title":"Advancement and challenges of reinforcement learning in lung cancer imaging","authors":"Jamin Rahman Jim ,&nbsp;Nasif Hannan ,&nbsp;Md Apon Riaz Talukder ,&nbsp;M.F. Mridha ,&nbsp;Md Mohsin Kabir","doi":"10.1016/j.imu.2026.101740","DOIUrl":"10.1016/j.imu.2026.101740","url":null,"abstract":"<div><div>Understanding how reinforcement learning can be applied to lung cancer imaging is essential for progress in this field. Despite increasing interest, there is a clear lack of focused survey papers that explore this intersection. To fill this gap, we conducted a comprehensive review that brings together current research across several key areas. We began by briefly outlining the primary forms of lung cancer to provide context for their imaging needs. Next, we explored RL algorithms that have been specifically adapted for lung cancer imaging tasks. We also reviewed widely used datasets and preprocessing techniques, highlighting their importance in building effective RL-based models. Furthermore, we analyzed recent state-of-the-art studies, focusing on their experimental setups, results, and limitations. This helped us map out the current research landscape. In addition, we identified major technical and practical challenges facing the field today. Based on our findings, we proposed several directions for future research that could address these gaps. Overall, this review offers a structured and in-depth overview of RL applications in lung cancer imaging, covering cancer types, RL models, datasets, preprocessing methods, current trends, open issues, and future prospects.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101740"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146075005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Who's afraid of synthetic data? Hybrid approaches to deliver medical digital twins","authors":"Joel Vanin ,&nbsp;Amit Hagar ,&nbsp;James A. Glazier","doi":"10.1016/j.imu.2026.101737","DOIUrl":"10.1016/j.imu.2026.101737","url":null,"abstract":"<div><div>Despite rapidly growing volumes of clinical data, precision medicine still faces a structural data deficit: most patients and rare disease variants are sparsely sampled, labels are noisy, and counterfactual outcomes for alternative treatments are fundamentally unobservable. This position paper argues that overcoming these limits will require hybrid systems that couple multiscale virtual tissue models, synthetic data generation, and AI/ML within risk-aware digital twin frameworks. Using a structured narrative synthesis of three literatures—synthetic health data, virtual tissues and medical digital twins, and hybrid mechanistic–AI architectures including numerical weather prediction—we develop a conceptual framework centered on a mechanistic core linked to AI via forward (mechanistic → synthetic data → AI), backward (AI → mechanistic), and closed (patient-anchored digital twin) loops. We analyze how complex-systems behavior, biological adaptability, and sparse observations bound what medical digital twins can meaningfully predict, motivating ensemble and population-level forecasts rather than exact individual replicas. We then survey emerging implementation patterns, parameter-space exploration methods, and computational envelopes for using virtual tissues to generate biologically constrained synthetic cohorts and to calibrate hybrid digital twins. Finally, we adapt risk- and context-informed verification, validation, and governance frameworks to a four-layer stack spanning mechanistic cores, synthetic data products, AI components, and clinical workflows, with explicit attention to bias, drift, and provenance. We conclude that near-term impact is most likely from population- and cohort-level digital twins that support stratification and short-horizon decision support, while laying the groundwork for more individualized, trustworthy hybrids as biological and methodological uncertainties are better characterized.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101737"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146025397","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AutoLungDx: A hybrid deep learning approach for early lung cancer diagnosis using 3D Res-U-Net, YOLOv5, and vision transformers","authors":"Samiul Based Shuvo ,&nbsp;Tasnia Binte Mamun","doi":"10.1016/j.imu.2026.101739","DOIUrl":"10.1016/j.imu.2026.101739","url":null,"abstract":"<div><h3>Objective:</h3><div>Lung cancer is a leading cause of cancer-related death worldwide, and early detection is crucial to improving patient outcomes. However, early diagnosis is a major challenge, particularly in low-resource settings with limited access to CT resources and trained radiologists. This study aims to propose an automated end-to-end deep learning-based framework for the early detection and classification of lung nodules, specifically for low-resource settings.</div></div><div><h3>Methods:</h3><div>The proposed framework consists of three stages: lung segmentation using the proposed 3D Res-U-Net, nodule detection using YOLO-v5, and classification with a Vision Transformer-based architecture. We evaluated the proposed framework on the publicly available dataset, LUNA16. The performance of the proposed framework was evaluated using task-specific metrics.</div></div><div><h3>Results:</h3><div>The proposed framework achieved a 98.82% lung segmentation dice score while achieving 0.76 mAP@50 for nodule detection from the segmented lung at a low false positive rate. Furthermore, our proposed Vision Transformer network achieved an accuracy of 96.29%, which is 4.25% higher than the state-of-the-art networks. The performance of all three networks in the proposed framework was compared with state-of-the-art studies, and they were found to outperform them across the reported metrics.</div></div><div><h3>Conclusion:</h3><div>Our proposed end-to-end deep learning-based framework can effectively segment the lung and detect and classify lung nodules, particularly in low-resource settings with limited access to radiologists. The proposed framework outperforms existing studies across all respective evaluation metrics.</div></div><div><h3>Significance:</h3><div>The proposed framework can potentially improve the accuracy and efficiency of lung cancer screening, ultimately leading to better patient outcomes, even in settings with limited medical resources and trained personnel.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101739"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170880","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fundus imaging based quantum computing: Grading of severity of vision threatening diabetic retinopathy","authors":"Amna Ikram ,&nbsp;Azhar Imran","doi":"10.1016/j.imu.2026.101744","DOIUrl":"10.1016/j.imu.2026.101744","url":null,"abstract":"<div><div>Diabetic Retinopathy (DR) is a major complication of diabetes that can result in irreversible vision loss if not detected early. Manual ophthalmologic screening is time-consuming and subject to subjectivity, underscoring the need for accurate and automated diagnostic systems. This study proposes QCNET, a hybrid quantum–classical neural network for multiclass grading of Diabetic Retinopathy from retinal fundus images. The model combines the feature extraction capability of a pre-trained DenseNet201 convolutional network with a four-qubit variational quantum layer that acts as a compact, non-linear feature transformation module within an end-to-end learning framework. The quantum-transformed features are subsequently refined through classical dense layers for final classification. To address class imbalance, advanced data augmentation techniques and random oversampling were used. Interpretability of model is enhanced using Gradient-weighted Class Activation Mapping (Grad-CAM), highlighting clinically relevant retinal regions that influence diagnostic decisions. Experimental evaluation on a publicly available dataset demonstrates that QCNET achieves an overall accuracy of 91.0%, precision of 91.7%, recall of 91.0%, and an F1-score of 90.9%. The Matthews Correlation Coefficient and Cohen’s Kappa values of 0.887 and 0.889 indicate a strong agreement with the ground-truth labels. Class-wise ROC analysis yields AUC values ranging from 0.95 to 1.00, confirming robust discriminative performance across disease stages. These results demonstrate the feasibility and potential of integrating variational quantum circuits within deep learning pipelines for medical image analysis, offering a promising and interpretable decision-support tool for the detection of diabetic retinopathy.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101744"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170876","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interpretable deep learning for rotator cuff tear diagnosis: A novel convolutional neural network with Grad-CAM visualization on MRI","authors":"Mohammad Amin Esfandiari ,&nbsp;Iman Ahanian ,&nbsp;Ali Broumandnia ,&nbsp;Nader Jafarnia Dabanloo","doi":"10.1016/j.imu.2026.101733","DOIUrl":"10.1016/j.imu.2026.101733","url":null,"abstract":"<div><div>Accurate diagnosis of rotator cuff tears from magnetic resonance imaging (MRI) is essential for effective clinical management and treatment planning. In this study, we propose a novel convolutional neural network (CNN) architecture specifically designed for classifying rotator cuff tears, integrated with gradient-weighted class activation mapping (Grad-CAM) to provide interpretable insights into the model's decision-making process. We utilized MRI data from 150 subjects, equally divided between normal and pathological cases, and applied data augmentation techniques, including rotation, scaling, and reflection, to enhance model generalization. The proposed CNN demonstrated superior performance, achieving an average accuracy of 94.5 %, sensitivity of 94.6 %, precision of 94.1 %, and specificity of 93.4 %, outperforming established lightweight models such as MobileNetV2 and SqueezeNet. Grad-CAM visualizations confirmed that the model accurately focused on anatomically relevant regions associated with tendon ruptures, thereby enhancing trust in its predictions. These results underscore the potential of our interpretable deep learning framework to deliver reliable, transparent, and clinically actionable diagnostic support for shoulder injuries, paving the way for improved decision-making in orthopedic care. This approach highlights the synergy of advanced CNN design and explainable AI for robust medical imaging applications.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101733"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146075113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Systematic review of transformers, graph neural networks, and federated learning in oncology: Applications, challenges, and pathways to clinical translation","authors":"Sasan Shafiei ,&nbsp;Atousa Saleknezhad ,&nbsp;Erfan Hashemi ,&nbsp;Amin Sedokani ,&nbsp;Emad Gholami ,&nbsp;Ali Parouhan ,&nbsp;Omid Nikoo ,&nbsp;Nikta Taghipour ,&nbsp;Iman Fereydooni ,&nbsp;Elnaz Olama ,&nbsp;Hamid Reza Saeidnia","doi":"10.1016/j.imu.2026.101743","DOIUrl":"10.1016/j.imu.2026.101743","url":null,"abstract":"<div><h3>Background</h3><div>Advanced deep learning models, including transformers, graph neural networks (GNNs), and federated learning (FL), are reshaping oncology by enhancing the analysis of complex, multimodal cancer data for diagnostic and prognostic purposes. Despite their promise, challenges in explainability, generalizability, and clinical readiness hinder widespread adoption.</div></div><div><h3>Methods</h3><div>This systematic review aims to evaluate the applications of transformers, GNNs, and FL in cancer diagnosis and prognosis, identify key technical challenges, and outline strategies for clinical translation. A comprehensive search across PubMed, Scopus, IEEE Xplore, Google Scholar, Web of Science, arXiv, and medRxiv identified 1004 articles, with 16 systematic reviews meeting inclusion criteria after rigorous screening.</div></div><div><h3>Results</h3><div>Data were extracted on model applications, cancer types, data modalities, and technical challenges, and the quality was assessed using the PROBAST tool. Transformers excel in image-based diagnostics, such as skin cancer detection, and multimodal data integration. GNNs are effective for modeling biological networks, aiding in cancer subtype classification and driver gene prediction. FL enables privacy-preserving, multi-institutional model training for cancers like breast, lung, and prostate. Major challenges include limited explainability, poor generalizability across diverse cohorts, data heterogeneity, and insufficient external validation.</div></div><div><h3>Conclusions</h3><div>Transformers, GNNs, and FL offer transformative potential in oncology but face significant hurdles in interpretability and generalizability. Future progress requires interdisciplinary frameworks, transparent benchmarking, and robust validation to ensure clinical trust and impact.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101743"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146170875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact assessment of digital ecosystem in healthcare services: A qualitative case study of hospital data management in Bikaner District in India","authors":"Nikhil Maurya ,&nbsp;Arya Veer Singh Chauhan ,&nbsp;Inder Puri ,&nbsp;Mukesh Kumar Rohil ,&nbsp;Sanjay Kumar Kochar ,&nbsp;Tanmaya Mahapatra","doi":"10.1016/j.imu.2026.101735","DOIUrl":"10.1016/j.imu.2026.101735","url":null,"abstract":"<div><div>The proliferation of digitalization, along with advanced computational techniques, in the healthcare ecosystem has expedited the process of patient care, treatment, and disease diagnosis globally. Medical research, especially involving computational techniques, is heavily dependent on the availability of high-quality datasets generated at the point of care for effective translational research. Our study aims to understand the state of the digital ecosystem (i.e., digitalization, usage of electronic health records (EHRs), and medical data) for the purpose of improving healthcare services and research in hospitals. We conducted a questionnaire-based survey at 16 upper-primary health care centers and public hospitals in the district of Bikaner, Rajasthan, India, to understand the current practices of medical data digitalization and data repository development. The survey results have been analyzed using Principal Component Factor Analysis (PCFA) and statistical tests, including Cronbach's Alpha, the Kaiser-Meyer-Olkin (KMO) measure, and Bartlett's test of sampling adequacy, which indicate that the state of digitalization is in its initial phase. Among technical professionals, 35.6 % agreed that digitalization has been implemented, while 12.3 % remained neutral and 52.1 % disagreed. For the same, 41.4 % agreed, 13.0 % remained neutral, and 45.6 % disagreed among non-technical professionals. These highlight that almost half of the groups recognize slow progress in this area, implying that digitalization is still in its initial phase. Our study also indicates that the lack of access to structured and semi-structured medical datasets is a key barrier to applying Artificial Intelligence (AI) and Machine Learning (ML) in Indian healthcare research, where these technologies could play a crucial role in improving healthcare diagnostics, outcome prediction, and enhancing clinical decision-making, for better healthcare services, esp. in resource-constrained settings.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101735"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145969414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of a neural network system for predicting in-hospital pressure ulcers in spinal trauma patients","authors":"Rezvan Raziee ,&nbsp;Mohsen Sadeghi-Naini ,&nbsp;Zahra Azadmanjir","doi":"10.1016/j.imu.2026.101742","DOIUrl":"10.1016/j.imu.2026.101742","url":null,"abstract":"<div><h3>Background and objectives</h3><div>This study aimed to developing and to deploying an optimal machine learning model to predict pressure ulcers (PUs) in hospitalized patients with spinal fractures, using data from the National Spinal Cord and Column Injury Registry of Iran (NSCIR-IR).</div></div><div><h3>Methods</h3><div>Data were from 4326 patients with traumatic spinal fractures. The preprocessing phase was included handling missing values, feature engineering, normalization, and addressing class imbalance (with a 3.4 % PU incidence) using the Synthetic Minority Oversampling Technique (SMOTE). Feature selection was carried out with univariate filtering methods such as ANOVA and chi-square tests, along with the random forest feature importance algorithm. Six traditional machine learning (ML) algorithms and six ensemble models were trained and evaluated using metrics such as the area under the receiver operating characteristic curve (AUC), recall, and Brier score.</div></div><div><h3>Results</h3><div>The multilayer perceptron neural network (MLP) emerged as the top-performing model, offers advantages for clinical use due to a higher AUC of 0.888 (0.85–0.92), a balanced accuracy, a good recall, calibration, and an acceptable net benefit on the decision curve. Key predictors identified included the ASIA Impairment Scale, the Glasgow Coma Scale score, SCI type, SaO2, and the number of damaged vertebrae. Shapley Additive Explanations (SHAP) analysis further highlighted the directional influence of these factors on PU risk.</div></div><div><h3>Conclusion</h3><div>The MLP model effectively predicts PU in patients with spinal fractures, outperforming other algorithms. Identified predictors align with clinical insights, are emphasizing the need for targeted preventive measures in hospitals. However, external validation with a larger multicenter cohort is recommended to confirm and to expand upon these findings.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101742"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146075003","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Epic overhaul at a Canadian hospital: Pre-Post evaluation insights from physicians and medical residents","authors":"Mirou Jaana ,&nbsp;Edward Riachy ,&nbsp;Erika MacPhee ,&nbsp;Heather Sherrard","doi":"10.1016/j.imu.2025.101725","DOIUrl":"10.1016/j.imu.2025.101725","url":null,"abstract":"<div><h3>Background</h3><div>The implementation of Electronic Medical Records (EMRs) in hospitals offers potential benefits but often disrupts clinician workflows, affecting care delivery and outcomes.</div></div><div><h3>Objective</h3><div>This study evaluates physicians' and medical residents’ perspectives on the impacts of introducing a new Epic system at a Canadian academic hospital.</div></div><div><h3>Methods</h3><div>A pre-post evaluation design was conducted using physician and resident surveys before (T0) and 4- and 9-months post-implementation (T1, T2) that assessed technology use, satisfaction with training and system use, and EMR's perceived impact on care delivery, work practices and quality.</div></div><div><h3>Results</h3><div>Satisfaction with training and system use declined for both groups in the first four months (more sharply for residents) but several measures improved at T2 as users readjusted to the system. There was a significant increase in physicians’ daily computer use (4 h at T0 to 6 h at T1; P &lt; .<em>001</em>). Limited early benefits of the Epic system were observed and a decline in perceived improvement in clinical documentation (P = .<em>006</em> and <em>.0012)</em>, order entry (P = .<em>018</em> and <em>.002)</em> and patient safety (P = .<em>044</em> and <em>.024)</em> were reported at T1 for physicians and residents, respectively. Although some medical practice/work indicators improved by 9 months for physicians, the changes were not statistically significant; these benefits were not observed for residents at T2. Medical training was not significantly affected by the new Epic system either immediately or later post implementation. At T1, 83% of physicians reported that the new system sometimes or often improved the quality of care, as opposed to only 33% of residents; no significant improvements were noted at 9 months post implementation by both groups.</div></div><div><h3>Conclusions</h3><div>Physicians and residents adapt differently to Epic and full system assimilation does not happen in one year. Early perceptions of Epic do not reflect its long-term potential, and meaningful benefits require prolonged stabilization periods for user satisfaction and efficiency gains. We caution hospital leaders not to rely heavily on a vendor-driven implementation, and recommend tailored training, rapid-cycle improvements, transparent communication, and monitoring of agreed-upon performance indicators to strengthen clinician engagement and support long-term success.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101725"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146075004","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Clinical validation of artificial intelligence for gastrointestinal diseases","authors":"Marjan Talebi ,&nbsp;Negar Bozorgchami ,&nbsp;Gauransh Mishra ,&nbsp;Gaurav Mishra ,&nbsp;Rouzbeh Almasi Ghale","doi":"10.1016/j.imu.2026.101736","DOIUrl":"10.1016/j.imu.2026.101736","url":null,"abstract":"<div><div>Artificial intelligence (AI), particularly deep convolutional neural networks (DCNNs), has demonstrated significant potential for transforming the diagnosis and management of gastrointestinal (GI) diseases. This review critically examines the evolving landscape of AI applications in gastroenterology, moving beyond a simple catalog of tools to analyze their pathway to clinical integration. We synthesize current evidence across key functional domains including computer-aided detection (CADe), computer-aided diagnosis (CADx), and predictive outcome modeling highlighting performance metrics and early clinical adoption. Crucially, we identify a pronounced translational gap between technical validation and demonstrable improvement in patient-centered outcomes. The narrative underscores that while AI systems show high diagnostic accuracy in controlled studies, their ultimate clinical utility remains unproven. The conclusion distills core challenges including the need for rigorous multicenter randomized trials, solutions for algorithmic generalizability, and effective human-AI collaboration and emphasizes the urgent imperative for structured clinical validation frameworks to realize AI's promise in routine GI care. We further synthesize evidence by validation stage and study design, highlighting clinical endpoints such as ADR, APC, and complication rates, with GI Genius and ENDOANGEL exemplifying the gap between technical metrics and patient outcomes.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"61 ","pages":"Article 101736"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146025398","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}