Artificial Intelligence in Gastrointestinal Disorders: Current Updates from Theory to Clinical Practice

Introduction: Colorectal cancer remains a leading cause of cancer-related morbidity and mortality, with adenomatous polyps representing a common precursor. Post-polypectomy polyp recurrence represents a significant risk of colorectal cancer, driving periodic colonoscopy surveillance and polypectomy as needed. In this study, we explore a multimodal machine learning approach that integrates endoscopic imaging with clinical and pathology data to improve recurrence risk prediction and support individualized surveillance planning. Methods: We developed and evaluated a multimodal artificial intelligence (AI) model to predict post-polypectomy colorectal polyp recurrence using the ERCPMP-v5 dataset. The cohort included 217 patients with 796 high-resolution endoscopic RGB images and 21 endoscopic videos; video data were converted to still frames at 2 frames per second. Images and frames were resized to 224 × 224 pixels and normalized. Patient-level demographic, morphological (Paris, Kudo Pit, JNET), anatomical, and pathological variables were encoded using standard scaling for continuous features and one-hot encoding for categorical features. Visual representations were extracted using a pretrained Vision Transformer backbone (ViT-Base-Patch16-224) with frozen weights. Structured metadata (79 variables) was encoded using a multilayer perceptron. A late fusion framework used image and metadata representations to generate a recurrence probability via a sigmoid classifier; probabilities were thresholded at 0.5 for binary prediction. Model performance was evaluated on a held-out test set using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC). We additionally compared fusion performance with image-only and metadata-only baselines. Predicted probabilities were translated to surveillance recommendations using risk tiers: low risk (0.00 ≤ p < 0.20), moderate risk (0.20 ≤ p < 0.50), and high risk (p ≥ 0.50). Results: On the test set, the multimodal fusion model achieved 90.4% accuracy, 86.7% precision, 83.1% recall, 84.9% F1-score, and an AUC of 0.920. The image-only model achieved 84.6% accuracy (AUC 0.880), and the metadata-only model achieved 81.9% accuracy (AUC 0.850), indicating improved performance with multimodal fusion. Risk stratification enabled surveillance recommendations of 1–3 years for low risk, 6–12 months for moderate risk, and 3–6 months for high risk. Conclusions: A late-fusion multimodal model integrating endoscopic imaging with structured clinical and pathology variables demonstrated excellent performance for predicting post-polypectomy recurrence and generated actionable risk-based surveillance intervals. This approach may support individualized follow-up planning and more efficient allocation of surveillance resources, while prioritizing timely evaluation for patients at higher predicted risk. Full article

Background: Current artificial intelligence (AI) systems in gastrointestinal (GI) endoscopy primarily emphasize binary detection or static classification, providing limited support for the graded assessment of malignant potential that underpins clinical decision-making. We developed GastroMalign, a transformer-based framework designed to stratify GI lesions according to ordinal disease severity while maintaining clinical interpretability, addressing this unmet need in endoscopic risk assessment. Methods: This retrospective development and validation study used the publicly available GastroVision dataset, comprising 8000 de-identified endoscopic still images from the upper and lower gastrointestinal tract, including the esophagus, stomach, duodenum, colon, rectum, and terminal ileum. GastroMalign integrates a Vision Transformer (ViT) encoder with a Sequential Feature Learner that explicitly models ordinal disease severity along a benign-to-malignant spectrum. The framework produces both categorical risk classification and a continuous malignancy risk score. Images were stratified into training (80%), validation (10%), and test (10%) sets. Performance was compared with convolutional neural network (CNN) baselines and a Swin Transformer. Interpretability was assessed using Score-CAM visualizations reviewed by blinded expert endoscopists. Results: On the held-out test set (n = 800 images), GastroMalign achieved an overall accuracy of 80.06%, precision of 79.65%, recall of 80.06%, and F1-score of 79.17%, with a micro-averaged AUC of 0.98. In comparison, ResNet-50 and DenseNet-121 achieved accuracies of 32.42% and 36.77%, respectively, while the Swin Transformer achieved 60.56% accuracy (AUC = 0.93). Ablation analyses demonstrated a 17% absolute reduction in High-Risk lesion recall when the progression-aware module was removed. Continuous malignancy risk scores increased monotonically across ordinal classes, with mean values < 0.18 for Benign and >0.72 for High-Risk/Malignant lesions. Score-CAM visualizations demonstrated 92% overlap with clinician-annotated lesion regions. Conclusions: GastroMalign delivers an interpretable, progression-aware AI framework for GI lesion risk stratification that outperforms existing CNN- and transformer-based models. Clinically, GastroMalign is intended as an adjunct decision-support tool during endoscopic review to standardize lesion risk stratification (benign to malignant spectrum), support management decisions (biopsy vs. resection vs. surveillance), and reduce operator-dependent variability by pairing ordinal risk outputs with interpretable visual explanations. Full article

Background/Objectives: Large language models (LLMs) have shown promising results in medical decision support; however, their effectiveness in managing acute cholecystitis and other gallbladder diseases remains insufficiently examined. This study evaluated the performance of a neuro-symbolic LLM system that integrates multiple AI agents with neural–symbolic reasoning for acute cholecystitis management and compared its diagnostic accuracy with that of human expert physicians across three clinical specialties. Methods: This multi-center cross-sectional study included 30 case-based questions covering acute cholecystitis and gallbladder diseases, stratified across eight predefined disease categories: acute calculous cholecystitis (n = 6), acute acalculous cholecystitis (n = 2), complicated cholecystitis including gangrenous, emphysematous, and perforated variants (n = 5), chronic cholecystitis and biliary colic (n = 4), gallbladder polyps and adenomyomatosis (n = 3), Mirizzi syndrome (n = 2), gallbladder carcinoma (n = 4), and post-cholecystectomy complications (n = 4). Questions were categorized into diagnosis (n = 10), treatment (n = 10), and complications/prognosis (n = 10). Gold standard answers were established through consensus by an expert panel consisting of two senior general surgery expert clinicians and one senior emergency medicine expert clinician, each with more than 20 years of clinical experience, utilizing the Tokyo Guidelines 2018 (TG18) as the reference standard for diagnostic criteria, severity grading, and management recommendations. The expert panel achieved unanimous consensus on all 30 gold standard answers. All responses were cross-referenced against the primary TG18 publications to ensure guideline-based rather than solely opinion-based reference standards. This consensus-based, guideline-anchored approach is consistent with established methodologies for gold standard establishment in AI diagnostic accuracy studies. Performance of a neuro-symbolic LLM system orchestrated via LangGraph v1.0 was compared against 10 general surgery specialists, 10 emergency medicine physicians, and 10 gastroenterology specialists from four tertiary centers in Turkey. The neuro-symbolic system incorporated the Tokyo Guidelines 2018 (TG18) as its symbolic knowledge base for diagnostic criteria, severity grading, and management algorithms. Results: The neuro-symbolic system attained the highest overall accuracy rate of 96.7% (29/30), markedly surpassing the performance of general surgery specialists (average 82.3% ± 6.8%), emergency medicine physicians (average 71.0% ± 8.2%), and gastroenterology specialists (average 78.7% ± 7.4%). Furthermore, the neuro-symbolic system exhibited superior performance across all clinical categories. Among human participants, general surgeons showed the highest accuracy in treatment decisions (88.0%), while gastroenterologists excelled in diagnostic questions (82.0%). Emergency medicine physicians showed comparable performance to other specialties in acute presentation scenarios. ROC analysis revealed excellent discrimination for the neuro-symbolic system (AUC = 0.983) compared to general surgery (AUC = 0.856), gastroenterology (AUC = 0.821), and emergency medicine (AUC = 0.764). Conclusions: The neuro-symbolic LLM system exhibited superior performance in standardized guideline-concordant case-based assessment of acute cholecystitis management compared to all human expert groups, reflecting its consistent application of encoded guideline criteria. These findings support its potential role as a clinical decision-support tool that augments, rather than replaces, physician expertise. The system’s consistent application of standardized guidelines indicates its potential utility as a clinical decision support tool, particularly in settings where specialist expertise is limited. However, these results should be interpreted within the constraints of a structured case-based evaluation and do not imply global clinical superiority over human experts. Full article

Background/Objectives: Vision–language models such as BiomedCLIP are increasingly investigated for their diagnostic potential in medical imaging. Although these foundation models show promise in general radiographic interpretation, their application in pediatric domains—particularly for subtle, postoperative findings like esophageal strictures—remains underexplored. This study aimed to evaluate the diagnostic performance of BiomedCLIP in classifying pediatric esophageal radiographs into three clinically relevant categories: presence of contrast agent, full esophageal visibility, and presence of esophageal stricture. Methods: We retrospectively analyzed 143 pediatric esophageal X-rays collected between 2021 and 2025. Each image was annotated by two pediatric radiology experts and categorized according to esophageal visibility, contrast presence, and stricture occurrence. BiomedCLIP was used in a zero-shot classification setup without fine-tuning. Model predictions were converted into binary outcomes and assessed against the ground truth using a comprehensive suite of 27 performance metrics, including accuracy, sensitivity, specificity, F1-score, AUC, and calibration analyses. Results: BiomedCLIP achieved high precision (88.7%) and a favorable AUC (85.4%) in detecting contrast agent presence, though specificity remained low (20%), leading to a high false-positive rate. The model correctly identified all cases of non-visible esophagus, but was untestable in predicting full visibility due to the absence of positive cases. Critically, its performance in detecting esophageal strictures was poor, with accuracy at 24%, sensitivity at 44%, specificity at 18%, and AUC of 0.26. Statistical overlap between contrast and stricture predictions indicated a lack of semantic differentiation within the model’s latent space. Conclusions: BiomedCLIP shows potential in detecting high-salience features such as contrast but fails to reliably identify esophageal strictures. Limitations include class imbalance, absence of fine-tuning, and architectural constraints in recognizing subtle morphologic abnormalities. These findings emphasize the need for domain-specific adaptation of foundation models before clinical implementation in pediatric radiology. Full article

Background/Objectives: Functional lumen imaging probe (FLIP) panometry allows real-time assessment of the esophagogastric junction opening and esophageal body contractile activity during an endoscopic procedure. Despite the development of the Dallas Consensus, FLIP panometry analysis remains complex. Artificial intelligence (AI) models have proven their benefit in high-resolution esophageal manometry; however, data on their role in FLIP panometry are scarce. This study aims to develop an AI model for automatic classification of motility patterns during a FLIP panometry exam. Methods: A total of 105 exams from five centers from both the European and American continents were included. Several machine learning models were trained and evaluated for detection of FLIP panometry patterns. Each exam was classified with an expert consensus-based decision according to the Dallas Consensus, with division into a training and testing dataset in a patient-split design. Models’ performance was evaluated through their accuracy and area under the receiver-operating characteristic curve (AUC-ROC). Results: Pathological planimetry patterns were identified by an AdaBoost Classifier with 84.9% accuracy and a mean AUC-ROC of 0.92. Random Forest identified disorders of the esophagogastric junction opening with 86.7% accuracy and an AUC-ROC of 0.973. The Gradient Boosting Classifier identified disorders of the contractile response with 86.0% accuracy and an AUC-ROC of 0.933. Conclusions: In this study, integrating exams with different probe sizes and demographic contexts, a machine learning model accurately classified FLIP panometry exams according to the Dallas Consensus. AI-driven FLIP panometry could revolutionize the approach to this exam during an endoscopic procedure, optimizing exam accuracy, standardization, and accessibility, and transforming patient management. Full article