Hepatocellular carcinoma (HCC) remains a major cause of cancer-related mortality worldwide, and its management is limited by heterogeneous risk profiles, suboptimal surveillance performance, diagnostic uncertainty in chronically diseased livers, and difficulty individualizing prognosis after treatment. The aim of this narrative review was to critically evaluate artificial intelligence (AI) applications across the HCC care continuum, with emphasis on their intended clinical role, reported performance, evidence maturity, and barriers to implementation. A major strength of this review is that it moves beyond a descriptive catalog of models by structuring the literature around clinically relevant decision points and by explicitly distinguishing emerging proof-of-concept tools from applications with stronger translational potential. Across risk stratification, surveillance, imaging-based diagnosis, pathology, treatment-response prediction, and prognostication, we found that AI consistently demonstrates promise, particularly for identifying patients at higher future HCC risk, improving lesion detection and characterization on ultrasound, CT, MRI, and contrast-enhanced ultrasound, assisting histopathologic classification, and predicting outcomes such as microvascular invasion, recurrence, survival, and response to locoregional therapies. However, we also found that the evidence base remains highly uneven: many diagnostic studies are retrospective and lesion-enriched rather than embedded in true surveillance populations, many prognostic models lack robust external validation and calibration assessment, and reference standards, imaging protocols, and dataset composition vary substantially across studies. These findings are clinically relevant because they highlight both where AI may offer near-term value and why most published systems are not yet ready for routine use. Overall, AI in HCC should be viewed as a rapidly evolving but still transitional field. Its future impact will depend not only on higher-performing algorithms but on clearly defined clinical use cases, multicenter and prospective validation, transparent reporting, workflow-aware evaluation, and implementation strategies that support safe, equitable, and scalable adoption.

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.

Goals: To compare a vision transformer with 2 convolutional neural network architectures for multiclass lesion classification in capsule endoscopy images. Background: Manual review of capsule endoscopy is time-consuming and subject to interobserver variability. Deep learning can automate lesion recognition; however, most prior capsule endoscopy work evaluates a small number of classes, and systematic comparisons between transformer and convolutional architectures across many lesion categories are limited. Study: Two publicly available data sets (SEE-AI and Kvasir-Capsule) were merged and preprocessed to create a 21-class image data set (∼58,000 frames). Images were resized to 224 × 224 pixels and split using stratified sampling into training (n=40,587), validation (n=8696), and test (n=8696) sets. A pretrained Vision Transformer, DenseNet121, and ResNet50 were fine-tuned using categorical cross-entropy loss and Adam optimization, with early stopping. Performance was assessed using accuracy, macroaveraged precision, recall, F1 Score, and the area under the receiver operating characteristic curve. Results: On the independent test set, the vision transformer achieved 92.2% accuracy with macroaveraged precision/recall/F1-score of 0.92 and an area under the receiver operating characteristic curve of 0.99. DenseNet121 achieved 74.0% accuracy (F1-score 0.78; area under the receiver operating characteristic curve 0.85). ResNet50 achieved 38.0% accuracy (F1-score 0.40; area under the receiver operating characteristic curve 0.55). Conclusions: In this merged 21-class capsule endoscopy image data set, the vision transformer achieved higher frame-level classification performance than DenseNet121 and ResNet50 under the present experimental conditions. Importantly, the data set was split at the image level rather than at the patient or procedure level, because frames from the same examination may be correlated; therefore, performance figures likely reflect benchmark results on this frame-level public data set and should not be interpreted as estimates of patient-level generalization or as evidence of definitive architectural superiority. These findings support further evaluation of transformer-based approaches, but grouped reanalysis, external validation, and workflow-oriented studies are required before clinical implementation.

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.