AI-Driven Early Detection of Gastric Ulcers Using High-Resolution Imaging and Multi-Scale Radiomic Feature Analysis

Gastric ulcers are a major gastrointestinal disorder whose early diagnosis remains challenging because conventional endoscopic assessment is subjective and prone to inter-observer variability. Recent advances in artificial intelligence (AI) and medical image analysis provide promising opportunities for improving diagnostic accuracy and enabling reliable early detection. This study proposes an AI-driven framework for early gastric ulcer detection using high-resolution endoscopic imaging and multi-scale radiomic feature analysis. The proposed system integrates deep learning–based hierarchical feature extraction with handcrafted radiomic descriptors to capture both global semantic information and subtle mucosal abnormalities associated with ulcerative lesions. Convolutional neural networks (CNNs) are employed to learn discriminative visual representations, while multi-scale radiomic analysis quantifies texture heterogeneity, intensity variation, and morphological irregularities. An attention-guided fusion mechanism is further introduced to adaptively combine deep and radiomic features according to their diagnostic relevance. Extensive experiments were conducted using a curated dataset of high-resolution gastric endoscopic images with preprocessing, augmentation, and cross-validation strategies. The proposed framework outperformed radiomics-only, deep learning–only, and conventional hybrid baseline models across multiple evaluation metrics, including accuracy, sensitivity, specificity, F1-score, and area under the curve (AUC). Ablation analysis additionally confirmed the importance of multi-scale radiomics and attention-based fusion in improving diagnostic robustness and generalization. The results demonstrate that combining deep learning with multi-scale radiomic analysis provides a robust and interpretable solution for early gastric ulcer detection. The proposed framework shows strong potential as a clinical decision-support system capable of improving diagnostic consistency, facilitating early intervention, and enhancing patient outcomes.