Abstract

The gastrointestinal (GI) tract, encompassing the stomach, liver, pancreas, and other digestive organs, is essential for digestion. Due to their prevalence and impact on quality of life, gastrointestinal diseases pose significant global public health challenges. The Global Cancer Observatory (GLOBOCAN), a project of the International Agency for Research on Cancer (IARC), part of the World Health Organization (WHO), reports over 4.2 million new GI cancer cases and 2.2 million deaths in 2020. Early detection and accurate diagnosis are vital for improving outcomes, supervised learning and pre-trained models like Xception, Alexnet, and VGG16, have demonstrated effectiveness; however, a hybrid model that integrates the capabilities of multiple approaches shows superior performance. This research introduces DenseResConcatenation, a hybrid deep‑learning model that combines ResNet50 and DenseNet121 to enhance diagnostic precision in GI disease prediction using endoscopic images. By integrating traditional ML techniques with modern deep learning, the model supports personalized treatment planning and robust classification. Our proposed model achieved a top-1 accuracy of 94.15% and an F1 score of 94.12% on an unseen dataset. The proposed framework significantly improves detection accuracy by leveraging visual patterns in GI‑tract imagery, opening the door to more effective and accurate solutions in endoscopic diagnostics.