Development of the Gastro-AI Model for Diagnosing Gastrointestinal Diseases: Based on the Hyperkvasir

Gastrointestinal diseases, including conditions such as polyps, inflammation, ulcers, and cancer, pose a significant global health challenge, often leading to severe complications when not detected early. This study presents the development and evaluation of a deep learning-based model, named Gastro-AI, designed to enhance diagnostic accuracy using the HyperKvasir dataset-a comprehensive collection of 10,662 endoscopic images representing 23 distinct disease categories. The model leverages the ResNet-50 architecture, a type of convolutional neural network pre-trained on a large image database, and employs transfer learning to adapt to medical imaging tasks. Trained on 8,529 images and validated on 2,133 images, Gastro-AI achieved an overall accuracy of 90.7 percent, with a sensitivity of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 9. 2}$</tex> percent for detecting diseased cases and a specificity of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 2. 1}$</tex> percent for identifying healthy tissues. The area under the receiver operating characteristic curve reached 0.93, indicating strong discriminative ability, while the model processed each image in 0.18 seconds, offering a significant time advantage over manual analysis. Notably, it excelled in detecting polyps (94.2 percent accuracy) and cancer (92.7 percent), though performance dipped to 87.8 percent for inflammation due to image noise challenges. These results underscore Gastro-AI's potential as a rapid and reliable diagnostic tool for clinical settings. Future improvements, including dataset expansion and advanced image preprocessing techniques, are proposed to address limitations and enhance its applicability in real-world healthcare environments. This research contributes to the growing field of artificial intelligence-assisted diagnostics, promising improved early detection and patient outcomes for gastrointestinal diseases.