Construction of gastroscope image recognition model and diagnosis system based on artificial intelligence technology

Abstract

In recent years, AI (Artificial Intelligence) technology has been applied to the diagnosis of diseases such as lung cancer, skin cancer and diabetic retinopathy, and has shown excellent diagnostic performance. AI technology with deep learning as the core can make full use of massive data, automatically learn the features in the data, and assist doctors in diagnosis accurately and quickly. Gastric cancer has a high degree of malignancy, hidden onset and no specific symptoms in the early stage. Its clinical manifestations are often similar to benign gastric diseases such as gastric ulcer and chronic gastritis, and it is easy to be ignored. In this paper, the gastroscope image recognition model and diagnosis system based on AI technology are developed. In this study, a gastroscope image recognition model based on GCN (Graph Convolutional Networks) is proposed. The GCN multi-label classification module is responsible for learning and representing the physiological and anatomical relationship of the upper digestive tract, and fusing the learning results with the output results of the detection module. Finally, a gastroscope image aided diagnosis system is designed and implemented. The results showed that the sensitivity and PPV of the model were 95.360% and 91.017%, respectively. It is suggested that the model is superior to the traditional method in the detection rate of early cancer lesions.